CN102541821A - Displaying method and system with notes for interactive multiple languages - Google Patents

Displaying method and system with notes for interactive multiple languages Download PDF

Info

Publication number
CN102541821A
CN102541821A CN2011101439898A CN201110143989A CN102541821A CN 102541821 A CN102541821 A CN 102541821A CN 2011101439898 A CN2011101439898 A CN 2011101439898A CN 201110143989 A CN201110143989 A CN 201110143989A CN 102541821 A CN102541821 A CN 102541821A
Authority
CN
China
Prior art keywords
phrase
sentence
language
chinese
multilingual
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN2011101439898A
Other languages
Chinese (zh)
Other versions
CN102541821B (en
Inventor
马特·斯科特
魏颢
徐维江
陈刚
周明
李志灏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Microsoft Technology Licensing LLC
Original Assignee
Microsoft Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Microsoft Corp filed Critical Microsoft Corp
Priority to CN201110143989.8A priority Critical patent/CN102541821B/en
Priority claimed from CN201110143989.8A external-priority patent/CN102541821B/en
Publication of CN102541821A publication Critical patent/CN102541821A/en
Application granted granted Critical
Publication of CN102541821B publication Critical patent/CN102541821B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Abstract

The invention discloses a displaying method and a system with notes for interactive multiple languages. First a sentence is displayed simultaneously in multiple languages through a user interface, the sentence in each language is segmented into phrases or word groups of the language, relevance is built among the phrases or the word groups in multiple languages, the principle for the relevance is that the relevant phrases or the word groups express the same semantics in the sentences, the sentence displayed in each language is generated and a note is displayed, and the note is segmented similarly and is relevant with the phrases or the word groups. A selection instruction is received to select part of the sentence displayed in some language. Due to the fact that the phrases or the word groups serve as processing units, the phrases or the word groups which the portions selected by the selection instruction belong to are searched for, and the phrases or the word groups are displayed in a highlighting mode. The relevant phrases or the word groups are search for in the sentences displayed in other languages and the corresponding notes and are displayed in a highlighting mode.

Description

The band note display method and the system of interactive multilingual language
Technical field
The present invention relates to the machine translation technology, more particularly, relate to the band note display method and the system of the multilingual language of a kind of interactive mode.
Background technology
Language is the human main tool that exchanges, globalization make deeply that further the interchange between the various language is day by day frequent.Learning foreign language is a kind of new trend, and in the process of learning foreign language, requisite link is with translating mutually between foreign language and the mother tongue of oneself, so that understand the implication of foreign language.
The appearance of machine translation technology makes this work become convenient, and computing machine utilizes the characteristic of its high-speed cruising and huge memory space to remedy human brain deficiency in these areas, for learning foreign language provides huge facility.
Summary of the invention
The present invention is intended to propose a kind of technology by the computer-aided learning (CAL) foreign language, helps user's understanding and uses foreign language through the content that shows the band note.
According to an embodiment, the band note display method of the multilingual language of a kind of interactive mode is proposed, at first showing a sentence simultaneously with multilingual through user interface.Each language sentence is cut apart, be divided into the phrase or the phrase of this kind language.Between multilingual phrase or phrase, set up association, related principle is that identical semanteme expressed in related phrase or phrase in sentence.For the sentence generation that shows with each language and show a note, note is cut apart equally and is associated with phrase or phrase.Receive selection instruction and select a wherein part with this sentence of wherein a kind of language demonstration.Since be with phrase or phrase as processing unit, search phrase or phrase under the part that selection instruction selects, with this phrase or phrase is outstanding shows.In the sentence that shows with other language and in the corresponding note, search phrase or the phrase that is associated, demonstration given prominence in these phrases that are associated or phrase.
According to one embodiment of the invention, the band note display system of the multilingual language of a kind of interactive mode is proposed, this system comprises user interface, statement sectioning, note generation device, associated apparatus and display device.User interface shows a sentence simultaneously with multilingual, and the reception selection instruction is selected the wherein part with this sentence of wherein a kind of language demonstration.The statement sectioning is cut apart this sentence that shows with each language, sentence is divided into the phrase or the phrase of this kind language.The note generation device is for the sentence generation that shows with each language and show a note, and note is cut apart according to phrase or phrase equally.Associated apparatus is set up related between multilingual phrase or phrase and note, and identical semanteme expressed in phrase that is associated or phrase in sentence.Display device is searched affiliated phrase or the phrase of part that selection instruction is selected; With this phrase or phrase outstanding demonstration on user interface; Display device is further searched phrase or the phrase that is associated through associated apparatus in the sentence that shows with other language and in the corresponding note, to these phrases that are associated or phrase outstanding demonstration on user interface.
According to one embodiment of the invention, disclosed the band note display method of the multilingual language of a kind of interactive mode, at first show a sentence simultaneously with Chinese and English through user interface.Centering sentence and English sentence are cut apart, and will be divided into Chinese phrase or phrase and english phrase or phrase.At Chinese phrase or phrase, and set up association between english phrase or the phrase, identical semanteme expressed in Chinese phrase that is associated or phrase and english phrase or phrase.According to Chinese sentence generation and show a pinyin sequence, pinyin sequence is cut apart according to Chinese phrase or phrase and is formed the phonetic segmentation, and the phonetic segmentation is associated with Chinese phrase or phrase.The selection instruction that receives the user is selected the wherein part with this sentence of Chinese or English demonstration.Search affiliated phrase or the phrase of part that selection instruction is selected, with this phrase or the outstanding demonstration of phrase.In the sentence that shows with another kind of language and search phrase or phrase and the phonetic segmentation that is associated in the pinyin sequence, demonstration given prominence in these phrases that are associated or phrase and phonetic segmentation.
Description of drawings
Above-mentioned and other characteristic, character and advantage of the present invention will become more obvious through the description below in conjunction with accompanying drawing and embodiment, in the accompanying drawings, identical Reference numeral is represented identical characteristic all the time, wherein:
Fig. 1 has disclosed the process flow diagram according to the band note display method of the interactive multilingual language of one embodiment of the invention.
Fig. 2 has disclosed the structural drawing according to the band note display system of the interactive multilingual language of one embodiment of the invention.
Fig. 3 has disclosed the process flow diagram of the band note display method of interactive multilingual language according to another embodiment of the present invention.
Fig. 4 has disclosed a display interface according to the band note display technique of the interactive multilingual language of one embodiment of the invention.
Fig. 5 has disclosed the example calculation foundation structure of the technology of the multilingual words aligning that the present invention utilizes.
Fig. 6 A-C has disclosed the graphical representation of exemplary of the words aligning user interface (UI) of the words aligning technology that the present invention utilizes.
Fig. 7 A-D is the graphical representation of exemplary of other words aligning UI, and UI shown in these realizes some improvement of words aligning technology or changes.
Fig. 8 has disclosed the data stream of foundation structure of the words aligning technology of utilization of the present invention and has represented.
Fig. 9-the 12nd, the process flow diagram of the example process of realization words aligning technology.
Embodiment
The Chinese phonetic alphabet is Chinese important component part, has learned the pronunciation that Chinese character just can be understood in the Chinese phonetic alphabet, so, provide the Chinese phonetic alphabet of Chinese can help the study of the user of the non-Chinese of mother tongue to a great extent in the process of underway translator of English simultaneously to Chinese.It is that the foreign user of learning Chinese tailors that phonetic phonetic notation function is provided, and the person that can help the Chinese studying better is to the grasp of relevant phonetic transcriptions of Chinese characters, and the time and efforts that significantly reduces the user and spent.
With reference to shown in Figure 1, disclosed band note display method 100 according to the interactive multilingual language of one embodiment of the invention.
In step 102, show a sentence simultaneously with multilingual through user interface.Chinese and English are that number of users is maximum in the world, most widely used bilingual, and therefore in one embodiment, user interface shows a sentence simultaneously with Chinese and English.With reference to shown in Figure 4, while Chinese display " today, weather was terribly cold " and English " It is very cold today ".
In step 104, this sentence that shows with each language is cut apart, sentence is divided into the phrase or the phrase of this kind language.Phrase and phrase are the principal ingredients of forming sentence, when learning a language, also are crucial for the study of phrase and phrase.In step 104, for the sentence of Chinese and English demonstration, grammer and the use habit according to Chinese and English is divided into Chinese phrase or phrase with them respectively, and English phrase or phrase.And, between each phrase or phrase, increase separator and be each phrase or the unique identifier of phrase distribution.
Continuation is that example describes with example sentence shown in Figure 4 " today, weather was terribly cold " and " It is very cold today ".According to Chinese grammer and use habit, Chinese statement " today, weather was terribly cold " is split into following phrase and phrase: " today ", " weather ", " very ", " cold ".Between each phrase and phrase, increase separator: " { * # " and “ $*} ", the first symbol of " { * # " expression phrase or phrase wherein, the identifier of this phrase or phrase is distributed in " * " wherein number representative.“ $*} " end mark of expression phrase or phrase, the identifier of this phrase or phrase is distributed in " * " wherein number representative.It should be noted that for phrase and phrase are effectively distinguished the identifier of distributing to each phrase or phrase must be unique.Such as, distribute to the identifier of Chinese phrase or phrase and can use alphabetical c to add numeral.So above-mentioned Chinese statement " today, weather was terribly cold " just is split into following phrase and phrase and uses following identifier to represent:
{ c1# Jin Tian $c1}{c4# Tian Qi $c4}{c2# Hen $c2}{c3# Leng $c3}
English " It is very cold today " is split into " it is ", " very ", " cold ", " today ".Equally also be endowed separator and identifier, separator is " { * # " and “ $*} equally ", identifier is that alphabetical e adds numeral.So above-mentioned english statement " It is very cold today " just is split into following phrase and phrase and uses following identifier to represent:
{e5#it?is$e5}{e2#very$e2}{e3#cold$e3}{e1#today$e1}
In step 106, between multilingual phrase or phrase, set up association, identical semanteme expressed in phrase that is associated or phrase in sentence.In one embodiment, set up association and realize through setting up related array, preserve the identifier of phrase or phrase in the related array, the semantic terms that expression in the various language is identical or the identifier of phrase carry out association in related array.Step 106 can also adopt multilingual word technique of alignment, and this technology will be described below specially.Because method of the present invention just is used to this technology, therefore should technology not describe herein.Carrying out the related principle of phrase or phrase is that these phrases or phrase have identical semanteme in different language.Continuation describes with above-mentioned example:
The semantic association of " today ", " weather ", " very ", " cold " and " it is ", " very ", " cold ", " today " concerns as follows:
" today " correspondence " today ";
" weather " does not have correspondence;
" very " corresponding " very ";
" cold " correspondence " cold ";
" it is " do not have correspondence.
Set up after the above-mentioned incidence relation, through a related array these phrases or phrase are carried out association, related array is preserved the identifier of these phrases or phrase.Because identifier all is unique, therefore the size that identifier helps reducing related array preserved in corresponding phrase of identifier or phrase.The identifier of in one embodiment, expressing identical semantic terms or phrase in the various language is placed on same row or same delegation in related array.
Be designed as follows such as related array:
1 2 3 4 5
c c1 c2 c3 c4 N/A
e e1 e2 e3 N/A e5
The capable identifier of preserving Chinese phrase or phrase of c wherein, the identifier of e capable preservation english phrase or phrase.And Chinese phrase or phrase and english phrase or phrase with incidence relation are arranged in the same row.Such as, that first row are preserved is identifier c1 and e1, respectively corresponding " today " and " today ".
In step 108, for the sentence that shows with each language, generate and show a note, note is cut apart according to phrase or phrase and is associated with phrase or phrase.In one embodiment, the note of generation is the pinyin sequence corresponding with the sentence of Chinese.Continue with reference to shown in Figure 4; Except the sentence " today, weather was terribly cold " of reality Chinese and English with " It is very cold today ", also shown the pinyin sequence " j ī n ti ā n ti ā n q ì h ě n l ě ng " of Chinese sentence " today, weather was terribly cold ".Cut apart according to phrase or phrase.With reference to above-mentioned Chinese phrase and phrase cut apart " today ", " weather ", " very ", " cold ", pinyin sequence correspondingly is divided into phonetic segmentation " j ī n ti ā n ", " ti ā n q ì ", " h ě n ", " l ě ng ".Between each phonetic segmentation, increase separator and be the unique identifier of each phonetic segment assignments.The form of separator also is " { * # " and “ $*} ", identifier is that alphabetical p adds numeral, here, the numeral in the identifier of phonetic segmentation should with corresponding Chinese phrase or the identifier in the phrase in digital identical.Be that pinyin sequence " j ī n ti ā nti ā n q ì h ě n l ě ng " is split into following phonetic segmentation and uses following identifier to represent:
{1#jīn?tiān$1}{4#tiān?qì$4}{2#hěn$2}{3#lěng$3}
The identifier of phonetic segmentation carries out related in related array with the corresponding Chinese phrase or the identifier of phrase equally.Because the phonetic segmentation is to derive from Chinese phrase or phrase, so the identifier of phonetic segmentation is placed on same row or same delegation with the identifier of Chinese phrase or phrase in related array.For example, in above-mentioned related array, it is capable to increase the p of delegation:
1 2 3 4 5
c c1 c2 c3 c4 N/A
e e1 e2 e3 N/A e5
p p1 p2 p3 p4 N/A
Like this, that first row are preserved is identifier c1, e1 and p1, respectively corresponding " today ", " today " and " j ī n ti ā n ".
In step 110, receive selection instruction, selection instruction is selected the wherein part with this sentence of wherein a kind of language demonstration.Continuation in Fig. 4, receives a selection instruction to choose " today " in the Chinese statement with reference to figure 4.Also can select the part in the English statement, for example select " very ".Might selected content be to cut apart good phrase or phrase yet, for example might only select " the present " perhaps " my god ".
In step 112, search affiliated phrase or the phrase of part that selection instruction is selected, with this phrase or the outstanding demonstration of phrase.In the present invention, the minimum unit of processing is to cut apart good phrase or phrase, rather than independent word.Therefore, phrase or the phrase under step 112 can be searched the selected part of selection instruction.For example, no matter the user selects be " the present ", " my god " or " today ", in step 112, all can be confirmed as is to have selected phrase " today ".Given prominence to demonstration selected " today ", for example, adopts high bright demonstration or employing inverse as shown in Figure 4 to show.
In step 114, in the sentence that shows with other language and in the corresponding note, search phrase or the phrase that is associated, demonstration given prominence in these phrases that are associated or phrase.In step 112, having confirmed selecteed is Chinese phrase " today ", searches related array then, and the identifier of " today " is c1, and the identifier that is in the same row all is correlated with.So search two other identifier e1 and the p1 that are in same row with c1, find corresponding phrase or phrase or phonetic segmentation according to identifier, be respectively " today " and " j ī n ti ā n ".So " today " and " j ī n ti ā n " also shown by outstanding.
In interface shown in Figure 4, also comprise a speech play button 402.When the user presses speech play button 402; Can hear phrase or the phrase chosen; For example; The Chinese phrase of being chosen is " today ", and the user presses Chinese " today ", English " today " and the phonetic " j ī n ti ā n " that 402 on speech play button can hear successively that computing machine is read out.If the user does not also select any phrase or phrase and directly presses speech play button 402, then can hear complete sentence " today, weather was terribly cold ", " It is very cold today " and pinyin sequence " j ī n ti ā n ti ā n q ì h ě n l ě ng ".The order of playing can be consistent with the order that languages show, is Chinese, English, phonetic herein, also can change order.Accordingly, the order of demonstration also can be changed.
With reference to shown in Figure 2, the present invention has also disclosed the band note display system 200 of the multilingual language of a kind of interactive mode, user interface 202, statement sectioning 204, note generation device 206, associated apparatus 208 and display device 210.
User interface 202 shows a sentence simultaneously with multilingual, and user interface 202 also receives selection instruction, and selection instruction is selected the wherein part with this sentence of wherein a kind of language demonstration.In conjunction with above-mentioned embodiment shown in Figure 4, user interface 202 shows a sentence simultaneously with Chinese and English.
204 pairs of these sentences that show with each language of statement sectioning are cut apart, and sentence are divided into the phrase or the phrase of this kind language.Statement sectioning 204 comprises Chinese sectionaliser 240, English sectionaliser 242 and identifier divider 244.240 pairs of Chinese statements of Chinese sectionaliser are cut apart, and for example Chinese statement " today, weather was terribly cold " is split into: " today ", " weather ", " very ", " cold ".242 pairs of english statements of English sectionaliser are cut apart, and for example english statement " It is very cold today " is split into " it is ", " very ", " cold ", " today ".Chinese sectionaliser 240 and English sectionaliser 242 increase separator between each Chinese phrase or phrase and each english phrase or phrase, for example adopt " * # " and “ $*} ".Identifier divider 244 is that each phrase or phrase distribute unique identifier, and the identifier of distributing to each phrase or phrase is unique.Statement sectioning 204 finally can become following sequence with above-mentioned Chinese and English statement:
{ c1# Jin Tian $c1}{c4# Tian Qi $c4}{c2# Hen $c2}{c3# Leng $c3}
{e5#it?is$e5}{e2#very$e2}{e3#cold$e3}{e1#today$e1}
Note generation device 206 is connected to statement sectioning 204, for the sentence that shows with each language, generates and shows a note, and note is cut apart according to phrase or phrase.Note generation device 206 comprises phonetic maker 260 and phonetic sectionaliser 262.Phonetic maker 260 generates and shows and the Chinese corresponding pinyin sequence of sentence.Such as, the pinyin sequence of " today, weather was terribly cold " " j ī nti ā n ti ā n q ì h ě n l ě ng ".Phonetic sectionaliser 262 is divided into the phonetic segmentation according to the phrase or the phrase of Chinese with pinyin sequence, between each phonetic segmentation, increases separator and is the unique identifier of each phonetic segment assignments.Such as; According to " today, weather was terribly cold " cut apart " today ", " weather ", " very ", " cold ", phonetic sectionaliser 262 correspondingly is divided into phonetic segmentation " j ī n ti ā n ", " ti ā n q ì ", " h ě n ", " l ě ng " with pinyin sequence.The adding separator " * # " and “ $*} ", the following sequence of note generation device 206 final generations:
{1#jīn?tiān$1}{4#tiān?qì$4}{2#hěn$2}{3#lěng$3}
Associated apparatus 208 is connected to statement sectioning 204 and note generation device 206; Associated apparatus 208 is set up related between multilingual phrase or phrase; Identical semanteme expressed in phrase that is associated or phrase in sentence, the note that associated apparatus 208 also will be cut apart is associated with phrase or phrase.Associated apparatus 208 is set up and the maintenance association array, preserves the identifier of phrase or phrase in the related array, and the semantic terms that expression in the various language is identical or the identifier of phrase carry out association in related array.Associated apparatus 208 is with perhaps same delegation of the same row of identifier in being placed on related array of expressing identical semantic terms or phrase in Chinese and the English; Associated apparatus 208 also carries out the identifier of phonetic segmentation and the identifier of corresponding Chinese phrase or phrase related in related array, the identifier of corresponding phonetic segmentation is placed on perhaps same delegation of same row with the identifier of Chinese phrase or phrase in related array.
With embodiment shown in Figure 4 is example, and the semantic association of " today ", " weather ", " very ", " cold " and " it is ", " very ", " cold ", " today " concerns as follows:
" today " correspondence " today ";
" weather " does not have correspondence;
" very " corresponding " very ";
" cold " correspondence " cold ";
" it is " do not have correspondence.
Pinyin sequence has and Chinese phrase or the consistent incidence relation of phrase.
Set up following related array:
1 2 3 4 5
c c1 c2 c3 c4 N/A
e e1 e2 e3 N/A e5
p p1 p2 p3 p4 N/A
The capable identifier of preserving Chinese phrase or phrase of c wherein, the identifier of e capable preservation english phrase or phrase, the segmentation of the capable preservation of p phonetic.And Chinese phrase or phrase, english phrase or phrase and phonetic segmentation with incidence relation are arranged in the same row.Such as, that first row are preserved is identifier c1, e1 and p1, respectively corresponding " today ", " today " and " j ī n ti ā n ".
Display device 210 is connected to user interface 202, statement sectioning 204, note generation device 206 and associated apparatus 208.Display device 210 is searched affiliated phrase or the phrase of part that selection instruction is selected, with this phrase or phrase outstanding demonstration on user interface; Display device 210 is further searched phrase or the phrase that is associated through associated apparatus in the sentence that shows with other language and in the corresponding note, to these phrases that are associated or phrase outstanding demonstration on user interface.In the present invention, the minimum unit of processing is to cut apart good phrase or phrase, rather than independent word.Therefore, display device 210 can be searched the selected part of selection instruction affiliated phrase or phrase.For example, no matter the user selects be " the present ", " my god " or " today ", in display device 210, all can be confirmed as is to have selected phrase " today ".Given prominence to demonstration selected " today ", for example, adopts high bright demonstration or employing inverse as shown in Figure 4 to show.Search related array then, the identifier of " today " is c1, and the identifier that is in the same row all is correlated with.So search two other identifier e1 and the p1 that are in same row with c1, find corresponding phrase or phrase or phonetic segmentation according to identifier, be respectively " today " and " j ī n ti ā n ".So " today " and " j ī n ti ā n " also shown by outstanding.
In interface shown in Figure 4, also comprise a speech play button 402.When the user presses speech play button 402; Can hear phrase or the phrase chosen; For example; The Chinese phrase of being chosen is " today ", and the user presses Chinese " today ", English " today " and the phonetic " j ī n ti ā n " that 402 on speech play button can hear successively that computing machine is read out.If the user does not also select any phrase or phrase and directly presses speech play button 402, then can hear complete sentence " today, weather was terribly cold ", " It is very cold today " and pinyin sequence " j ī n ti ā n ti ā n q ì h ě n l ě ng ".The order of playing can be consistent with the order that languages show, is Chinese, English, phonetic herein, also can change order.Accordingly, the order of demonstration also can be changed.
If from the angle of application-specific, it is following a kind of method of protection that the present invention can not be regarded as yet, and with reference to shown in Figure 3, is the band note display method 300 of the multilingual language of a kind of interactive mode, comprising:
In step 302, show a sentence simultaneously with Chinese and English through user interface.
In step 304, centering sentence and English sentence are cut apart, and will be divided into Chinese phrase or phrase and english phrase or phrase.
In step 306, at Chinese phrase or phrase, and set up association between english phrase or the phrase, identical semanteme expressed in Chinese phrase that is associated or phrase and english phrase or phrase.
In step 308, according to Chinese sentence generation and show a pinyin sequence, pinyin sequence is cut apart according to Chinese phrase or phrase and is formed the phonetic segmentation, and the phonetic segmentation is associated with Chinese phrase or phrase.Being included between each Chinese phrase or phrase, english phrase or phrase and the phonetic segmentation increases separator and is each Chinese phrase or phrase, english phrase or phrase and the unique identifier of phonetic segment assignments, sets up related array and related relevant Chinese phrase or the identifier of phrase, english phrase or phrase and phonetic segmentation in related array.
In step 310, receive user's selection instruction, selection instruction is selected the wherein part with this sentence of Chinese or English demonstration.
In step 312, search affiliated phrase or the phrase of part that selection instruction is selected, with this phrase or the outstanding demonstration of phrase.
In step 314, in the sentence that shows with another kind of language and search phrase or phrase and the phonetic segmentation that is associated in the pinyin sequence, demonstration given prominence in these phrases that are associated or phrase and phonetic segmentation.
The detail of this method can be with reference to the description of figure 4 and top combination Fig. 1 and Fig. 2.
Multilingual words aligning technology introduction
As a technology of being utilized among the present invention, multilingual words aligning is simply introduced at this.
Described herein is to be used for interactively to present the words aligning of multilingual translation and come automatically from the technology of improving these translations based on user feedback.Words aligning user interface (UI) is used to for example help newspeak of a self-study.This UI shows a pair of bilingual sentence simultaneously.Each sentence all is intended on implication basic identical with another sentence, but a sentence becomes another kind of language from another sentence translation.
This UI allows the user through using cursor mutual with sentence.When the user with the word of cursor hovers in a sentence above the time, outstanding this word and the corresponding word in another sentence thereof of showing.These two outstanding demonstrations and corresponding word are words aligning each other, because these words are translated into another kind of language from a kind of language.Therefore, the user can check fast that which word translates each other, and the therefore implication of learning word in another kind of language better.
In one or more other are realized, can adopt another kind of like words aligning UI make supposition understand a multilingual people to not only a pair of, but hope to many more bilingual sentences the feedback about the quality of existing words aligning is provided.In addition, this UI can to this understand multilingual people be provided at one or more bilingual sentences between align the again chance of word.In this case, this understand multilingual people use bilingual sentence that this UI selects to be shown between preferred alignment.
In addition, based on the above user feedback technology of sketching, described herein more another other technologies can particularly be come from understanding that in a large number multilingual user collects such user feedback through the Internet and WWW (being called as " web scale " here).Such large-scale consumer feedback data has been arranged, can readjust and improve thus existing words aligning model.The model of readjusting will reflect the nuance of the word implication that the mankind perceive and more more economical than employing a large amount of language specialists better.The model utilization of readjusting is called as the notion of " masses' wisdom " sometimes.These other technologies have been arranged, can realize that this model of readjusting is to produce new and from improved multilingual text data set on to data set at new or existing bilingual sentence.
Obvious unless otherwise or from context, otherwise term as used herein is by its usual implication when submitting this paper to as those of ordinary skill in the art understands, understand.Requiring the additional place of understanding, below may be useful to the explanation of term.
Bilingual sentence is right: two sentences passing on identical in fact implication with two kinds of different natural languages.From the description purpose here, this bilingual is called as L1 and L2 respectively, and these two sentences are called as L1 sentence and L2 sentence.Certainly, each in these sentences can be a part that forms other bigger sentence set of paragraph, article, chapters and sections, books etc.Use a technical term " bilingual ", because it describes this to sentence here.Certainly, if consider more than two sentences and to its operation, then bilingual will be understood that to mean " multilingual " and " to " will be greater than two " set ".
Word: the people that a language generally scolded mostly in word understands.It is independently minimum bounded unit in the language.Word generally includes one or more morphemes, and morpheme is the minimum implication unit in the language.Because it possibly be thorny in some language that word and other words in the sentence are separated, so the sentence of phantom order word alignment must define word boundary clearly here.
Phrase: be used as the set of more than word of the individual unit in the sentence together.Word in the phrase is normally adjacent one another are in sentence, but is not like this all the time.Phrase comprises proverb, Chinese idiom and many speech title.
Words aligning: because two sentences of bilingual sentence centering mean identical things in fact, the word in the upward L1 sentence therefore directly perceived is probably corresponding to a certain word in the L2 sentence.This corresponding relation between L1 word and the L2 word is called as the words aligning link or abbreviates " link " as.The set of the all-links of sentence centering is called as the right words aligning of sentence (or abbreviating alignment as).
Be based on character or Chinese character like some language of Chinese.These characters are treated as word herein.Yet, not necessarily have the one of English word to Chinese character.In fact, one or more Chinese character can be corresponding to such as the one or more words in other language such as English.
The words aligning computing system: the words aligning computing system comprises and is configured to describe, show, show, appear and/or improve about the one or more computing equipments of one or more bilingual sentences to the words aligning data of data set.Words aligning generates through using the words aligning model.
Mechanical translation: mechanical translation becomes another kind of natural language with text or voice from a kind of natural language translation.Mechanical translation for example can be, rule-based (for example, dictionary), the statistics, based on example or its certain combination.For all translations described herein, adopt supervised learning algorithm at least in part to generate the words aligning model.
The words aligning model: generally speaking, the words aligning model is supervised learning algorithm/method of from sign/training data, learning about the something of words aligning.After a while, it is right to use this model that the words aligning that has defined (that is the word link/label that, is distributed) is become non-sign sentence.
In other words, one skilled in the art will appreciate that the words aligning model indicates sentence to regarding a branch of characteristic as with each.Each characteristic all possesses feature weight.Use for example log-linear model, characteristic and feature weight thereof are combined to produce mark.This mark is the basis how the words aligning model produces the words aligning link.The words aligning model is learnt the feature weight value from training data.In case know the feature weight value, this model just can produce for right the aligning linked of any sentence.
Training dataset: the one group of data that is used to train the words aligning model.Words aligning model described herein is based on certain machine learning and therefore it uses training dataset to learn how to produce words aligning.Each training data of this data centralization all is to come the sentence of note right with its words aligning.This training dataset is the unlabeled data collection.
The bilingual text data set: this bilingual text data set comprises that a large amount of sentences with the words aligning that produces through use words aligning model are right.The bilingual text data set is non-sign normally.Typical bilingual text data set can comprise for example 5,000,000 pairs of sentences or more.
People's natural language: people's natural language is normally human be used for exchanging each other pass through the language that organ draws.People's natural language (or abbreviating natural language as) with as computer programming language etc. is artificial or the structure language forms and contrasts.Though the main example of the language that English and Chinese are this places to be discussed can be used anyone natural language that can particularly carry out words aligning via mechanical translation.
Other termsUse word " exemplary " to mean at this as example, instance or illustration.Be described to aspect " exemplary " any or design might not be interpreted as than other aspects or designs preferred or favourable at this.On the contrary, use word " exemplary " to be intended to present each notion with concrete mode.As employed among the application, term " or " mean comprising property " or " but not alternative " or ".That is, specify or can be clear only if having in addition, otherwise " X uses A or B " means any natural comprising property arrangement from context.That is, if X uses A; X uses B; Or X use A and B, then in office how going up under the situation all satisfied " X uses A or B ".In addition, among the application with appended claims in employed article " " and " one " generally should be interpreted as and be meant " one or more ", refer to singulative only if appointment is arranged in addition or can know from context.
Fig. 5 shows and can realize that the said words aligning that is used for describes the exemplary networked computing infrastructure 500 with improved technology.Foundation structure 500 can comprise at least one the terminal user's computing equipment 502 with display screen 504, and this display screen 504 has exemplary words aligning user interface (UI) 506.This UI 506 comprises and comprises nonsensical and the text that can't understand and two sentences of symbol as shown in the figurely.Yet in this accompanying drawing, these two sentences are intended to represent that UI 506 is with the right example of bilingual sentence that shows at least.
Terminal user's computing equipment 502 normally is connected in a large amount of such equipment of language translation and learning system 520 via network 510.User-operable computing equipment 502 and this user use anyone of the Internet, WWW (" web ") etc. usually.Ground all is not shown, and computing equipment 502 has processor, storage system and I/O subsystem, such as keyboard, mouse, monitor, loudspeaker etc.Terminal user's computing equipment 502 moves such as one or more application programs such as web browsers to check exemplary words aligning UI 506 and mutual with it usually.
Simultaneously, network 510 expression interconnects and as any or combination in a plurality of networks of different type of single catenet (for example, the Internet, web or Intranet) each other.In fact, network 110 can comprise network (for example, Ethernet, cable, dial telephone cable etc.) and/or wireless network based on wiring (for example, local radio network hub, hotspot, move, honeycomb, satellite etc.).
As shown in the figure, language translation and learning system 520 comprise one or more processor 522, one or more storage system 524 and one or more storer 526.Language translation and learning system 520 comprise backend systems 530; This backend systems 530 has various assemblies, comprises User feedback collection device 532, structural data assembler 534, align factor resolver (factorizer) 536, training data renovator 538 and mechanical translation learner 540 again.Fig. 1 also illustrates system 520 and also comprises front terminal system 550.
Generally speaking, the purpose of front terminal system 550 is to allow the user to check the words aligning data effectively and mutual and the mechanism of collecting user feedback is provided with it.Front terminal system 550 generates and provides the user-interactive characteristic of words aligning UI 506.More generally, front terminal system 550 can directly not appear or show, visually appears or shows such as one or more words aligning UI described herein but can change into feasible or help.
The data of being obtained by front terminal system 550 are collected and handled to backend systems 530.More specifically, User feedback collection device 532 uses the UI as 506 to collect and store the feedback of for example using the user of terminal user's computing equipment 502 freely.User feedback for example can comprise that the user is to the grading of the existing words aligning of bilingual sentence centering.User feedback also can comprise the suggestion of user to the difference alignment of these words.
After collecting user feedback, structural data assembler 534 is handled the record of the right bilingual text data centralization of bilingual sentence.Each sentence is checked all contrasting the data structure that is produced by User feedback collection device 532.
When the user carries out the alignment again of word but not only existing alignment graded, call the factor resolver 536 that aligns again and explain to align again and produce one group " calibrated " and align again from this of many users.
538 pairs of user feedback gradings of training data renovator and user align the factor of carrying out decomposition again to form new training dataset.
Use this new training dataset, mechanical translation learner 540 operation learning algorithms are to produce new and possible improved words aligning model.Then this model is applied to existing or new bilingual text data set.If be applied to the available data collection, it is right then to come to align again to have now sentence according to this new words aligning model.Make that now multilingual text data set that upgraded or new can obtain to be used for showing to the user for front terminal system 550.
As shown in the figure, front-end and back-end subsystem (550 and 530) and each assembly are to reside in the working storage 526 and the software module of the computer executable instructions carried out by processor 522 when needed.Generally speaking, computer executable instructions is the instruction that can on the processor of one or more computing machines, computing equipment or computing machine, carry out.Though be illustrated as module at this, these subsystems and assembly can be embodied as hardware, software or its combination in any.Equally, reside on the single computing equipment (that is, language translation and learning system 520) though be illustrated as at this, these subsystems and assembly can be distributed on the many computing equipments in distributed system or the network.Data set and model can for example be stored in and store in 524.
Fig. 6 A-C shows three different views 600,620 and 640 that identical in fact interactive words aligning is described user interface (UI).This interactive mode words aligning is described UI and can be generated by a front terminal system 550 and/or a part of describing as the words aligning at 602 places.
Each view of this UI comprises that all corresponding bilingual sentence appears in the time of right.This sentence is to the sentence 622 of the sentence 602 and 604 that comprises UI view 600, UI view 620 and 624 and the sentence 642 and 644 of UI view 640.Each is English and pronounces first sentence listed in the sentence: " The sun was shining; the wind was blowing; the light and shadows were passing across the field, the birds were singing. " each is the translator of Chinese of first sentence to second sentence of listing (it is listed below first sentence) in the sentence.
Fig. 6 A shows the UI view 600 that graphic cursor 606 hovers over the top of the word interested 608 in the English sentence 602.This action is called as " mouse-over " or " putting mouse in the top (mouseover) " sometimes.This does not have that Any user is clicked or some of additional move is passive that the action of hovering can trigger an incident when graphic cursor is placed in the zone of definition.Realize that for this zone of this definition comprises the defined zone, border of the letter that forms word 608 interested and the blank between the letter at least.
Cursor 606 moves in this UI, to move about through user's the correspondence such as pointing devices such as computer mouses usually.Certainly, mouse just is used in an example of the suitable pointing device of the cursor 606 that moves about in this UI and the entire display screen.Other examples include, but is not limited to wait the cursor that drives through user's usage track ball, touch pad, graphic tablet, touch-screen, operating rod, fixed point bar, light pen, eye tracking, motion tracking telepilot, keyboard.Cursor 606 is examples of the denotable location pointer of user.Certainly, can in other are realized, be used for can be by other suitable options of position indication on the movably screen of user's indication.
Hover on the word 608 interested outstanding this interested word that shows of UI in response to cursor 606.Shown in UI view 600, word 608 interested is that " wind " and this word are given prominence to demonstration through reverse painted each letter and direct background thereof.In response to identical mouse-over, also outstanding in a similar manner another word that shows of this UI.Yet this another word is the part of second non-English sentence 604.This word is called as linked word 610 here, because it arrives word 608 interested via the words aligning data " link " that are associated.Therefore, linked word 610 supposition are translations (vice versa) of this word interested.
For for simplicity, the discussion is here described word interested, linked word and other such words as just a word.Yet, in this realization and other are realized, can relate to a plurality of words that are called as phrase.Really, shown in UI view 600, linked word 610 is actually the phrase that comprises two adjacent Chinese characters.
This section is the general introduction of the words aligning shown in Fig. 6 A being described the interactive mode action of UI.(such as the user of terminal user's computing equipment 502 of Fig. 5) moves her mouse so that cursor 608 moves to the top (or very approaching) of word 608 interested " wind " as long as the user, just outstanding simultaneously show word interested 608 in first sentence 602 and the words aligning in second non-English sentence thereof and the word 610 of link thus.This outstanding demonstration with vision matching or corresponding mode takes place.Equally, with opposite mode, when the user moves apart cursor 606 word 608 interested, no longer outstanding word 608 interested and the linked word 610 of showing.
Outstanding demonstration word means so that the mode that this word highlights from other words shows this word herein on screen.For example, can give prominence to this word of demonstration through the letter and the direct background thereof of reverse painted word.Certainly, conceive many other outstanding display techniques here.For example, overstriking, different fonts, size text, flicker, different colours, spacing, italic, underscore etc.
Fig. 6 A also shows two other words 612 and 614 in each in sentence 602 and 604 respectively.In these words each (that is overstriking) is all in a corresponding way stressed.As as used herein, " stressing " just outstanding another word that shows.Yet emphasical word will be given prominence to demonstration with the word that is described to " the outstanding demonstration " differently.
Go out as shown here, these emphasical words 612 and 614 are not word interested or linked word.Yet, each the emphasical word in sentence all with another sentence in emphasical word words aligning.
Stress that word is a looking up words here.Suppose that the user selects to carry out based on as the search of " shining " that stress word 612 or to its search through the word 614 of translation.The result of this search comprises that here the bilingual sentence of being represented by sentence 602 and 604 is right.There are the many other reasonses that can stress the word of these words aligning why.For example, because instant translation, context translation and because structure understanding.
Instant translation has avoided independent translation to search.Therefore, this translation is faster and more convenient.
Under the situation of context translation, be the translation each other in the context of sentence through the word of stressing.Compare with returning searching separately of one group of various translation, this makes the user to understand quickly.In this case, the translation scope is dwindled in the context translation, and this is of value to user's understanding and throughput rate, and is especially true when a term (such as word " make ") can have the set of very large possibility translation definition.
Under the situation that structure is understood, show the structure of sentence translation through the word of stressing.This is of value to study and how comes oneself to write with foreign language or express.This structure refers to the word ordering in the sentence.This for such as non-Latin languages such as Chinese and englishes to particularly useful, because the word order in the sentence is not necessarily 1:1 mapping between each language.In these cases, when translation English and Spanish, can't sentence structure (or word ordering) be assumed to its possible appearance generally.For example,
a.w1w2w3w4
b.w2w4w1w3
Wherein sentence " a " can be an English and sentence " b " can be a Chinese.When the user placed mouse the top of first word (w1) of a, they can see the word of the outstanding demonstration among the b in different positions.The textural difference of this this bilingual sentence centering of indication.
UI view 600 also comprises audio frequency activation selecting device 616.When the user selected this option, she depended on other options and selection and hears in many things.For example, she can hear one or two sentence that the language that is associated with her is read out loudly.Word in each sentence can show with the word of reading is outstanding synergistically.Perhaps, when she with mouse-over each in the word of the outstanding words aligning that shows above the time, she can hear this word.Perhaps, under the situation of selecting this option, can speech recognition be used for the function of the denotable location pointer of user.In other words, this UI can discern the user and read which word, in a sentence, finds this word, and also therefore mouse-over such continuation above this word is elected as interested word in this word as the user then.
Fig. 6 B shows UI view 620, and it has the corresponding sentence of translating into Chinese at English sentence 622 and 624 places as UI view 600.Cursor 626 is illustrated as the top that hovers over the word interested 628 in the Chinese sentence 624.Given this, with English sentence 622 in the outstanding simultaneously Chinese word 628 interested that shows of English word " light " 630 (it is the linked word of interested Chinese word 628) of words aligning.These sentences also have looking up words 632 and 634, and audio frequency activation selecting device 636.
Fig. 6 C shows UI view 640, and it has the corresponding sentence of translating into Chinese at English sentence 642 and 644 places as UI view 600.Its cursor 646 is illustrated as the top of the word interested 648 (" across ") that hovers in the English sentence 642.
Yet, with taken place in the describing of Fig. 6 A and 6B different, come the outstanding word that shows in response to this action of hovering.This is because there is not the words aligning that is associated with word 648 interested (" across ").In this case, there is not linked word, therefore do not have the outstanding linked word that shows certainly.Also, show interested word so also do not give prominence to because there is not words aligning.In this way, the user knows the words aligning that does not exist about word interested fast.These sentences also have looking up words 652 and 654, and audio frequency activation selecting device 656.
In addition, use so to be in the UI shown in the UI view 640, the user can select any word (for example, " across " at 648 places) and this UI to initiate additional move based on this word.For example, can carry out inquiry via the on-line search engine.This inquiry can be for example in order to obtain the dictionary enquiring of word definitions.
This UI view 640 comprises user feedback grading instrument 658.This instrument has been arranged, and the bilingual sentence that the user can select to show simultaneously is to the grading of 642 and 644 existing words aligning.This rating system can have some grades and be provided with via the slider bar shown in 658.Perhaps, this rating system can be two hierarchical systems.For example, thumb is represented up, and thumb representative down is bad, and not grading of supposition under the user does not have about this situation of option.
Use is from the feedback of a large number of users of web scale, and " masses' wisdom " effect will help to produce and reflect the answer of words aligning more accurately best.
Fig. 7 A-D shows four different views 700,720,740 and 760 of the interactive words aligning improvement user interface (UI) of counterpart.This UI has been arranged, the user have an opportunity to redistribute bilingual sentence between words aligning.This interactive mode words aligning is improved UI and can described to generate with the part of the data aggregation (discussing after a while) at 804 places by front terminal system 550 and/or as the words aligning at 802 places.
Each view of this UI comprises that all corresponding bilingual sentence appears in the time of right.This sentence is to the sentence 702 and 704 that comprises UI view 700, the sentence 722 and 724 of UI view 720, the sentence 742 and 744 of UI view 740, the sentence 762 and 764 of UI view 760.The sentence that first of each centering listed is English and pronounces: " Thus, Einstein took the job in the Swiss Patent Officein 1902.Each is the translator of Chinese of first sentence to second sentence of listing (it is listed below first sentence) in the sentence.
Fig. 7 A shows has user's UI view 700 of the alignment option 706 again.When the user selects this additional user again during the alignment option 706, present the UI shown in UI view 720 to her.
Fig. 7 B shows UI view 720.This UI has English sentence 722 and at the identical in fact sentence of translating into Chinese at 724 places as UI view 700.This UI has been arranged, the user have an opportunity to redistribute bilingual sentence between words aligning.Really, the indication 726 that provides at the top of UI view 720 indicates this to redistribute through saying " click you and hope the English word that aligns ".Also illustrating herein but becoming shown in the ash is to retreat-advance navigational arrows 728, supplies the user to redistribute and can between these are redistributed, retreat and advance once the words aligning of making this user.
Fig. 7 C shows UI view 740, and it has the identical in fact sentence of translating into Chinese at English sentence 742 and 744 places as present other UI views of discussing.Equally, as other UI views, this UI view has indication 746 and navigational arrows 748.In case the user makes words aligning and redistributes, this user just can use navigational arrows 748 between these are redistributed, to navigate backward and forward.
This UI view 740 illustrates user's mobile graphics cursor 750 and select the word interested 752 (" job ") in the English sentence 742.This selection is described the more active of only hovering that UI describes such as the interactive words aligning shown in above reference 600 and 620.Usually, the user clicks or presses a certain key and will cause this selection.In case the user selects word 752 interested, this UI is just outstanding to show that this word and this word keep outstanding and show.
Fig. 7 D shows UI view 760, and it has two identical in fact bilingual sentences as other UI views right: 762 and 764.Equally, as other UI views, this UI view has indication 766 and navigational arrows 768.In this UI view 760, the outstanding word interested 770 (" job ") that shows in the English sentence 762 was because the user had before selected this word in Fig. 7 C.As shown in the figure, the user uses graphic cursor 772 to select the linked word 774 in the Chinese sentence 764.
Therefore, through these actions, this user has defined the link between English word 770 interested and the Chinese linked word 774.At this moment, this user can select to indicate the hyperlink option " continuation " in 766 so that between these two sentences, align more words again.Perhaps, this user can select hyperlink option " submission data " to align again with completion and send data so that handle at language translation and learning system 520 places of for example Fig. 5.
Fig. 8 is from the expression that improves words aligning circulation 800.Though Fig. 5 shows the physical basis structure of exemplary realization, Fig. 8 shows the data stream of same or analogous foundation structure and representes.According to this data stream, improve the words aligning existing words aligning data of 800 illustrated that circulate certainly, improve these data and redeploy the cyclic process of improved words aligning data.
802, show existing words aligning data.Usually, a pair of bilingual sentence comes to appear simultaneously via the UI shown in 506.This can be realized that this front terminal system 550 can send instruction and data to terminal user's computing equipment 502 by front terminal system 550, and this instruction and data makes this equipment on display screen 504, produce UI 506.For example, can in the context of html web page environment, show these words aligning data.
804, collect and the storage user feedback data.Usually, this user feedback data is collected via UI and is stored by for example User feedback collection device 532.In case collect user feedback, just it be stored in the words aligning data structure that has defined to follow the tracks of sentence, user, grading, align data and any other useful metadata again.
The example of adoptable here suitable words aligning data structure comprises following field: sentence identifier, user identifier, user feedback grading, user's word align and user's specific value again.At least one sentence of sentence identifier word segment identification bilingual sentence centering.The user identifier field identification provides the user of feedback.The indication of user feedback grading field about bilingual sentence between the user option of quality of existing words aligning.User's word field of aliging again is the right one group of user feedback words aligning link of sentence.Redistributing that each is new all has metadata, redistributes the THE MEASUREMENT OF TIME that words aligning spends such as the user.This time measure can use when degree of confidence factor calculating after a while.For example, if make alignment too soon, then the degree of confidence to the link accuracy is lower.User's special use is put the letter value and is associated with the user.This can be the overall application program private subscribers confidence value that helps the contribution between other user's contributions of calibration.
806, after collecting user feedback, this user feedback is assembled in the data structure.This can be realized by for example structural data assembler 534.Each sentence is checked the data structure that all produces when impinging upon feedback capture.Can generate following corresponding to the right record of each sentence of data centralization:
● give its existing alignment how many (for example, " thumb up " or be higher than the grading of fixed amount) feedbacks certainly;
● give its existing alignment what negate (for example, " thumb down " or be lower than the grading of fixed amount) feedback.
Then; 806; Calculate following record corresponding to each word in first (" L1 ") sentence of sentence centering appointment: the link of existing this word that alignment model gives and the user that each is different give the link (that is alignment again) of this word.
808, user's factor of carrying out of aliging is again decomposed.This can be realized by the factor resolver 536 that for example aligns again.When the user carries out alignment again but not only existing alignment graded, produce one group " calibrated " and align again.This factor is decomposed and can be calculated by following mode:
1) supposes that the L1 sentence comprises m word (that is, this sentence has " w1 (1) w1 (2) w1 (m) " form) and hypothesis L2 sentence comprises n word (that is, this sentence has " w2 (1) w2 (2) w2 (n) " form).
2) suppose to exist and to give this sentence with feedback x the user of (comprising L1 and L2) (is called as u 1, u 2,, u x).Then there be x+1 voter.First ticket is (to be called as ' user 0 or u by existing words aligning model 0) launch, and other tickets are launched by human user.
3) for each L1 word w1 (i), there be n+1 ballot candidate in 1≤i≤m, promptly w2 (0), and w2 (1), w2 (2),, w2 (n) }.Candidate ' w2 (0) ' mean w1 (i) not with the L2 sentence in any words aligning, and other candidate w2 (j) (each candidate among 1≤j≤n) means that w1 (i) aligns with certain words w2 (j) in the L2 sentence.
4) if the user does not give any correction of link of w1 (i), also therefore her ballot is identical with the ballot of existing words aligning model then to suppose she to agree existing link.
5) from user u k(ballot of 1≤k≤x) is assigned with weights W kW kConfirm through the blocking factor tabulation.These factors are divided into two groups:
(i) first group of relevant degree of confidence of factor to the user.A factor is based on user's credibility of user's precedence record; Another factor is the time that the user was spent before the correction of making her.
(ii) second group of factor is about throwing the degree of confidence of the link of its ticket to the user.These factors can be whether (a) link obtains dictionary support, (b) link whether to obtain support, (c) link to the statistical study of bilingual text data set be enough under given its contextual situation, to look reasonable.
6) for each w1 (i), each candidate w2 (j) is assigned with a mark:
Σ k = 1 x W k · δ ( j , k ) Equality (1)
If wherein user k votes to candidate w2 (j), then (j k) is defined as 1 to δ, otherwise is 0.Obtain the new url of the highest candidate of branch as w1 (i).Can obtain new alignment thus for whole L1 sentence.
810, the grading of handling existing words aligning is alignd again with the new word of being advised and should grade and align again that to be inserted into new training data concentrated.This can be realized by for example training data renovator 538.
Sentence for being given user feedback grading is right, with these sentences to being divided into two groups: reservation group and want revised group.The reservation group will be the part of new training dataset and comprise that the sentence that its existing words aligning has been given many asserts feedback is right.To want revised group will can not be the part of new training dataset and comprise that the sentence that its existing words aligning has been given many negative feedback is right.This alignment group again can checked by lineup's speech like sound expert after a while.Certainly,, other can adopt other statistical computations and threshold value in realizing.
For right, decompose result that the user aligns again 808 based on factor and calculate calibrated word and align again through the sentence of alignment again.The sentence that this calibrated word aligns again is to being included as the part of new training dataset.
Use from User feedback collection to data produce new unlabeled data collection so as the training.Based on this new training dataset, can train new words aligning model.All sentences in the glossary of the example that non-unlabeled data collection is multilingual text data set are right.The words aligning model that this is new is applied to non-unlabeled data collection to produce new aligning linked.
812, the learning algorithm/method of on the training dataset that has upgraded, moving is to produce new and possible improved words aligning model.This can be realized by for example mechanical translation learner 540.
814, consume and dispose improved words aligning model and use for the user.This means new model is applied to available data collection or new multilingual text data set.If be applied to the available data collection, it is right then to come to align again to have now sentence according to this improved words aligning model.Make that now multilingual text data set that upgraded or new can obtain to be used for showing to the user for front terminal system 550.
At last, circulation turns back to 802, there via UI to the user show based on improved words aligning model upgrade or new multilingual text data set so that the user can for example learn a language or provide feedback and improve current words aligning.
Fig. 9-the 12nd illustrates and realizes described hereinly being used for that words aligning is described and/or the process flow diagram of the example process of improved technology 900,1000,1100 and 1200.UI shown in Fig. 6 A-C and the 7A-D is generated and/or is utilized by example process 900,1000 and 1100.
In these processes each all is illustrated as the set of logical flow chart center, and the set of this frame representes that available hardware, software or its make up the sequence of operation that realizes.In the context of software, frame table shows the computer instruction that is stored on one or more computer-readable recording mediums, and set operation is being carried out in these instructions when carrying out such as the one or more processors of computing machine.Notice that the order of describing process is not intended to be interpreted as restriction, and the said process frame of any number can be by any order combination to realize this process or replacement process.In addition, can from process, delete each frame, and not deviate from the spirit and the scope of theme described herein.
Fig. 9 shows words aligning and describes process 900.This process is carried out by the words aligning computing system at least in part.This computing system comprises the one or more computing equipments that are configured to describe, show, show, appear and/or improve the right words aligning of one or more bilingual sentences.This words aligning computing system for example comprises, computing equipment 502, language translation and learning system 520 or its certain combination.So the words aligning computing system of place description ground configuration is used as particular machine or device.
Go out as shown here, process 900 starts from operating 902, and it is right that the words aligning computing system obtains at least one bilingual sentence there.
In operation 904, the words aligning computing system is gone up each sentence that comes to show simultaneously bilingual sentence centering via UI at Output Display Unit (display screen 504 as shown in Figure 5).Can stress the word or expression of the one or more words aligning in each sentence.The example of this situation is seen in Fig. 6 A, stresses the word " shining " 612 in the sentence 602 together with words aligning alliance (confederate) word or expression as the word in the sentence 604 614 there.Should stress to pass on specific meanings, for example, the word 612 and 614 among Fig. 6 A is looking up words.It is right that in these two words one has been used to search the specific sentence shown in Fig. 6 A.
In operation 906, the words aligning computing system waits for that the user produces the incoming event that this user of indication has selected the interested word or expression in the sentence.This incoming event for example can be, mouse cursor hovers is above a word or expression or near this word or expression.No matter this interested word can be in arbitrary sentence and the language or the order of sentence.For the sake of clarity, the sentence that has a word interested is called as " first " sentence here.
In operation 908,, whether there is corresponding word in just definite another (that is, the second) sentence of this system with this words aligning interested in case selected word interested.If be not, it is selected to wait for another word interested that then this process turns back to operation 906.If for being that then this process proceeds to next operation.
In operation 910, in another (that is, second) sentence and specific corresponding word words aligning interested are located by this system.Specific corresponding word in this another sentence is called as linked word.
In operation 912, two words of the outstanding simultaneously demonstration of this system.In other words, this system outstanding simultaneously word interested and linked word of showing on screen.
Perhaps, operation partly or entirely can be described among the 906-912 comprises and confirms that the denotable location pointer of user is whether near the interested word or expression in the right sentence of bilingual sentence.When the denotable location pointer of user when the interested word or expression, linked word or phrase corresponding to this interested word or expression are searched another sentence centering by this system.This linked word or phrase based on bilingual sentence between the booking list word alignment search.Then, this system comes the outstanding simultaneously interested word or expression of demonstration and linked word or phrase via UI on display, and in each sentence that still shows bilingual sentence centering via this UI simultaneously, does like this.
In operation 914, this system receives indication about the user's grading of the quality of the right existing words aligning of the bilingual sentence that shows simultaneously or the input of suggestion from the user.
Figure 10 shows another process 1000 that words aligning is described.This process is carried out by the words aligning computing system at least in part.Process 1000 adopts like process 900 employed words aligning computing systems.Equally, process 1000 can or be separated employing with 900 whiles of process.
Go out as shown here, process 1000 starts from operating 1002, and it is right that the words aligning computing system obtains at least one bilingual sentence there.In operation 1004, the words aligning computing system is gone up each sentence that comes to show simultaneously bilingual sentence centering via UI at Output Display Unit (display screen 504 as shown in Figure 5).
In operation 1006, the words aligning computing system waits for that the user produces the incoming event that this user of indication has selected the interested word or expression in the sentence.This incoming event for example can be, mouse cursor hovers is above a word or expression or near this word or expression.This incoming event can be such as the result who clicks (left click or right click), hot key or select through the users more initiatively such as text selecting that above required text, pull cursor.Pulling here comprises that moving cursor presses the button (usually mouse originally on one's body) simultaneously.
In operation 1008, in case selected word interested, this system just carries out inquiry based on this word or expression interested.This inquiry can be via the BING like Microsoft TMOn-line search engines such as the search engine of brand.Perhaps, this inquiry can be to dictionary, multilingual dictionary or translator.Perhaps, can make this inquiry to the database of wherein describing or detail the word implication.
But, rather than only search for based on interested word with other realization query link word or expressions and/or the interested word or expression of process 900 combinations.Therefore, in this case, in response to selecting English word, the user can obtain the English definition of Russian word of the words aligning of this English word in Russia's sentence of bilingual sentence centering.
In operation 1010, this system presents Query Result.
Figure 11 shows the process 1100 that is used to help to improve words aligning.This process is carried out by the words aligning computing system at least in part.The words aligning computing system that process 1100 adopts like process 900 and 1000.Equally, process 1100 can adopt or separately adopt together with other processes described herein.
Go out as shown here, process 1100 starts from operating 1102, and it is right that the words aligning computing system obtains at least one bilingual sentence there.In operation 1104, the words aligning computing system is gone up each sentence that comes to show simultaneously bilingual sentence centering via UI at Output Display Unit (display screen 504 as shown in Figure 5).
In operation 1106, the words aligning computing system waits for that the user produces the incoming event that this user of indication has selected first word or expression in the sentence.No matter this first word or expression can be in arbitrary sentence and the language or the order of sentence.For the sake of clarity, the sentence that has this first word or expression is called as " first " sentence here.
In operation 1108, in case selected first word or expression, just outstanding this first word or expression that shows of this system.
Then, in operation 1110, this system wait user produces another incoming event that this user of indication has selected second word or expression of two another sentences in the sentence.For the sake of clarity, the sentence that has this second word or expression is called as " second " sentence here.
In case the user has selected first and second word or expressions, she just indicates these two word or expressions that should align.This is referred to herein as, and word aligns again or the user feedback word aligns again.
In operation 1112, in case selected second word or expression, just outstanding this second word or expression that shows of this system.
In operation 1114, this system's storage user feedback word aligns again.
In addition, in operation 1116, other characteristics that this system is associated with the user that this user feedback word of execution aligns again.For example, other characteristics can comprise the time measure between the word selection, and this time measure can be the designator whether this user considered the context of word in earnest.
Figure 12 shows the process 1200 that is used to help to improve words aligning.This process is carried out by the words aligning computing system at least in part.The words aligning computing system that process 1200 adopts like other processes of having described.Equally, process 1200 can adopt or separately adopt together with other processes described herein.
Go out as shown here, process 1200 starts from operating 1202, and system obtains the user feedback grading to the quality of the existing words aligning of data set of the bilingual sentence that perceives about the user there.
In operation 1204, this system grades based on user feedback and selects one group of sentence that keeps right.This also can be described to this system bilingual sentence is divided at least two groups to data set.The reservation group comprises satisfied or right above the sentence of the definition quality standard of grading based on the user feedback of being obtained.It is right to want revised group to comprise to have to be lower than based on the sentence of the existing words aligning of the definition quality standard of the user feedback grading of being obtained.
Defined quality threshold can be provided with (via statistical study) automatically or manually is provided with by human operator.In case be provided with, each sentence is to just having satisfied or surpassing threshold value, certain total grading that perhaps is lower than this threshold value.The right total grading of each sentence can be based on calculating such as a plurality of factors such as the intermediate value of the user feedback grading of specific right words aligning or averages.Also can use other statistical considerations, comprise the amount (for example, at least 900 " good " grading) of specific grading or based on the weight of the degree of confidence that is associated with the specific user.
In operation 1206, the words aligning computing system obtains about the right user feedback word of many sentences of data centralization align data again.
This user feedback word align data and user feedback grading again can obtain from understanding multilingual user in a large number through the Internet.Because possible global scale therefore can be from thousands of, hundreds thousand of and even possibility millions of customer reception feedback.
In operation 1208, the right user feedback word of sentence that aligns again based on the warp that is obtained aligns again, and this system-computed is alignd through the right calibrated word of sentence of alignment again again.These sentences are to being called as " calibrated " group or " through alignment again " group.
As the part of this operation, the special-purpose confidence value of user can calculate based on the factor that is associated with the user at least in this system.This system repeats this calculating to each user in the group of being considered then.Each link all comes weighting based on the special-purpose confidence value of each user's among the many users that calculate user.This system fetches the right calibrated word of selecting through alignment again of sentence based on the chain through weighting and aligns again then.
Then, in operation 1210, the sentence that this system keeps based on this group to through the sentence of alignment again to generating new and the improved words aligning model of possibility.
In operation 1212, this system will this new word again alignment model be applied to same or another multilingual text data set.This is improved multilingual text data set.
In operation 1214, this system demonstration is right from the sentence of improved multilingual text data set.This accomplishes extraordinary image here about that kind that process 900,1000 and 1100 is discussed.
The foregoing description provides to being familiar with personnel in this area and realizes or use of the present invention; Being familiar with those skilled in the art can be under the situation that does not break away from invention thought of the present invention; The foregoing description is made various modifications or variation; Thereby protection scope of the present invention do not limit by the foregoing description, and should be the maximum magnitude that meets the inventive features that claims mention.

Claims (20)

1. the band note display method of the multilingual language of interactive mode is characterized in that, comprising:
Show a sentence through user interface simultaneously with multilingual;
This sentence to showing with each language is cut apart, and sentence is divided into the phrase or the phrase of this kind language;
Between multilingual phrase or phrase, set up association, identical semanteme expressed in phrase that is associated or phrase in sentence;
For the sentence that shows with each language, generate and show a note, said note is by according to said phrase or phrase is cut apart and be associated with said phrase or phrase;
Receive selection instruction, said selection instruction is selected the wherein part with this sentence of wherein a kind of language demonstration;
Search affiliated phrase or the phrase of part that said selection instruction is selected, with this phrase or the outstanding demonstration of phrase;
In the sentence that shows with other language and in the corresponding note, search phrase or the phrase that is associated, demonstration given prominence in these phrases that are associated or phrase.
2. the band note display method of the multilingual language of interactive mode as claimed in claim 1 is characterized in that, through user interface with multilingual show simultaneously a sentence comprise with Chinese and English show a sentence simultaneously.
3. the band note display method of the multilingual language of interactive mode as claimed in claim 2 is characterized in that, said generation also shows that a note comprises the corresponding pinyin sequence of sentence of generation and demonstration and Chinese.
4. the band note display method of the multilingual language of interactive mode as claimed in claim 3; It is characterized in that, phrase or phrase that sentence is divided into this kind language cut apart in this sentence that shows with each language comprise: between each phrase or phrase, increase separator and distribute unique identifier for each phrase or phrase.
5. the band note display method of the multilingual language of interactive mode as claimed in claim 4; It is characterized in that; Between multilingual phrase or phrase, setting up association comprises: set up related array; Preserve the identifier of phrase or phrase in the related array, the semantic terms that expression in the various language is identical or the identifier of phrase carry out association in related array.
6. the band note display method of the multilingual language of interactive mode as claimed in claim 5 is characterized in that, the identifier of expressing identical semantic terms or phrase in the various language is placed on same row or same delegation in related array.
7. the band note display method of the multilingual language of interactive mode as claimed in claim 5; It is characterized in that; Note is comprised by cutting apart according to phrase or phrase: cut apart said pinyin sequence according to the phrase of Chinese or phrase and form the phonetic segmentation, between each phonetic segmentation, increase separator and be the unique identifier of each phonetic segment assignments.
8. the band note display method of the multilingual language of interactive mode as claimed in claim 7; It is characterized in that note is associated with phrase or phrase and comprises: in related array, carry out the identifier of phonetic segmentation and the corresponding Chinese phrase or the identifier of phrase related.
9. the band note display method of the multilingual language of interactive mode as claimed in claim 8; It is characterized in that note is associated with phrase or phrase and comprises: the identifier of corresponding phonetic segmentation and the identifier of Chinese phrase or phrase are placed on same row or same delegation in related array.
10. the band note display system of the multilingual language of interactive mode is characterized in that, comprising:
User interface, user interface shows a sentence simultaneously with multilingual, and user interface also receives selection instruction, and selection instruction is selected the wherein part with this sentence of wherein a kind of language demonstration;
The statement sectioning is cut apart this sentence that shows with each language, sentence is divided into the phrase or the phrase of this kind language;
The note generation device is connected to the statement sectioning, for the sentence that shows with each language, generates and shows a note, and said note is cut apart according to said phrase or phrase;
Associated apparatus; Be connected to statement sectioning and note generation device; Between multilingual phrase or phrase, set up association, identical semanteme expressed in phrase that is associated or phrase in sentence, and the note that said associated apparatus also will be cut apart is associated with said phrase or phrase;
Display device is connected to said user interface, statement sectioning, note generation device and associated apparatus, and display device is searched affiliated phrase or the phrase of part that selection instruction is selected, with this phrase or phrase outstanding demonstration on user interface; Display device is further searched phrase or the phrase that is associated through associated apparatus in the sentence that shows with other language and in the corresponding note, to these phrases that are associated or phrase outstanding demonstration on user interface.
11. the band note display system of the multilingual language of interactive mode as claimed in claim 10 is characterized in that, said user interface shows a sentence simultaneously with Chinese and English.
12. the band note display system of the multilingual language of interactive mode as claimed in claim 11 is characterized in that said note generation device comprises the phonetic maker, generates the pinyin sequence corresponding with Chinese sentence and shows in user interface.
13. the band note display system of the multilingual language of interactive mode as claimed in claim 12; It is characterized in that; Said statement sectioning comprises Chinese sectionaliser and English sectionaliser; Between each Chinese phrase or phrase and each english phrase or phrase, increase separator respectively, said statement sectioning also comprises the identifier divider, is each phrase or the unique identifier of phrase distribution.
14. the band note display system of the multilingual language of interactive mode as claimed in claim 13; It is characterized in that; Said associated apparatus is set up and the maintenance association array; Preserve the identifier of phrase or phrase in the related array, the semantic terms that expression in the various language is identical or the identifier of phrase carry out association in related array.
15. the band note display system of the multilingual language of interactive mode as claimed in claim 14; It is characterized in that said associated apparatus is with perhaps same delegation of the same row of identifier in being placed on related array of expressing identical semantic terms or phrase in Chinese and the English.
16. the band note display system of the multilingual language of interactive mode as claimed in claim 14; It is characterized in that; Said note generation device comprises the phonetic sectionaliser; The phonetic sectionaliser is divided into the phonetic segmentation according to the phrase or the phrase of Chinese with pinyin sequence, between each phonetic segmentation, increases separator and is the unique identifier of each phonetic segment assignments.
17. the band note display system of the multilingual language of interactive mode as claimed in claim 16 is characterized in that, said associated apparatus in related array carries out related with the corresponding Chinese phrase or the identifier of phrase the identifier of phonetic segmentation.
18. the band note display system of the multilingual language of interactive mode as claimed in claim 17; It is characterized in that said associated apparatus is placed on same row or same delegation with the identifier of the phonetic segmentation of correspondence with the identifier of Chinese phrase or phrase in related array.
19. the band note display method of the multilingual language of interactive mode is characterized in that, comprising:
Show a sentence through user interface simultaneously with Chinese and English;
Centering sentence and English sentence are cut apart, and will be divided into Chinese phrase or phrase and english phrase or phrase;
At Chinese phrase or phrase, and set up association between english phrase or the phrase, identical semanteme expressed in Chinese phrase that is associated or phrase and english phrase or phrase;
According to Chinese sentence generation and show a pinyin sequence, said pinyin sequence is cut apart according to Chinese phrase or phrase and is formed the phonetic segmentation, and the phonetic segmentation is associated with Chinese phrase or phrase;
Receive user's selection instruction, said selection instruction is selected the wherein part with this sentence of Chinese or English demonstration;
Search affiliated phrase or the phrase of part that said selection instruction is selected, with this phrase or the outstanding demonstration of phrase;
In the sentence that shows with another kind of language and search phrase or phrase and the phonetic segmentation that is associated in the pinyin sequence, demonstration given prominence in these phrases that are associated or phrase and phonetic segmentation.
20. the band note display method like the interactive multilingual language of claim 19 is characterized in that,
Between each Chinese phrase or phrase, english phrase or phrase and phonetic segmentation, increase separator and be each Chinese phrase or phrase, english phrase or phrase and the unique identifier of phonetic segment assignments, set up related array and related Chinese phrase of being correlated with or the identifier of phrase, english phrase or phrase and phonetic segmentation in related array.
CN201110143989.8A 2010-12-30 2011-05-20 The band note display method of interactive multiple languages and system Active CN102541821B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201110143989.8A CN102541821B (en) 2010-12-30 2011-05-20 The band note display method of interactive multiple languages and system

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
CN201010624605.X 2010-12-30
CN201010624605X 2010-12-30
CN201010624605 2010-12-30
CN201110143989.8A CN102541821B (en) 2010-12-30 2011-05-20 The band note display method of interactive multiple languages and system

Publications (2)

Publication Number Publication Date
CN102541821A true CN102541821A (en) 2012-07-04
CN102541821B CN102541821B (en) 2016-12-14

Family

ID=

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103198722A (en) * 2013-03-15 2013-07-10 肖云飞 English training method and English training device
CN105190489A (en) * 2013-03-14 2015-12-23 微软技术许可有限责任公司 Language model dictionaries for text predictions
CN105933096A (en) * 2016-06-30 2016-09-07 广东小天才科技有限公司 Information processing method and device based on error correction feedback
CN106201509A (en) * 2016-07-08 2016-12-07 南京贝湾教育科技有限公司 A kind of method for information display, device and mobile terminal
CN107194490A (en) * 2016-03-14 2017-09-22 商业对象软件有限公司 Predict modeling optimization
CN107451119A (en) * 2017-07-26 2017-12-08 上海智臻智能网络科技股份有限公司 Method for recognizing semantics and device, storage medium, computer equipment based on interactive voice
CN108520644A (en) * 2017-12-14 2018-09-11 李敏 A kind of E-learning system to learn to speak Chinese and method
CN109165389A (en) * 2018-07-23 2019-01-08 北京搜狗科技发展有限公司 A kind of data processing method, device and the device for data processing
CN109190128A (en) * 2018-09-28 2019-01-11 郭派 A kind of method and system constructing English paraphrase group
CN109902295A (en) * 2019-02-01 2019-06-18 杭州晶一智能科技有限公司 A kind of foreign language word library self-training method based on the network information
CN110751947A (en) * 2018-11-13 2020-02-04 北京嘀嘀无限科技发展有限公司 Method for prompting user, electronic equipment and computer readable storage medium

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6131082A (en) * 1995-06-07 2000-10-10 Int'l.Com, Inc. Machine assisted translation tools utilizing an inverted index and list of letter n-grams
US20050267734A1 (en) * 2004-05-26 2005-12-01 Fujitsu Limited Translation support program and word association program
CN101118540A (en) * 2006-08-02 2008-02-06 苗玉水 Chinese characters phonetic and Chinese voice code bidirectional reversible transform method
CN101271451A (en) * 2007-03-20 2008-09-24 株式会社东芝 Computer aided translation method and device

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6131082A (en) * 1995-06-07 2000-10-10 Int'l.Com, Inc. Machine assisted translation tools utilizing an inverted index and list of letter n-grams
US20050267734A1 (en) * 2004-05-26 2005-12-01 Fujitsu Limited Translation support program and word association program
CN101118540A (en) * 2006-08-02 2008-02-06 苗玉水 Chinese characters phonetic and Chinese voice code bidirectional reversible transform method
CN101271451A (en) * 2007-03-20 2008-09-24 株式会社东芝 Computer aided translation method and device

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105190489A (en) * 2013-03-14 2015-12-23 微软技术许可有限责任公司 Language model dictionaries for text predictions
CN103198722A (en) * 2013-03-15 2013-07-10 肖云飞 English training method and English training device
CN107194490A (en) * 2016-03-14 2017-09-22 商业对象软件有限公司 Predict modeling optimization
CN105933096A (en) * 2016-06-30 2016-09-07 广东小天才科技有限公司 Information processing method and device based on error correction feedback
CN106201509B (en) * 2016-07-08 2019-02-15 南京贝湾教育科技有限公司 A kind of information display method, device and mobile terminal
CN106201509A (en) * 2016-07-08 2016-12-07 南京贝湾教育科技有限公司 A kind of method for information display, device and mobile terminal
CN107451119A (en) * 2017-07-26 2017-12-08 上海智臻智能网络科技股份有限公司 Method for recognizing semantics and device, storage medium, computer equipment based on interactive voice
CN108520644A (en) * 2017-12-14 2018-09-11 李敏 A kind of E-learning system to learn to speak Chinese and method
CN109165389A (en) * 2018-07-23 2019-01-08 北京搜狗科技发展有限公司 A kind of data processing method, device and the device for data processing
CN109190128A (en) * 2018-09-28 2019-01-11 郭派 A kind of method and system constructing English paraphrase group
CN110751947A (en) * 2018-11-13 2020-02-04 北京嘀嘀无限科技发展有限公司 Method for prompting user, electronic equipment and computer readable storage medium
CN110751947B (en) * 2018-11-13 2021-05-07 北京嘀嘀无限科技发展有限公司 Method for prompting user, electronic equipment and computer readable storage medium
CN109902295A (en) * 2019-02-01 2019-06-18 杭州晶一智能科技有限公司 A kind of foreign language word library self-training method based on the network information

Similar Documents

Publication Publication Date Title
US8930176B2 (en) Interactive multilingual word-alignment techniques
CN110750959B (en) Text information processing method, model training method and related device
US9208144B1 (en) Crowd-sourced automated vocabulary learning system
US8386240B2 (en) Domain dictionary creation by detection of new topic words using divergence value comparison
Shima et al. Overview of NTCIR-9 RITE: Recognizing Inference in TExt.
CN104102720B (en) The Forecasting Methodology and device efficiently input
Tahsin Mayeesha et al. Deep learning based question answering system in Bengali
CN105814557A (en) Systems and methods for providing context based definitions and translations of text
CN101084500A (en) Computer-implemented methods and systems for entering and searching for non-roman-alphabet characters and related search systems
Zhou et al. Chinese grammatical error correction using statistical and neural models
CN102982021A (en) Method for disambiguating multiple readings in language conversion
KR20100052461A (en) Word probability determination
Anthony Visualisation in corpus-based discourse studies
WO2009026850A1 (en) Domain dictionary creation
Kumaran et al. Compositional machine transliteration
CN111414561B (en) Method and device for presenting information
Shoaib et al. Integrating multiwordnet with Italian sign language lexical resources
Hu Usability evaluation of E-Dunhuang cultural heritage digital library
Zhao Chinese character modernisation in the digital era: A historical perspective
Samanta et al. An approach to design virtual keyboards for text composition in Indian languages
Bhattacharya et al. Bengali text input interface design for mobile devices
Rajan et al. A survey of Konkani NLP resources
Hussain et al. Usability evaluation of english, local and plain languages to enhance on-screen text readability: A use case of Pakistan
Ghosh et al. Designing an efficient virtual keyboard for text composition in Bengali
CN102541821A (en) Displaying method and system with notes for interactive multiple languages

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
ASS Succession or assignment of patent right

Owner name: MICROSOFT TECHNOLOGY LICENSING LLC

Free format text: FORMER OWNER: MICROSOFT CORP.

Effective date: 20150717

C41 Transfer of patent application or patent right or utility model
TA01 Transfer of patent application right

Effective date of registration: 20150717

Address after: Washington State

Applicant after: Micro soft technique license Co., Ltd

Address before: Washington State

Applicant before: Microsoft Corp.

C14 Grant of patent or utility model
GR01 Patent grant