WO1999044345A2 - Controlling navigation paths of a speech-recognition process - Google Patents

Abstract

A method and computer program for controlling navigation paths in a speech-recognition process in which prompts are provided to a user, and actions are taken based on comparison of the user's spoken responses to stored speech elements. The method includes organizing the prompts, actions, and speech elements in a hierarchy of nodes that include group nodes and terminal nodes, and displaying the nodes in a hierarchical list that indicates navigation paths in the speech-recognition process. A user can alter the navigation paths in the speech-recognition process by editing the position of the nodes in the hierarchical list. The speech-recognition process may be a call routing process.

Description

CONTROLLING NAVIGATION PATHS OF A SPEECH-RECOGNITION

PROCESS

Background of the Invention

Businesses often offer main phone numbers as an entry point into their

phone network. Customers who call a main number frequently need assistance in

reaching a particular person or department. Speech-based automated attendant

(autoattendant) software uses voice-recognition technology to quickly forward

calls from a main number to a particular extension. For example, an autoattendant

might ask "What is the name of the employee you are trying to reach?" then

analyze a caller's spoken reply to determine a call destination. Automating this

task reduces the burden of offering main business phone numbers.

Unfortunately, information management tasks can partially offset the

advantages of installing a speech-based autoattendant. For example, scripting the

autoattendant's queries and responses to caller speech ("call flow") can become

time consuming. Additionally, coordinating the different speech generating,

speech recognizing, and phone directory data can require significant database

management efforts.

Summary of the Invention

In general, the invention facilitates control of navigation paths in a

speech-recognition process in which prompts are provided to a user, and actions

are taken based on comparison of the user's spoken responses to stored speech elements. The invention organizes the prompts, actions, and speech elements in a

hierarchy of nodes comprising group nodes and terminal nodes and displays the

nodes in a hierarchical list that indicates navigation paths in the

speech-recognition process. A user can alter the navigation paths in the

speech-recognition process by editing the position of the nodes in the hierarchical

list.

Embodiments may include one or more of the following features. Altering

the navigation paths can be done by interacting with a graphical user interface.

The user can edit the position of a node in the hierarchical list by a drag and drop

operation that can also edit the position of hierarchically included nodes. A user

can add, delete, or edit nodes. The prompts, actions, and speech elements can be

organized as records in a relational database.

The user may also be presented with a display of hierarchically included

nodes of a selected group node in a separate list of hierarchically included nodes.

The user may be able to collapse or expand group nodes to alter the display of the

hierarchical list.

The speech-recognition process can be a call routing process. The call

routing process can include forwarding calls to phone extensions or playing

speech files.

Advantages may include one or more of the following.

Controlling call flow with an intuitive user interface reduces the burden of

call flow management. Further, the ability to process call flow based on hierarchical data can not only help callers reach an appropriate extension but can

also speed autoattendant response by narrowing acceptable responses to any

particular prompt.

The invention may be implemented in hardware or software, or a

combination of both. Preferably, the technique is implemented in computer

programs executing on programmable computers that each include a processor, a

storage medium readable by the processor (including volatile and non-volatile

memory and/or storage elements), at least one input device, and at least one

output device. Program code is applied to data entered using the input device to

perform the functions described above and to generate output information. The

output information is applied to one or more output devices.

Each program is preferably implemented in a high level procedural or

object oriented programming language to communicate with a computer system.

However, the programs can be implemented in assembly or machine language, if

desired. In any case, the language may be compiled or interpreted language.

Each such computer program is preferably stored on a storage medium or

device (e.g., ROM or magnetic diskette) that is readable by a general or special

purpose programmable computer for configuring and operating the computer

when the storage medium or device is read by the computer to perform the

procedures described in this document. The system may also be considered to be

implemented as a computer-readable storage medium, configured with a computer program, where the storage medium so configured causes a computer to

operate in a specific and predefined manner.

Other features and advantages will be apparent from the following

detailed description, including the drawings, and from the claims.

Brief Description of the Drawing

FIGS. 1A-1D are diagrams illustrating autoattendant functions.

FIG. 2 is a diagram of a computer platform that includes autoattendant

components.

FIG. 3 is a diagram of autoattendant components.

FIG. 4 is a diagram of table interrelations in an autoattendant relational

database.

FIG. 5 is a diagram of hierarchy records.

FIG. 6 is a flowchart illustrating the relationship between hierarchically

organized nodes and call flow navigation paths.

FIG. 7 is a screen display of a graphical user interface (GUI) that manages

an autoattendant.

FIGS. 8A-8D are screen displays of autoattendant GUI dialogs.

Description of the Preferred Embodiments

Referring to FIGS. 1A-1D, an autoattendant configuration 10 forwards

incoming calls 12 received by a switch 20, such as a PBX (private branch

exchange), to an autoattendant 24. The autoattendant 24 can ask a caller questions

to determine which extension 14, 16, 18, or 22 the caller wants to reach. The autoattendant 24 instructs the switch 20 to connect the incoming call 12 with the

determined extension as shown in FIG. IB.

Different extensions may belong to different groups. For example,

extensions 16 and 18 may be phones of employees in a sales department 26. If

provided with a description of this hierarchical department/employee relationship,

the autoattendant 24 can ask a caller which department they would like to reach.

After determining the caller's response, the autoattendant 24 can either forward

the call to an extension in the sales department 26 or ask the caller which

extension 16 or 18 within the sales department 26 the caller wishes to reach.

As shown in FIG. 1C, the autoattendant 24 can also process calls

originating from extensions connected to the switch 20, for example, when an

employee 14 needs to talk with another employee in a particular department 26.

After analyzing the caller's responses to questions, the autoattendant 24 can

connect the caller as shown in FIG. ID.

Instead of routing an incoming call 12, the autoattendant 24 can perform

further call processing. For example, the autoattendant 24 can play speech files of

technical support information or road directions.

Referring to FIG. 2, an autoattendant 24 can include a computer system

32 that includes a processor 36, memory 34, and other components such as bus

interface circuits (not shown). The computer platform 24 includes a standard PC

type keyboard 28, a pointing device such as a mouse 30, and a monitor 27. The

computer system 32 includes a mass storage element 38 such as a CD, floppy disk, hard disk, etc. The computer system 32 receiving incoming calls through a

line card 37. Portions of mass storage element 38 are transferred to memory 34

and processor 36 in the course of operation.

Mass storage element 32 includes autoattendant management software 40,

data 44, and voice user interface (VUI) software 42. The management software

40 provides a graphical user interface (GUI) that displays autoattendant

information on the monitor 27 and enables a manager to quickly edit and

configure data 44 to provide a desired call flow. The VUI 42 processes incoming

calls based on the data 44 as arranged by the management software 40. Locking

techniques permit the management software 40 to alter data 44 without

interrupting VUI 42 service.

Referring to FIG. 3, data 44 includes different relational databases 50 and

52, however, other implementations use a single relational database. Each

database 50 and 52 corresponds to a different call flow and produces different

prompt 54 and 58 and grammar 56 and 60 files.

Prompt files 54 and 58 include indexed signal information used by the

VUI 42 to produce autoattendant speech. For example, after accessing a relational

database 50 or 52 to determine which prompt to play, the VUI 42 may retrieve

prompt file 54 or 58 information needed to produce a particular prompt (e.g.,

"Thank you for calling our business"). Prompt files 54 and 58 can include both

prerecorded and site specific prompts. Grammar files 56 and 60 include indexed signal information that

represents the different speech elements the autoattendant can recognize in

response to a particular prompt (e.g., "sales, please"). After analyzing caller

speech, the VUI 42 can access a relational database 50 or 52 to determine how the

autoattendant should respond (e.g., forwarding the call to an extension or playing

another prompt).

Compiling software 64 produces grammar files 56 and 60 from relational

database 50 and 52 records. Compiling can occur either incrementally, en masse

for better run-time retrieval, or both when the manager initiates an incremental

compilation and allows automatic scheduling of an off hours compilation.

The management software 40 service manager 66 enables a manager to

assign different call flows, as embodied in different databases 50 and 52, to

different incoming channels 62. For example, a business may have a set of phone

lines for customer inquires and another set for technical support. The VUI 42 can

examine a call directing file 48 to determine which relational database 50 or 52

provides assigned call flow. Many channels 62 can simultaneously use the same

relational database 50 or 52.

Referring to FIG. 4, each relational database, such as database 50,

includes a call flow configuration (configuration) record 64. The configuration

record 64 stores data describing general parameters such as the type of switch

connected to the autoattendant and the language being used (e.g., English or

Spanish). The configuration record 64 can also store information that indicates normal business hours, holidays, and an extension (e.g., voice mail or an

operator) for handling messages received after hours.

The configuration record 64 also stores a configuration type identifier that

controls the prompts the VUI 42 uses to query a caller. For example, a "Basic"

configuration produces prompts for a fiat configuration that does not nest

extensions or other groups within groups. For example, a basic configuration type

might prompt "Please say the name of the person or department you would like to

reach" and forward the call based on the callers response.

A "Department-Name" configuration produces prompts for a multi-level

configuration that nests extensions, groups, etc. within other groups. A

department-name configuration might prompt a caller: "To reach a party please

say the name of their department or wait for a list of departments." If the caller

responds with a department name, the VUI 42 can list the names of people in that

department for selection.

Each relational database 50 includes a table of hierarchy records 66.

Referring also to FIG. 5, each hierarchy record 66 describes a node in a hierarchy

68. A node can be a group node 26, 130, 132, 134, or a terminal node. A terminal

node can represent an extension 14, 16, 18, 24, 140, a speech file 136, 138, or a

file that includes further call processing instructions (not shown). A group 26,

130, 132, 134 can hierarchically include (i.e., parent) any of the other node types.

The connections between nodes form different navigation paths a caller can navigate. Referring again to FIG. 4, a hierarchy table 66 record includes a unique,

identification number used as an index 5 into other database 50 tables.

Records in name detail 70 and group detail 72 tables further describe each

hierarchy table 66 record. For example, a name detail table 70 record includes an

employee's name and an extension the VUI 42 can use to forward an incoming

call. A group detail table 72 record includes a group name, but does not include

extension data since a group in a department-name typed configuration does not

result in call forwarding until further caller querying (e.g., "who in the department

would you like to speak with?").

A pronunciation table 74 describes words in both the name detail 70 (e.g.,

a person's name) and group detail 72 (e.g., the name of the group) tables. For

example, a name of "John Doe" contains two words and is represented by two

corresponding pronunciation table 74 records. A pronunciation table 74 record

includes both the word and the phonemes that construct the word. For example,

the phonemes "j", "ah" and "n" describe the word "John." The compilation

process (64 in FIG. 3) stores the collected phonemes as an entry in a grammar file

along with the unique identification number of the hierarchy table 66 record that

corresponds to the pronunciation record 74. When a caller speaks, the VUI 42

detects phonemes in the caller's speech, checks the phonemes against phonemes

in the grammar file, and retrieves the hierarchy table 66 record that corresponds

with the grammar file phonemes that match the caller's speech. For example,

when a caller says "John", the VUI 42 finds j-ah-n in a grammar file and searches the relational database 50 hierarchy table 66 for the corresponding hierarchy 66

record. If the hierarchy 66 record stores a name, the VUI 42 can forward the

caller to the extension stored in a corresponding name detail 70 record. If the

hierarchy 66 record stores a speech file reference, the VUI 42 can play the speech

file to the caller. If the hierarchy 66 record stores a group, the VUI 42 can play a

group-level prompt to further query a caller.

The management system 40 can import data into a database 50 from a

variety of sources including CSV (Comma Separated Value) files and any ODBC

(Open Database Connectivity) compliant data source (e.g., Microsoft Excel™ or

Accesses) provided a manager supplies an appropriate ODBC driver. After a

manager links fields in the imported data source with autoattendant database

fields, the autoattendant can load each data source record into hierarchy 66, name

detail 70, and group detail 72 records. The autoattendant automatically produces

corresponding pronunciation table 74 records. The manager can also specify

whether an import record (e.g., a particular person's information) overwrites an

existing record or is ignored. By importing pre-existing human resources data

files, managers can quickly begin using an autoattendant without laborious data

entry.

The database 50 also includes data that controls the prompts the VUI 42

plays to a caller. The autoattendant data 44 includes pre-recorded prompts for

responses to predefined events (e.g., the caller fails to respond or the caller says

something not in the grammar file). Referring again to FIG. 5, the prompts correspond to caller navigation to different hierarchy nodes (records). For

example, navigating to a configuration node 64 can trigger a message telling a

caller that a call occurred after hours. Navigating to a group node 26, 130, 132,

134 can trigger a group prompt telling a caller to choose a particular employee or

other.node within a group. Each node can have several associated prompts. The

VUI 42 can choose prompts based on caller behavior. For example, the VUI 42

can keep track how many times a caller fails to respond and play a series of

different prompts before terminating the call.

Referring back to FIG. 4, the template prompt table 73 stores references to

pre-recorded prompts in a prompt file. A manager can record over a pre-recorded

prompt, perhaps including business specific information, producing an overriding

prompt table 75 record that references a different prompt in the prompt file.

During a call, the VUI 42 first checks the prompt table 75 for a prompt record

before checking the template prompt table 73. The VUI 42 can then retrieve

corresponding prompt file information to produce autoattendant speech.

Referring to FIG. 6, call flow follows the hierarchy defined in the

database. After receiving an incoming call, the VUI 42 positions the caller at the

root of the hierarchy, the configuration node 64 (110). Thereafter, the VUI 42

engages the caller in "conversation" that controls navigation through the hierarchy

(114-124).

As discussed, each node has an associated set of prompts. The VUI 42

plays a prompt for the caller's current node position (112) based on caller behavior (e.g., how many times the caller a visited the same node). The VUI 42

analyzes a caller's response to a prompt by checking the node's associated

grammar file (114). The VUI 42 identifies a hierarchy table 66 record that

corresponds to the caller's speech. If the caller has specified a name record (i.e., a

record with an extension) (118) the autoattendant can forward the call (120). If

the caller has instead specified a group record, the VUI 42 advances the caller to

the group node corresponding to the caller's speech (124) and begins the

prompt/response exchange again (112).

Referring to FIG. 7, the description of the autoattendant data architecture

and call flow may have seemed complicated. Fortunately, management software

40 significantly reduces the complexity of managing such a system.

The management software 40 includes a graphical user interface (GUI) 84

constructed from different Microsoft Foundation Class (MFC) controls (e.g.,

buttons, list controls, and dialogs). The GUI 84 provides a manager with the

ability to quickly define and alter the database node hierarchy (FIG. 6) in addition

to providing an intuitive relational database management system. The GUI 78

includes a menu bar 86, toolbar buttons 88, and a side-by-side display that

includes a hierarchical list display 90, and a display of hierarchically included

nodes 92 of a selected group in the hierarchical list display 90.

The hierarchical list display 90 shows an outline of call flow as embodied

in the node hierarchy. The hierarchical list display 90 lists the names of the

configuration and hierarchy nodes. Alongside each listed name appears a folder icon 94 and a sign 93 (e.g., "+" or "-"). A sign 93 indicates whether the

hierarchical list display 90 shows nodes included in a particular node.

Expanding (e.g., clicking a listing "+" sign 93) a hierarchy node in the

hierarchical list display 90 expands the hierarchical list display 90 to show nodes

hierarchically included within the expanded node. For example, expanding group

node 96 produces a hierarchical list display 90 that includes listings of included

group nodes 95 indented relative to the expanded group node 96. Closing (e.g.,

clicking a listing "-" sign) a hierarchy node conceals nodes within the closed

hierarchy node from the hierarchical list display 90. For example, closing a group

node 96 would conceal group nodes 95 from display on the hierarchical list

display 90.

A manager can manipulate groups from the hierarchical list display 90.

For example, a manager can add and delete groups nodes from a configuration.

The hierarchical list display 90 also offers a "drag-and-drop" capability. For

example, a manager can drag a selected group into another group. Doing so, alters

a hierarchy, nesting the selected group within the other group. While presenting a

caller with more levels of navigation may seem undesirable, this technique can

help a caller quickly winnow through a tremendous amount of information

whether extensions, technical support information, etc. Nesting also narrows the

possible responses the VUI 42 needs to consider to determine a grammar file

match for caller speech speeding VUI 42 response. The hierarchically included node display 92 shows the contents of a

selected hierarchical list display 90 element. For example, selecting a group node

97 in the hierarchical list display 90 changes the selected group elements icon to

an open folder and lists hierarchically included groups and extensions in the

hierarchically included node display 92. The display 92 can include node

information (e.g., name, extension, or remarks). The display 92 can further

include management information about each node. For example, if an employee

has not recorded a pronunciation of her name, the display 92 can indicate this by

marking a node with an exclamation point (not shown). A manager can sort the

display 92 by a variety of criteria such as alphabetical order, when the node was

added to the hierarchy, etc.

A manager can add, delete, and edit display 92 elements. A manager can

also move elements (i.e., groups or names) into a different positions in the call

hierarchy by dragging-and-dropping the element into a different group in either

the hierarchically included node display 90 or the hierarchical list display 92. The

management system 40 alters database contents based on these actions. This

allows a manager to quickly reorganize data and alter call flow.

Referring to FIGS. 8A and 8B, GUI dialogs provide easy management of

database information. For example, selecting a configuration node (from FIG. 7)

for editing produces the tabbed dialogs shown in 8 A and 8B. In FIG. 8 A, a

manager can edit information in dialog fields that describe a configuration record.

In FIG. 8B, a manager can alter the configuration level prompt messages issued by the VUI 42 in response to events caused by navigation to a configuration node

including initial call processing.

Selecting a dialog's "OK" button 100 saves the edited information in the

relational database, in this case, potentially updating the configuration record and

adding new prompt records. The management software further records - when

database changes occur to coordinate record locking. The "Close" button 102

discards edits. The GUI presents a manager with a "Keep changes made" dialog

when the manager ends a management session. Those familiar with database

concepts will recognize that the "OK" button resembles a SQL INSERT

command while the "Keep changes made" dialog causes a database commit or

rollback. Familiarity with such database concepts, however, is unnecessary for

system management since the GUI presents these database concepts in dialog

buttons familiar even to casual word-processing users.

Referring to FIG. 8C, another dialog permits editing of group node

information and optionally recording a pronunciation of the group name. Altering

group node information can alter the node's description in the hierarchy table and

potentially add, delete, or modify prompt and pronunciation records. Again, the

management system 40 conceals this cascade of database changes from a

manager to ease autoattendant management.

Referring to FIG. 8D, selecting a name node produces a name properties

dialog. In this dialog, a manager can alter an employee's extension or alter the

phonemes that construct the employee's name. Similar to the group node dialog, a manager can record a pronunciation of the employee's name or let the

management software generate a pronunciation based on spelling. The

management software 40 also allows individual employees to remotely (i.e., from

any phone) record pronunciation of their own name. Any alterations can produce

or alter pronunciation records and produce entries in a prompt file.

Other embodiments are within the scope of the following claims. The

techniques described should not be considered limited to autoattendant functions,

but instead can be incorporated into a variety of applications.

Claims

What is claimed is:

1. A method of controlling navigation paths in a speech-recognition process in

which prompts are provided to a user, and actions are taken based on comparison of the user's

spoken responses to stored speech elements, the method comprising:

organizing the prompts, actions, and speech elements in a hierarchy of nodes

comprising group nodes and terminal nodes;

displaying the nodes in a hierarchical list that indicates navigation paths in the

speech-recognition process; and

altering the navigation paths in the speech recognition process in response to the user's

editing of the position of the nodes in the hierarchical list.

2. The method of claim 1, wherein altering the navigation paths comprises

interacting with a graphical user interface.

3. The method of claim 1, wherein editing the position of a node in the

hierarchical list comprises a drag and drop operation which comprises editing the position of

hierarchically included nodes.

4. The method of claim 1, wherein organizing the prompts, actions, and speech

elements in a hierarchy of nodes comprises organizing records in a relational database.

5. The method of claim 1, wherein organizing the prompts, actions, and speech

elements in a hierarchy of nodes comprises adding, deleting, or editing nodes.

6. The method of claim 1, further comprising displaying hierarchically included

nodes of a selected group node in a separate list of hierarchically included nodes.

7. The method of claim 1, wherein group nodes in the hierarchical list can be

expanded to display hierarchically included nodes or collapsed to hide hierarchically include

nodes.

8. The method of claim 1, wherein the speech-recognition process comprises a

call routing process.

9. The method of claim 8, wherein terminal nodes comprise phone extensions.

10. The method of claim 9, wherein actions comprise forwarding a call to a phone

extension when the user navigates to a phone extension terminal node.

11. The method of claim 8, wherein terminal nodes comprise speech files.

12. A computer program, residing on a computer readable medium, for controlling

navigation paths in a speech-recognition process in which prompts are provided to a user, and

actions are taken based on comparison of the user's spoken responses to stored speech

elements, the program comprising instructions for:

organizing the prompts, actions, and speech elements in a hierarchy of nodes

comprising group nodes and terminal nodes;

displaying the nodes in a hierarchical list that indicates navigation paths in the

speech-recognition process; and

altering the navigation paths in the speech-recognition process in response to the

user's editing of the position of the nodes in the hierarchical list.

13. The computer program of claim 12, wherein altering the navigation paths

comprises interacting with a graphical user interface.

14. The computer program of claim 12, wherein editing the position of a node in

the hierarchical list comprises a drag and drop operation which comprises editing the position

of hierarchically included nodes.

15. The computer program of claim 12, wherein organizing the prompts, actions,

and speech elements in a hierarchy of nodes comprises organizing records in a relational

database.

16. The computer program of claim 12, wherein organizing the prompts, actions,

and speech elements in a hierarchy of nodes comprises adding, deleting, or editing nodes.

17. The computer program of claim 12, further comprising displaying

hierarchically included nodes of a selected group node in a separate list of hierarchically

included nodes.

18. The computer program of claim 12, wherein group nodes in the hierarchical

list can be expanded to display hierarchically included nodes or collapsed to hide

hierarchically included nodes.

19. The computer program of claim 12, wherein the speech-recognition process

comprises a call routing process.

20. The computer program of claim 19, wherein terminal nodes comprise phone

extensions.

21. The computer program of claim 20, wherein actions comprise forwarding a

call to a phone extension when the user navigates to a phone extension terminal node.

22. The computer program of claim 19, wherein 2 terminal nodes comprise speech

files.