WO2000011571A1 - Adaptive natural language interface - Google Patents

Adaptive natural language interface Download PDF

Info

Publication number
WO2000011571A1
WO2000011571A1 PCT/US1999/019255 US9919255W WO0011571A1 WO 2000011571 A1 WO2000011571 A1 WO 2000011571A1 US 9919255 W US9919255 W US 9919255W WO 0011571 A1 WO0011571 A1 WO 0011571A1
Authority
WO
WIPO (PCT)
Prior art keywords
natural language
agent
user
application
user input
Prior art date
Application number
PCT/US1999/019255
Other languages
French (fr)
Inventor
Rachmat Hartono
Zeeshan Khan
Timotius Tjahjadi
Hassan Alam
Giac Vu
Original Assignee
Bcl Computers, Inc.
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
Priority to US9763098P priority Critical
Priority to US10542898P priority
Priority to US60/097,630 priority
Priority to US37271099A priority
Priority to US60/105,428 priority
Priority to US09/372,710 priority
Application filed by Bcl Computers, Inc. filed Critical Bcl Computers, Inc.
Publication of WO2000011571A1 publication Critical patent/WO2000011571A1/en

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/20Handling natural language data
    • G06F17/27Automatic analysis, e.g. parsing
    • G06F17/2785Semantic analysis
    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/20Handling natural language data
    • G06F17/27Automatic analysis, e.g. parsing
    • G06F17/2705Parsing
    • G06F17/271Syntactic parsing, e.g. based on context-free grammar [CFG], unification grammars

Abstract

A system and method for providing natural language interface for a computer system that interprets natural language user input and outputs responses using natural language are disclosed. The system (102) includes a natural language agent adapted to receive and interpret the natural language user input and to output an output command and at least one application agent adapted to receive and further interpret the output command from the natural language agent and to output an executable instruction to an application program. The natural language agent includes a syntactic parser (102b) adapted to generate a parsed sentence from the natural language user input, a semantic interpreter (102c) adapted to generate the output command from the parsed sentence, and an agent communication manager (102d) adapted to provide communication between the semantic interpreter. Each application agent may include a semantic task interpreter and at least one application wrapper.

Description

ADAPTIVE NATURAL LANGUAGE INTERFACE

CROSS-REFERENCE TO RELATED APPLICATIONS

The present application claims priority to Provisional Patent Application Nos. 60/105,428 entitled "Adaptive Natural Language Interface for Use in Applications" filed on October 3, 1998, and 60/097,630 entitled "Adaptive Personal Assistant (APA)" filed on August 21, 1998.

STATEMENT OF RIGHTS TO INVENTIONS MADE UNDER FEDERALLY SPONSORED RESEARCH

The U.S. Government has a paid-up license in certain claims of this invention and the right in limited circumstances to require the patent owner to license others on reasonable terms as provided for by the terms of DARPA Contract Numbers DAAH01-96-C-R241 and DAAH01-99-C-R057.

BACKGROUND OF THE INVENTION

1. Field of Invention

The present invention relates generally to an adaptive natural language interface for use in applications. More specifically, the present invention provides a method for receiving commands, executing received commands and adaptively interacting with the user using a natural language interface, such as a natural language speech interface.

2. Description of Related Art Making computers more user-friendly has long been a goal. More and more people, include people in non-technical fields and even young kids, use computers for various purposes, such as personal, school and/or business. Computer systems are also handling more complex tasks resulting in more increasingly more complex operations. Even conceptually simple tasks require users to execute multiple complex steps for completion of the tasks. Further, when a user switches between different application programs or vendors, e.g. from MICROSOFT EXCHANGE to NETSCAPE, the same conceptual task requires the operator to learn a new set of steps to complete the same task. For example, a conceptually simple task such as finding out whether the user has received a certain message, the user must be trained in the platform-specific graphical user interface of scrolling and the vendor- specific method of viewing new mail. As is evident, a conceptually simple task may require the user to execute multiple complex steps.

As the number of users and the complexity of computer systems increase, there is an increased need for computer systems and computer applications that require little or no training for the user to use. There is also an increased need for a method to efficiently and productively use, manipulate and control computers and applications running on computers.

Natural or spoken language is an efficient method for people to communicate and express commands. For example, voice-recognition method and software have been developed and are commercially available. Although some of these method and software allow the user to speak certain commands for the computer to execute, these voice-recognition method and software support only a predetermined set of commands at a very low-level of abstraction. The user must learn the precise words and syntax that the software can accept. In other words, the voice communication cannot handle and interpret high-level, abstract, natural language commands.

Because natural language is an efficient and easy method for people to communicate and express commands, there is a long felt need for a voice- based command system and interface that can handle high-level, abstract commands and that responds to natural language.

The Air Force Institute of Technology, MIT Media Lab, Oregon Graduate Institute, Microsoft and IBM are examples of groups conducting research in the area of spoken language input. See, for example, Ball, "Mixing Scripted Interaction with Task-Oriented Language Processing in a

Conversational Interface," International Conference on Intelligent User Interfaces, Jan. 5-8, 1999, Redondo Beach, CA, pg. 101-104.

U.S. Patent No. 5,748,974 assigned to IBM Corp. describes an example of spoken language input and, more specifically, a multimodal natural language interface for cross-application tasks. The multimodal natural language interface interprets user requests by combining natural language input from the user (spoken, typed or handwritten) with information selected from an application currently in use by the user to perform a task in another auxiliary application for processing. The information is selected by a standard technique from the current application.

Copending U.S. Patent Application Serial No. 08/919,138, assigned to the assignee of the present application and incorporated herein in its entirety by reference, describes a natural-language speech control method. The natural-language speech control method produces a command for controlling the operation of a computer from words spoken in a natural language. The method includes processing an audio signal representing the spoken words of a user to generate textual digital computer data (e.g. ASCII text), processing the textual digital computer data with a natural language syntactic parser to produce a parsed sentence that includes a string of words with each word being associated with a part of speech in the parsed sentence, and generating the command from the parsed sentence.

SUMMARY OF THE INVENTION

The present invention comprises a method for receiving commands and/or adaptively outputting results and responses using a natural language interface, such as a natural language speech interface. The method utilizes an agent-based architecture comprising a front-end natural language agent and one or more application task agents for each class of applications.

It should be appreciated that the present invention can be implemented in numerous ways, including as a process, an apparatus, a system, a device, a method, or a computer readable medium such as a computer readable storage medium or a computer network wherein program instructions are sent over optical or electronic communication lines. Several inventive embodiments of the present invention are described below. In one embodiment, the natural language interface for a computer system includes a natural language agent adapted to receive and interpret the natural language user input and to output an output command and at least one application agent adapted to receive and further interpret the output command from the natural language agent and to output an executable instruction to an application program. The natural language agent includes a syntactic parser adapted to generate a parsed sentence from the natural language user input, a semantic interpreter adapted to generate the output command from the parsed sentence, and an agent communication manager adapted to provide communication between the semantic interpreter. Each application agent may include a semantic task interpreter adapted to generate the executable instruction from the output command of the natural language agent, and at least one application wrapper, each wrapper configured to communicate with a corresponding application program.

In another embodiment, a computer readable medium on which are stored natural language interface instructions executable on a computer processor is disclosed. The natural language interface instructions generally comprises receiving natural language user input, generating a parsed sentence from the natural language user input, mapping the parsed sentence into a semantic action, and generating an instruction from the semantic action, the instruction being executable by an application.

In yet another embodiment, a method for receiving, interpreting and executing natural language user input is disclosed. The method generally comprises receiving natural language user input, generating a parsed sentence from the natural language user input, semantically interpreting the parsed sentence and generating an output command from the parsed sentence, outputting the output command to an application class agent, semantically interpreting the output command and generating an executable instruction from the output command, and outputting the executable instruction to an application program for execution by the application program. The present invention is a method for abstracting complex sequence computer operations into a conceptually simple task. The natural language interface parses the users' input and semantically maps it into a knowledge concept structure. The system then determines which application context should be responsible for interpreting and executing that command concept.

The system utilizes task application wrappers to map the complex application tasks to vendor-specific executable tasks. Thus, the natural language interface system of the present invention allows users to control multiple desktop applications by abstract commands. The system of the present invention lowers the barrier to entry to computing and greatly increases productivity by combining a spoken language system with the ability to handle higher order abstract commands in naturally spoken language. The system combines a spoken language interface with a knowledge-based semantic interpretation such that semantically equivalent abstractions result in the same operation. Syntactic and semantic interpretation of spoken language enable ease of use and complexity abstraction and provides the user access to computing through spoken language.

The system and method can be adapted to user preferences with feedback using active and passive relevance feedback techniques. Further, the present invention may include a natural language based help system in the natural language agent and each application class agent that collaborate with the user in offering assistance. For example, the system may prompt the user for semantically correct input, help the user complete tasks, and remind the user on tasks that need to be done. The system of the present invention may be utilized and is compatible with existing software applications and platforms. The system uses a set of application class agents and wrappers that provide interface between the application class agent and different applications in the class. Each agent works with a class of applications, such as electronic mail, and communicates with specific applications through application wrappers. Thus, with a modular distributed agent architecture, the system and method of the present invention is extendable to multiple applications and is scalable to large sets of networked computer systems. These and other features and advantages of the present invention will be presented in more detail in the following detailed description and the accompanying figures which illustrate by way of example the principles of the invention.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic illustration of the system and method of the present invention comprising an adaptive natural language interface for use in executing computer applications;

FIG. 2 is a schematic illustration of the natural language agent; FIG. 3 shows a simplified model of a traditional dialog manager for ordering pizza through an interactive system;

FIG. 4 is a schematic illustration of the application class agent; FIG. 5 illustrates the mapping of natural language into a set of semantic tasks by each task agent; FIG. 6 illustrates an example of a personality assessment grid; FIG. 7 illustrates an example of a computer system that can be utilized to execute the software of an embodiment of the invention and use hardware embodiments; and

FIG. 8 illustrates a system block diagram of the computer system of FIG. 7.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

The present invention comprises a system and method for receiving commands and/or adaptively outputting results using a natural language speech interface, such as a natural language speech interface. The system and method are an agent-based architecture comprising a front-end natural language agent and an application class task agents for each class of applications. The system and method may include adapting to each user, including the user's speech pattern, the current or recent commands issued by the user and the user's preferences. The following description is presented to enable any person skilled in the art to make and use the invention. Descriptions of specific embodiments and applications are provided only as examples and various modifications will be readily apparent to those skilled in the art. The general principles defined herein may be applied to other embodiments and applications without departing from the spirit and scope of the invention. Thus, the present invention is to be accorded the widest scope encompassing numerous alternatives, modifications and equivalents consistent with the principles and features disclosed herein. For purpose of clarity, details relating to technical material that is known in the technical fields related to the invention have not been described in detail so as not to unnecessarily obscure the present invention.

Referring to the schematic illustration of FIG. 1, there is shown an adaptive natural or spoken language user interface system 100 for use in executing computer applications. The interface system 100 generally comprises a voice or front-end natural language agent 102 and one or more task agents 104a-d. As is generally shown, the user 106 communicates an input phrase, command or sentence 108 to the natural language agent 102 which processes the input sentence and sends the input sentence to the appropriate one of the back-end application class task agents 104a-d.

Examples of the task agents 104a-d shown in FIG. 1 are meeting agent 104a, personal information manager agent 104b, email agent 104c, and voice training agent 104d. Each task agent 104a-d outputs to the natural language agent 102 which then delivers the natural language output 110 to the user 106. Each of the back-end application class task agents 104a-d works with a class of one or more existing computer applications. The interface system can be adapted to existing computer applications so that users can operate a computer using spoken language as well as other input devices such as keyboard and pointing devices, giving full multi-modal interface to existing computer applications.

Although the natural language user interface system 100 is generally described as one interacting in spoken natural language, the system 100 may be configured to receive and/or output using one or more alternative input and/or output mechanisms while utilizing natural language for such input and/or output interactions. Suitable alternative modes of input and/or output include keyboard, mouse, touch or contact sensitive screen, and or screen display.

FIG. 2 is a schematic illustration of the natural language agent 102. The natural language agent 102 communicates with the user 106 through spoken language. The natural language agent 102 preferably includes:

• automatic speech recognition system 102a;

• natural language syntactic parser 102b;

• natural language semantic interpreter 102c;

• agent communication manager 102d;

• adaptive preference manager 102e;

• dialog manager 102f; and

• text-to-speech synthesizer 102g.

The natural language agent 102 executes a first level interpretation of the natural language input. The front end natural language agent 102 receives all natural language input and determines which of the available task agents

104 to pass the natural language input interpreted by the front end natural language agent 102. The task agent 104 to which the natural language input was passed may return a response such as an output to the front end natural language agent 102. The front end natural language agent 102 then outputs the response from the particular task agent 104 to the user 106. The natural language agent 102 may itself return a response if it determines that the original natural language input is incomplete, incorrect, or otherwise cannot be properly interpreted. Each of these components 102a-g of the natural language agent 102 are described in more detail below.

Automatic Speech Recognition System 102a Automatic speech recognition systems for speech input are readily and commercially available off the shelf. Any suitable off the shelf speech recognition systems may be used as the automatic speech recognition system 102a in the natural language interface system 100 of the present invention. Thus, details of speech recognition methods and systems are not described herein. In addition, error correcting techniques and cue words may be utilized to improve accuracy and allow for dialog management to effectively recognize speech input.

Natural Language Syntactic Parser 102b There are generally three basic approaches to natural language syntactic processing: simple grammar, statistical and Government- and- Binding (GB-based). Simple grammar is used for simple, non-complicated syntax. Statistical approach examines word patterns and word co-occurrence and attempts to parse natural language sentences based on the likelihood of such patterns. Statistical approach uses a variety of methods such as neural networks and word distribution. The statistical approach is limited by an upper limit on error rate and it is very difficult to handle wide varieties of linguistic phenomena such as scrambling, NP (noun phrase) movement, binding between question words and empty categories. The GB-based approach is described in, for example, "Some Concepts and Consequences of the Theory of Government and Binding," Cambridge, MA, MIT Press, the entirety of which is incorporated herein by reference. The GB-based approach is a more robust approach to natural language parsing using computational methods based on linguistic theory of a universal language. GB-based approach reveals implied syntactic structure in English language sentences and thus better facilitates in resolving ambiguous syntactic structures. By using generalized principles and parameters, the GB-based approach allows a customizable and portable parser that can be tailored to different environments and languages with little modification.

Preferably, the natural language syntactic parser 102b utilizes a GB- based principle and parameters framework to parse natural language computer commands. Hageman, L. Introduction to Government and Binding Theory, incorporated by reference herein, for example, describes this concept. With the generalized principles and parameters, GB-based approaches can describe a large syntax and vocabulary relatively easily and thus may result in higher robustness than other approaches. With a GB-based approach, commands to computers can be seen as verb phrases that are a sub-set of complete English sentences. The sentences have an implied second person singular pronoun subject and the verb is active present tense.

For example, to resume work on a previous project, the user 106 may state "show me the first message." This request would be parsed into the following structure:

(VP (Vbar (V (V_IP (V_IP show [present sg]) (IP

(NP (Nbar (N me [goal animate sg]))) (Ibar (NP [these inanimate sg] (Det the) (Nbar (AP (Abar (A first)))

(N message))))))))) The parse would allow the computer to map the verb into a computer command action, with the noun phrase (NP) as the object and the adjective phrase (AP) as properties of the object.

Natural Language Semantic Interpreter 102c

The natural language semantic interpreter or interpretation engine 102c is preferably a frame-based command interpretation system. The natural language semantic interpreter 102c may interpret the syntactic parse using context sensitive methodologies. The natural language semantic interpreter 102c uses a knowledge base populated with concept-interfaces that each application can handle. The natural language semantic interpreter 102c takes the syntactic parse of a spoken language request and maps it into a generic concept frame used to invoke the appropriate application method. TABLE I lists examples of concept-interfaces.

Figure imgf000015_0001
Requests input to the computer are preferably transformed by the semantic interpretation engine 102c from a syntactic parse into a variable length verb-head frame. The process has variable length noun phrases as arguments. The noun phrases in turn have arguments that are adjective phrases. The verb-head describes an action-concept. The noun phrases describing the objects on which the actions are performed are topic-concepts and the adjective phrases describing the type of objects are modifier concepts.

Reverse Grammar Generation Mechanism The semantic interpretation engine 102 c may also include a reverse grammar generation mechanism. The reverse grammar generation mechanism may be implemented in each agent, i.e. the natural language agent and or each of the task agents. The reverse grammar generation mechanism includes a list or vector for each word and corresponding probabilities for each word in the list. For example, for the word "I," "eye," or "aye," the associated vector or list includes those words, i.e. "I," "eye," and "aye," and may have corresponding probabilities of 80%, 15%, and 5%. These probabilities may be predetermined and may be adjusted depending upon each user's selection of words used or depending upon a subset or all of the users' selection of words used.

Upon receiving the syntactic parse of the spoken language request, the semantic interpretation engine 102c determines the permutations of the syntactic parse using the list for each word. For example, using the exemplary vector above and ignoring the lists for all the other words, if an input request is "I want to go home," the permutations of the syntactic parse may include: "I want to go home," "Eye want to go home"; and "Aye want to go home." Using the permutations, the semantic interpretation engine 102c

determines which words best fit the grammar of the syntactic parse. To determine each best fit word, the word with the highest probability, i.e. "I" in the example above, is evaluated and determined if that word is suitable given the context. If the word is not suitable given the context, the next word having each next highest probability is then evaluated and determined if that word is suitable given the context, until a suitable word is determined. Of course, if no suitable word is determined, then the natural language agent may request clarification or correction from the user.

A combination of the action-concept and the topic-concepts are used to determine which task agent should handle the request. If the request is for a specialized task agent, the request is routed to that specialized task agent. If the request is for the natural language agent 102 itself, a routine associated with the command is invoked with the topics and modifiers as arguments. Using the arguments for routing commands allows for better disambiguation than with the verb alone. The above-described interpretation approach has the advantage of allowing the natural language agent 102 to query the user for clarification, for example, if the original request is incomplete, or otherwise cannot be properly interpreted by the natural language agent 102. For instance, if arguments do not match the verb, then a clarification can be requested by the natural language agent 102. Further, the above-described interpretation approach has the advantage of allowing the natural language agent 102 to properly interpret natural language requests without requiring the user to input the request conforming to specific structures. For example, in requesting an airline ticket from Portland to Boston, the user may state "I'd like a ticket to Boston from

Portland" or "I'd like a ticket from Portland to Boston." In response, the natural language agent 102 may request clarification as to Portland, Oregon or Portland, Maine, for example. The above-described interpretation approach has the advantage of that it does not rely upon certain key words in order to properly interpret the user's requests. Further, the interpretation technique may be context base or context sensitive.

Agent Communication Manager 102d

The application class task agents 104 may communicate with each other, preferably using Knowledge Query Manipulation Language (KQML) or any other suitable language, via the agent communication manager or module 102d. The contents of a message between application class agents 104 may be coded in any suitable format, preferably the Knowledge Interchange Format (KLF). When a KQML message with an "achieve" performative is received by an agent 104, the KLF encoded concept structure is interpreted further by the agent 104 through a semantic interpretation knowledge base similar to the one described above with reference to the semantic interpreter 102c. In this case, the knowledge-base only includes information on how to map application- specific modifiers to application-task parameters. Using KQML and KIF allows the different agents 104 to easily communicate with each other. In particular, the natural language agent 102 sends the user's request to the application class agent 104 via the agent communication manager 102d and the application class agent 104 sends requests back to the natural language agent 102, or some other agent, via the agent communication manager 102d. Thus an email class agent 104c can request information from a file manager class agent (not shown) using a KQML/KIF statement via the agent communication manager 102d.

The above-described approach has the advantage of modularizing the different ontologies by allowing different application class agents 104 to have different subset dictionaries and task specific semantic interpretation knowledge-bases. It also allows the class agents 104 to handle vendor-specific application features by easily modifying the local semantic interpretation tables. This is described in more detail below with reference to the application class agent 104. The task routing mechanism is similar to Galaxy II, a voice controlled system integrating three separate voice controlled systems, as discussed in Seneff et al. "Galaxy-II: A Reference Architecture for Conversational System Development," 5th International Conference On Spoken Language Processing, Nov. 30 - Dec. 4, 1998, Sydney, Australia, p.931, the entirety of which is incorporated by reference herein. Currently, the Galaxy II requires a user to explicitly switch from one domain to another.

Adaptive Preference Manager 102e

The adaptive preference manager 102e is associated the natural language each agent 102 and with each user 106. The task of the adaptive preference manager 102e is to learn what default conditions are preferred the user 106 by either monitoring the users action implicitly (i.e. observing in the background) and/or by being instructed explicitly by the user 106 on positive and/or negative preferences. These preferences may be shared by different users 106 running similar application class agents 104.

The adaptive preference manager 102e uses relevance feedback techniques. Relevance feedback technique is widely used for preference optimization with declarative preferences. A request for executing an action based on preferences can be modeled as a query to locate a document in a collection of documents. In this technique widely used in information retrieval, the relevance of a document to a query is measured by how many matches the document has with query terms. In the realm of preference requests, the result of an action is analogous to a document where the preference is analogous to a query. Using this substitution, information retrieval techniques for ranking results of action requests can be adapted according to user preferences. Criteria specified in the spoken request are also factored as preferences. For preference matching, the information retrieval formula can be adapted for preference ranking by simplifying the equation for small queries as expressed in equation (1):

∑(lDF, * wv ) similarity (Q, D) = (1)

Figure imgf000020_0001

where t = total number of independent terms wιq = (.5 + (.5 qfreqιq/maxfreqq)) x IDF, ,j = dfreq,j x IDF, qfreqιq = Frequency of term i in request q dfreqy = Frequency of term i in result j maxfreq, = Maximum frequency of any term in query maxfreqq

IDF, = log2 (maxn/n,) + 1

N = Number of results n, = Total number of occurrences of term i in the results maxn = Maximum frequency of any term in the results

The qualitative ranking can be quantified by adding a set of weights to the ranking equations (2) and (3) as set forth below to incorporate the weights applied to terms in the definition of IDF,.

maxn = max ((∑j(wι * nJ ))\/j (2)

1=1

n, = ] , * *, ) (3)

Relevance feedback techniques have been used in information retrieval techniques for improving the precision and recall of queries. In relevance feedback the query terms are reweighted by the selection of the retrieved items by the user. For the case where the user does not exhaustively select all the relevant responses, the reweighting of the term weights can be done by equations (4) and (5).

Initial Weights Wljk = (C + IDF,) * f,k (4)

Feedback W,jk = (C + log p,j ( 1 -qu)/ (l-py)qij) f* (5) where:

W,jk - weight for term i in preference j and result k

IDF, = the IDF weight for the term i in the entire set of result p,j = probability of the term i within the set of relevant results for preference j q,j = probability that term i is assigned with the set of non relevant results for preference j f,k = K + (1 - K )* freq,k/maxfreqk freq,k = the frequency of term i in result k maxfreqk = maximum frequency of any term in result k.

As noted above, the execution of a task with variable parameters can be modeled as an information retrieval query. In this case the weights for the query term can be modeled as the user's preference weights.

Help System

With a natural language based system that abstracts the semantic concepts from the complexity of tasks, much of the help system is implicitly encoded in the knowledge base. Instead of asking "How can I send my spreadsheet to John", the user asks the natural language agent 102 to "Send the spreadsheet to John." If invalid parameters are given, the user 106 is prompted for the correct parameters. However, the natural language interface system 100 is also able to handle requests for help by generating an explanation of how the request functions. It can also show an example of a typical user request to accomplish the task. Dialog Manager 102f

The natural language agent 102 further includes a dialog manager 102f. The dialog manager 102f of the natural language agent 102 controls the interactions between the user 106 and the natural language interface system 100. The dialog manager 102f is an finite state machine (FSM) similar to the one described in Cohen, et al. "The Efficiency of Multimodal Interaction: A Case Study," 5th International Conference On Spoken Language Processing, Nov. 30 - Dec. 4, 1998, Sydney, Australia, p.253, incorporated herein in its entirety by reference. The dialog manager 102f handles tasks such as accepting user inputs, obtaining parameter for tasks, requesting clarification and asking for confirmation on tasks.

The ability to handle natural language commands extends the concept of traditional dialog managers. Traditional dialog managers function like finite state machines (FSM) accepting dialog. For example, as shown in FIG.

3, ordering a pizza through an interactive system requires the user to specify the type of pizza, such as the size and topping of the pizza. A simplified model may be adopted where the user must select the size of the pizza (small, medium, or large) and the toppings (cheese, Hawaiian or pepperoni), and confirm the order. If changes are to be made to the size while selecting the topping, then either this ability to make such a change must to be written into the FSM or the user must to wait until the end of the ordering sequence.

In contrast, with spoken language commands, many of these dialog steps are unnecessary. Some FSMs can be generalized to a set of Boolean operations on a set of choices. In this case choosing a pizza is an AND operation (size, topping and confirmation) on a set of XOR operations (e.g. small, medium or large size). Thus in spoken natural language, the user may simply say "I wish to order a large cheese pizza."

As is evident, one natural language sentence completes all the choices and only a confirmation is necessary. However, additional dialog issues are created under various circumstances. For instance, the user may make an incomplete request such as "I want a cheese pizza," make an incorrect request such as "can you send veggie pizza," request information such as "what types of pizza do you have," change a request "I'd like to make that a small one," or make an out of context request such as "I want to see my email."

A global state variable may be introduced to allow the dialog manager 102f the flexibility to handle such spoken language requests. The global state variable uniquely identifies the state of the interaction between the user 106 and the natural language agent 102. The state of the natural language agent 102 can be in one of two classes: IDLE or DEFINED. Ifthe natural language agent 102 is in IDLE, the natural language agent 102 is not actively engaged in a dialog with the user 106 and interprets the request in the default global context. Ifthe natural language agent 102 is in a DEFINED state SI, the designer of the dialog has the option of specifying a set of semantic frames it will accept and the actions. Ifthe semantic frame is not defined, the action would be deemed out of context.

With the above-described scheme, if an incomplete request was made, the user 106 is prompted for more information; if an incorrect request is made, the user 106 is given a set of options from which to chose; and if a change request is made, the order is changed; if an out of context request is made, then the user 106 is asked if a context switch is indeed desired with a warning that the current context will be lost.

Text-To-Speech Synthesizer 102g The natural language agent 102 may offer the user 106 the option of receiving messages as text-on-screen or as synthesized speech with text-to- speech synthesizer 102g. The text-to-speech synthesizer 102g preferably uses commercial off the shelf technology to communicate messages to the user 106 by speech. The text-to-speech synthesizer 102g may utilize intonation to make the synthesized speech sound more natural to the user 106. In addition, the natural language interface system 100 may use Avatars for output. The text and speech messages are transmitted in conjunction with other graphical items that may be displayed by the applications and/or the agents.

Application Class Agent 104

As shown in FIG. 4 and described above, the agent communication module 102d of the natural language agent 102 allows communication between the application class agents 104 and the natural language agent 102. Each application class agent 104 preferably works with a single class of applications 112 that have similar conceptual operations. For example, different email applications generally perform the same conceptual actions of sending and receiving mail but performs these actions through different sets of steps.

Each application class agent 104 preferably includes a set of application wrappers 104A, a semantic or task interpretation engine 104B, an application class communication or dialog manager 104C, an adaptive application class preference manager 104D, and an application class help system (not shown).

The communication between the application class agent 104 and each different type of vendor-specific applications programs 112 is via an application wrapper 104 A that translates the conceptual action to a set of application specific operations. The task application wrapper 104 A is the interface between the application class agent 104 and different applications 112 in the class. With wrappers 104A, the application class agent 104 communicates with specific applications 112 allowing the incorporation of existing applications into the architecture of the system 100. For example, an e-mail agent would have a wrapper for interacting with each email system such as NETSCAPE and MICROSOFT EXCHANGE.

To interface with existing applications, the wrapper 104 A is preferably written in one of the platform specific macro languages. Examples of platform specific macro languages are listed in TABLE II.

Figure imgf000026_0001

The task or semantic interpretation engine 104B is similar to the semantic interpretation engine 102c of the natural language agent 102 described above. The task interpretation engine 104B serves as the knowledge base for each agent 104. The task interpretation engine 104B receives the semantic frame representation as input. Based on the frame's head verb (action request) and noun phrases (parameters), the task interpretation engine 104B invokes a routine that sends a set of requests to the task application wrapper 104 A.

The application class dialog manager 104C is similar to the natural language agent dialog manager 102f of the natural language agent 102 described above. The application class dialog manager 104C manages the interaction between the user 106 and the application class agent 104, requests clarification for ambiguous requests, asks for confirmation, and obtains incomplete parameters. The application class adaptive preference manager 104D records the user-preferences for each task. The preference is computed in a manner similar to the general natural language agent preference calculation for the natural language agent adaptive preference manager 102e as described above. While the natural language capability of the natural language interface system 100 desirably removes most of the user's need for help, each application class preferably has a help capability to enhance the minimum training feature of the natural language interface system 100 of the present invention. The help system can be encoded in the application class interpretation engine 104B such that the request will result in communications of instructions and explanation from the application class agent 104. For example, requests such as "How do I," "Can you show me," "What are the possible values for" will result in a response from the application class agent 102 with instructions and explanation on how to perform the task.

The help system may provide various types of help information. The help system may provide description of the agent capabilities such as the general uses of the application and the tasks the agent can perform. While the natural language interface system 100 is designed for unconstrained input, ambiguity resolution may require constraints in syntax and the help system may provide syntax for different tasks to the user 106. Thus, ifthe user 106 is unable to get the application class agent 104 to perform a task, the user 106 may ask how to execute an operation. The help system can respond with a sample natural language sentence. In addition, the help system can also provide suitable parameter values and ranges as well as the typical general help information normally included with the application on, for example, how to use the specific application

Example: Address Book Agent

The operation of the system 100 will be brief described with reference to an address book agent as an example. The address book agent comprises a task interpretation engine, a dialog manager and one or more task wrappers.

The typical key actions of an address book include show (to display all or part of an address), change (to change all or part of an address), add (to add a new address), delete (to delete an existing address), sort (to arrange address by a given category), open/close (to open or close an address book), save (to save an address book), copy/paste (to copy and paste data from one part of an address book to another part).

These actions can be interpreted by the address book agent with reference to a semantic frames knowledge-base. The frames are also inserted in the natural language agent routing table. An example of a listing of such frames is shown in TABLE III. An application wrapper interfaces with the particular address book application. The routines will handle the tasks as described above and will interface to the address book module for, e.g., MICROSOFT EXCHANGE, and NETSCAPE.

Semantic Mapping

FIG. 5 schematically illustrates the mapping of the user's input phrase, command or sentence within a large set of syntactically correct natural language phrases, commands or sentences 140 into a set of semantic tasks or actions 142 by the semantic mapper 144. Preferably, a semantic mapper 144 is provided for the natural language semantic interpreter 102c of the natural language agent 102 and/or the semantic interpretation engine 104B of each application class agent. For example, a different semantic mapper 144 may be provided for word processing applications, e-mail applications and spreadsheet applications. TABLE IV provides an illustrative listing of task agents for a class of applications and a list of sample tasks corresponding to each task agent.

Figure imgf000029_0001

Each task agent for a class of applications is preferably provided with its own set of semantically correct sentences, semantic actions and semantic mapping. Each task agent thus serves as the common user interface for the corresponding class of applications under the assumption that each application within the class accomplishes the same or generally overlapping set of tasks. In other words, in a given class of applications, there is a finite and relatively small set of semantically equivalent actions or tasks 142 which may be performed by each application within the class.

Figure imgf000030_0001
Figure imgf000031_0001

For example, for the word processor class of applications, a user may input "compose a letter to John Smith," "please begin drafting a letter to John

Smith," or "can you create a letter for my friend, John Smith?", each of which is a syntactically correct sentence within the large set of syntactically correct sentences 140. These user commands are semantically equivalent. In each case, the semantic mapper 144 maps the user input to a specific action within the small set of semantic actions 142. In this example, the semantic mapper 144 maps each of these user inputs to the same action, draft a letter to John Smith, and the same task is performed. Thus, the semantic mapper 144 ensures that the same task is performed in a given class of applications regardless of the specific user input.

Each application in the class may have a different method for accomplishing the same semantic task. In response to any of the user inputs in the example above, a word processor application composes or drafts a letter to John Smith although the particular word processor application may utilize an approach different from the approach used by another word processor application. By using a core set of semantically equivalent tasks 142 for each class of applications, the present invention allows the user to accomplish the same semantic task independent of the specific application utilized. Although a single task agent is preferably provided for each class of applications, the task engine of each task agent includes a process-specific execution module for each application. For example, the word processing task agent may include an execution module for MICROSOFT WORD and another execution module for WORD PERFECT. The process-specific execution module translates the semantic action for the specific application.

The semantic mapper 144 is capable of reducing idiomatic sentences and output a mapped semantic action. Input sentences may be generally classified as wh-query, request, let construction, infinitive, embedded clause, semantic mappings and context-dependent. Examples of the classifications of input sentences are shown in Table V. Regardless of the classification of the input sentence, each input sentence is mapped into a semantic action. Preferably, each mapped semantic action is in the form of a verb phrase or the imperative case with an implied non-phrase. "Show me mail message" is an example of a imperative verb phrase having "you" as the implied non-phrase.

TABLE V

Wh-query

What I wouldn't give to see the back of a blue car

What is stopping you from showing me my mail

What would I give to see my mail

Why don't you show me my mail How is my email situation today

Why can't I see my mail Why don't you clean my mailbox

Where is my mail How about attaching this file

Request

Can I see my mail May I see may mail

Can you show me my mail Would you allow me to see may mail

Can you like show me my mail Will you show me my mail

Let construction

Let me see my mail Let me mail be seen by me Let me mail be shown Let me know when I get mail

Infinitive

I want to check out my mail I want to know if I got new mail

I would love to check out my mail I want to access my mail

We want to see our mail now He wants to see his mail

I need to see my mail now Susan wants to see her mail

Embedded clause

I will be upset if you do not show me mail right now

I would really appreciate it if you could show me my mail

I think it would be great if you could show me my mail

I want you to tell me if I have new mail

I'm looking to see whether I have new mail we would like it if you could show us our mail

I want you to inform me if I have new mail

Semantic mappings

I would like for you to give me a view where I can see my mail

I would love it if I could play with my mail

Can you show me something like mail

Can you show me what evertone has sent to me

Context-dependent

Let me see what you have I want you to summarize these

What else can you do Who sent the last one

I want to know all about the things you can do In addition to the various input sentences, the spoken input sentences 108 given by the user 106 may contain one or more of several types of errors which may occur. These errors include unrecognized word, bad parse, unhandled verb, unhandled object, unhandled verb/object attribute and/or task- specific error. Some errors may be better handled and addressed at the natural language agent 102 while other errors may be better handled and addressed at the appropriate task agent 104. For example, errors relating to unrecognized word, bad parse and unhandled verb are preferably handled and addressed at the natural language agent 102. Errors relating to unhandled object may be handled and addressed at either the natural language agent 102 or the task agent 104. Further, errors relating to unholdled verb/object attribute and task- specific errors are preferably handled and addressed at the task agent 104. As discussed above, the interface 100 of the present invention is an adaptive natural language interface 100. The output of the natural language agent 102 is preferably adaptive to the personality of the user 106 by first identifying the personality type, personality trait or characteristics of the user and utilizing that identification for responding to the user. FIG. 6 illustrates an example of a personality assessment grid where users may be groups into one of four types: analytical, driver, amiable and expressive, which are defined depending upon the relative levels of assertiveness and responsiveness. The natural language agent may make a determination as to which of the four types best characterizes the user from factors such as the user's tone, pitch, speed and the actual words used by the user. Of course, the natural language agent may utilize any other factors, personality assessment methods and/or personality characterization schemes. The natural language agent 102 is adaptive in that it utilizes that determination of the user 106 in responding to the user by delivering the output to the user or in requesting additional information from the user using simplified emotional response. The determination thus may affect the tone, pitch, speed and/or the actual words used to respond to the user. For example, in delivering the output to the user or in requesting additional information from the user, the natural language agent may be empathetic, for example, and express similar levels of assertiveness and/or responsiveness by varying the words used, the speed at which the words are delivered, the tone and/or the pitch of the words. In addition, the form of as well as the specific graphical interface seen by the user may be determined by the user, the application currently utilized and/or based upon the determination of the user's personality.

Although the adaptive natural or spoken language user interface system 100 is described above in terms of natural language speech input, the interface system can also recognize and interpret natural language non-speech command, such as text. The natural language interface is preferably embodied in a computer program product in the form of computer coded instructions executable by a computer processor and stored in a computer readable medium.

FIG. 7 illustrates an example of a computer system that can be utilized to execute the software of an embodiment of the invention and use hardware embodiments. FIG. 7 shows a computer system 201 that includes a display

203, screen 205, cabinet 207, keyboard 209, and mouse 211. Mouse 211 can have one or more buttons for interacting with a GUI. Cabinet 207 houses a CD-ROM drive and/or a floppy disc drive 213, system memory and a hard drive (see FIG. 8) which can be utilized to store and retrieve software programs incorporating computer code that implements aspects of the invention, data for use with the invention, and the like. Although a CD-ROM and a floppy disk 215 are shown as an exemplary computer readable storage medium, other computer readable storage media including magnetic tape, flash memory, system memory, RAM, other types of ROM, and hard drive can be utilized. Additionally, a data signal embodied in a carrier wave (e.g., in a network including the Internet) can be the computer readable storage medium. FIG. 8 shows a system block diagram of computer system 201 used to execute a software of an embodiment of the invention or use hardware embodiments. As in FIG. 7, computer system 201 includes monitor 203 and keyboard 209, and mouse 211. Computer system 201 further includes subsystems such as a central processor 251, system memory 253, fixed storage 255 (e.g., hard drive), removable storage 257 (e.g., CD-ROM drive), display adapter 259, sound card 261, transducers 263 (speakers, microphones, and the like), and network interface 265. Other computer systems suitable for use with the invention can include additional or fewer subsystems. For example, another computer system could include more than one processor 251 (i.e., a multi-processor system) or a cache memory.

The system bus architecture of computer system 201 is represented by arrows 267. However, these arrows are illustrative of any interconnection scheme serving to link the subsystems. For example, a local bus could be utilized to connect the central processor to the system memory and display adapter. Computer system 201 shown in FIG. 8 is but an example of a computer system suitable for use with the invention. Other computer architectures having different configurations of subsystems can also be utilized.

While the prefeπed embodiments of the present invention are described and illustrated herein, it will be appreciated that they are merely illustrative and that modifications can be made to these embodiments without departing from the spirit and scope of the invention. Thus, the invention is intended to be defined only in terms of the following claims.

Claims

CLAIMSWhat is claimed is:
1. A natural language interface for a computer system that interprets natural language user input, the natural language interface comprising: a natural language agent adapted to receive and interpret the natural language user input and adapted to output an output command; and at least one application agent adapted to receive and further interpret the output command from the natural language agent and adapted to output an executable instruction to an application program, the natural language agent including: a syntactic parser adapted to generate a parsed sentence from the natural language user input, a semantic interpreter adapted to generate the output command from the parsed sentence, and an agent communication manager adapted to provide communication between the semantic interpreter and the at least one application agent, each of the at least one application agent including: a semantic task interpreter adapted to generate the executable instruction from the output command of the natural language agent, and at least one application wrapper, each wrapper configured to communicate with a coπesponding application program.
2. The natural language interface of claim 1, wherein the semantic interpreter of the natural language agent includes a semantic mapper adapted to map the parsed sentence to a semantic action as the output command.
3. The natural language interface of claim 1, wherein the natural language agent further includes a speech recognition system adapted to receive and recognize natural language user speech input and to generate the natural language user request therefrom.
4. The natural language interface of claim 1, wherein the natural language agent further includes a dialog manager adapted to provide feedback to the user indicating the natural language agent's understanding of the natural language user input and to interact with the user in natural language to clarify the natural language user input as necessary.
5. The natural language interface of claim 4, wherein the natural language agent further includes a text to speech synthesizer adapted to provide the speech feedback to the user in speech.
6. The natural language interface of claim 1 , wherein the natural language agent further includes an adaptive preference manager adapted to generate default conditions prefeπed by the user, the default conditions being specific to each user and/or common to multiple users.
7. The natural language interface of claim 1, wherein the semantic task interpreter of each application agent further includes a semantic mapper for mapping the output command to a semantic action as the executable instruction.
8. The natural language interface of claim 1 , wherein each of the at least one application agent further includes a dialog manager adapted to provide natural language feedback to the user indicating the application agent's understanding of the natural language user input and to interact with the user in natural language to clarify the natural language user input as necessary.
9. The natural language interface of claim 1 , wherein each of the at least one application agent further includes an adaptive preference manager adapted to generate default conditions prefeπed by the user for the specific application, the default conditions being specific to each user and/or common
to multiple users.
10. The natural language interface of claim 1, comprising one of the at least one application agent for each class of application programs, each class of application programs being selected from the group consisting of electronic mail, fax, letter, file, operating system, address, games, flight simulators, vehicle simulators, naval simulators, sports simulators, war game and strategy simulators, role play simulators, action simulators, personal information manager, printer, calendar, terminal, travel, encyclopedia, image viewer, C++, Basic, graphical user interface, presentation, charting, meeting, scheduler, telephone, voice mail, word processor, spreadsheet, drawing, web, network, mathematical, directory assistance, internet retail sales, common household utility agent, K-12 education, general education, hands-on training for job-based tasks, internet event lookup, internet product-information lookup, internet-based meeting scheduler, and hardware manager.
11. A computer readable medium on which are stored instructions executable on a computer processor, the instructions comprising: receiving natural language user input; generating a parsed sentence from the natural language user input, mapping the parsed sentence into a semantic action; and generating an instruction from the semantic action, the instruction being executable by an application.
12. The computer readable medium of claim 11 , wherein receiving natural language user input includes receiving natural language speech input.
13. The computer readable medium of claim 11 , the instructions further comprising: providing feedback to the user indicating the processor's understanding of the natural language user input; and interacting with the user in natural language to clarify the natural language user input as necessary.
14. The computer readable medium of claim 13 , wherein providing feedback to the user includes providing speech feedback to the user.
15. The computer readable medium of claim 11 , the instructions further comprising generating a set of default conditions for executing the instruction by an application, the default conditions being specific to each user and/or common to multiple users.
16. The computer readable medium of claim 11 , wherein the application is one of one or more applications selected from the group consisting of electronic mail, fax, letter, file, operating system, address, games, flight simulators, vehicle simulators, naval simulators, sports simulators, war game and strategy simulators, role play simulators, action simulators, personal information manager, printer, calendar, terminal, travel, encyclopedia, image viewer, C++, Basic, graphical user interface, presentation, charting, meeting, scheduler, telephone, voice mail, word processor, spreadsheet, drawing, web, network, mathematical, directory assistance, internet retail sales, common household utility agent, K-12 education, general education, hands-on training for job-based tasks, internet event lookup, internet product-information lookup, internet-based meeting scheduler, and hardware manager.
17. The computer readable medium of claim 11 , wherein the computer readable medium is selected from the group consisting of CD-ROM, zip disk, floppy disk, tape, flash memory, system memory, hard drive, and data signal embodied in a carrier wave.
18. A method for receiving, interpreting and executing natural language user input, comprising: receiving natural language user input; generating a parsed sentence from the natural language user input, semantically interpreting the parsed sentence and generating an output command from the parsed sentence, outputting the output command to an application class agent, semantically interpreting the output command and generating an executable instruction from the output command, and outputting the executable instruction to an application program for execution by the application program.
19. The method for receiving, interpreting and executing natural language user input of claim 18, wherein receiving natural language user input includes receiving natural language speech input.
20. The method for receiving, interpreting and executing natural language user input of claim 18, further comprising: providing feedback to the user indicating the processor's understanding of the natural language user input; and interacting with the user in natural language to clarify the natural language user input as necessary.
21. The method for receiving, interpreting and executing natural language user input of claim 20, wherein providing feedback to the user includes providing speech feedback to the user.
22. The method for receiving, interpreting and executing natural language user input of claim 18, further comprising generating a set of default conditions for executing the instruction by an application, the default conditions being specific to each user and/or common to multiple users.
23. The method for receiving, interpreting and executing natural language user input of claim 18, wherein semantically interpreting the parsed sentence and generating the output command includes mapping the parsed sentence into a semantic action as the output command.
24. The method for receiving, interpreting and executing natural language user input of claim 18, wherein semantically interpreting the output command and generating the executable instruction includes mapping the output command into a semantic action as the executable instruction.
25. The method for receiving, interpreting and executing natural language user input of claim 18, wherein the application is one of one or more applications selected from the group consisting of electronic mail, fax, letter, file, operating system, address, games, flight simulators, vehicle simulators, naval simulators, sports simulators, war game and strategy simulators, role play simulators, action simulators, personal information manager, printer, calendar, terminal, travel, encyclopedia, image viewer, C++, Basic, graphical user interface, presentation, charting, meeting, scheduler, telephone, voice mail, word processor, spreadsheet, drawing, web, network, mathematical, directory assistance, internet retail sales, common household utility agent, K-
12 education, general education, hands-on training for job-based tasks, internet event lookup, internet product-information lookup, internet-based meeting scheduler, and hardware manager.
PCT/US1999/019255 1998-08-24 1999-08-20 Adaptive natural language interface WO2000011571A1 (en)

Priority Applications (6)

Application Number Priority Date Filing Date Title
US9763098P true 1998-08-24 1998-08-24
US10542898P true 1998-10-23 1998-10-23
US37271099A true 1999-08-11 1999-08-11
US60/105,428 1999-08-11
US09/372,710 1999-08-11
US60/097,630 1999-08-11

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2000566759A JP2002523828A (en) 1998-08-24 1999-08-20 Adaptive natural language interface
EP19990943868 EP1116134A1 (en) 1998-08-24 1999-08-20 Adaptive natural language interface

Publications (1)

Publication Number Publication Date
WO2000011571A1 true WO2000011571A1 (en) 2000-03-02

Family

ID=27378419

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US1999/019255 WO2000011571A1 (en) 1998-08-24 1999-08-20 Adaptive natural language interface

Country Status (4)

Country Link
EP (1) EP1116134A1 (en)
JP (1) JP2002523828A (en)
CN (1) CN1313972A (en)
WO (1) WO2000011571A1 (en)

Cited By (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2002027712A1 (en) * 2000-09-29 2002-04-04 Professorq, Inc. Natural-language voice-activated personal assistant
EP1199630A2 (en) * 2000-09-14 2002-04-24 Microsoft Method and apparatus for performing planbased dialog
JP2002229591A (en) * 2001-02-02 2002-08-16 Hitachi Ltd Speech interpretation system, and speech interpretation method
GB2402507A (en) * 2003-06-03 2004-12-08 Canon Kk A user input interpreter and a method of interpreting user input
EP1589440A2 (en) 2004-04-23 2005-10-26 Microsoft Corporation Semantic programming language and linguistic object model
US7681186B2 (en) 2004-04-23 2010-03-16 Microsoft Corporation Resolvable semantic type and resolvable semantic type resolution
US7689410B2 (en) 2004-04-23 2010-03-30 Microsoft Corporation Lexical semantic structure
DE102010049869A1 (en) 2010-10-28 2012-05-03 Volkswagen Ag Method for providing voice interface in vehicle, involves determining hit list from stored data depending on assigned category and comparison result
CN102902664A (en) * 2012-08-15 2013-01-30 中山大学 Artificial intelligence natural language operation system on intelligent terminal
WO2013102892A1 (en) * 2012-01-06 2013-07-11 Technologies Of Voice Interface Ltd A system and method for generating personalized sensor-based activation of software
US20130346992A1 (en) * 2012-06-13 2013-12-26 Casio Computer Co., Ltd. Computing system, method for controlling thereof, and computer-readable recording medium having computer program for controlling thereof
WO2014078747A1 (en) * 2012-11-15 2014-05-22 Maxwell Worthington, Llc Natural language command string for controlling target applications
CN104360897A (en) * 2014-10-29 2015-02-18 百度在线网络技术(北京)有限公司 Conversation processing method and conversation management system
US9229974B1 (en) 2012-06-01 2016-01-05 Google Inc. Classifying queries
US9323722B1 (en) * 2010-12-07 2016-04-26 Google Inc. Low-latency interactive user interface
WO2016196701A1 (en) * 2015-06-05 2016-12-08 Microsoft Technology Licensing, Llc Natural language engine for coding and debugging
US9614790B2 (en) 2012-07-25 2017-04-04 Casio Computer Co., Ltd. Apparatus for controlling execution of software, method for controlling thereof, and computer-readable recording medium having computer program for controlling thereof
EP3213224A4 (en) * 2014-10-27 2017-09-06 Identifor Inc. Identifying an individual's abilities, skills and interests through gaming data analytics

Families Citing this family (21)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7613719B2 (en) 2004-03-18 2009-11-03 Microsoft Corporation Rendering tables with natural language commands
CN100461109C (en) 2004-04-28 2009-02-11 富士通株式会社 Semantic task computing
US8725505B2 (en) 2004-10-22 2014-05-13 Microsoft Corporation Verb error recovery in speech recognition
JP4508917B2 (en) * 2005-03-24 2010-07-21 株式会社ケンウッド Information presentation apparatus, information presentation method and an information presentation program
JP4986301B2 (en) * 2008-09-30 2012-07-25 Kddi株式会社 Content search apparatus, program, and method using voice recognition processing function
CN101655716B (en) * 2009-07-17 2011-04-06 赵维 Remote labor division and cooperative system and method comprising idiom
US9760566B2 (en) 2011-03-31 2017-09-12 Microsoft Technology Licensing, Llc Augmented conversational understanding agent to identify conversation context between two humans and taking an agent action thereof
EP2691885A4 (en) * 2011-03-31 2015-09-30 Microsoft Technology Licensing Llc Augmented conversational understanding architecture
US9858343B2 (en) 2011-03-31 2018-01-02 Microsoft Technology Licensing Llc Personalization of queries, conversations, and searches
US9842168B2 (en) 2011-03-31 2017-12-12 Microsoft Technology Licensing, Llc Task driven user intents
US9244984B2 (en) 2011-03-31 2016-01-26 Microsoft Technology Licensing, Llc Location based conversational understanding
US9454962B2 (en) 2011-05-12 2016-09-27 Microsoft Technology Licensing, Llc Sentence simplification for spoken language understanding
CN103534696B (en) * 2011-05-13 2018-02-16 微软技术许可有限责任公司 Click for domain detection using a lookup record spoken language comprehension
US9064006B2 (en) 2012-08-23 2015-06-23 Microsoft Technology Licensing, Llc Translating natural language utterances to keyword search queries
US20140115456A1 (en) * 2012-09-28 2014-04-24 Oracle International Corporation System for accessing software functionality
US9330090B2 (en) * 2013-01-29 2016-05-03 Microsoft Technology Licensing, Llc. Translating natural language descriptions to programs in a domain-specific language for spreadsheets
US9123345B2 (en) * 2013-03-14 2015-09-01 Honda Motor Co., Ltd. Voice interface systems and methods
CN104243666B (en) * 2013-06-13 2017-10-31 腾讯科技(深圳)有限公司 Language processing method and apparatus
CN104301500A (en) * 2013-07-16 2015-01-21 中兴通讯股份有限公司 Terminal control method and device and terminal
US9558176B2 (en) 2013-12-06 2017-01-31 Microsoft Technology Licensing, Llc Discriminating between natural language and keyword language items
CN106328135A (en) * 2016-08-19 2017-01-11 北京智能管家科技有限公司 Application switching method and application switching device based on voice

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4688195A (en) * 1983-01-28 1987-08-18 Texas Instruments Incorporated Natural-language interface generating system
US5555169A (en) * 1992-05-20 1996-09-10 Hitachi, Ltd. Computer system and method for converting a conversational statement to computer command language
US5642519A (en) * 1994-04-29 1997-06-24 Sun Microsystems, Inc. Speech interpreter with a unified grammer compiler
US5748974A (en) * 1994-12-13 1998-05-05 International Business Machines Corporation Multimodal natural language interface for cross-application tasks
US5748841A (en) * 1994-02-25 1998-05-05 Morin; Philippe Supervised contextual language acquisition system

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4688195A (en) * 1983-01-28 1987-08-18 Texas Instruments Incorporated Natural-language interface generating system
US5555169A (en) * 1992-05-20 1996-09-10 Hitachi, Ltd. Computer system and method for converting a conversational statement to computer command language
US5748841A (en) * 1994-02-25 1998-05-05 Morin; Philippe Supervised contextual language acquisition system
US5642519A (en) * 1994-04-29 1997-06-24 Sun Microsystems, Inc. Speech interpreter with a unified grammer compiler
US5748974A (en) * 1994-12-13 1998-05-05 International Business Machines Corporation Multimodal natural language interface for cross-application tasks

Cited By (27)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1199630A2 (en) * 2000-09-14 2002-04-24 Microsoft Method and apparatus for performing planbased dialog
EP1199630A3 (en) * 2000-09-14 2005-02-02 Microsoft Corporation Method and apparatus for performing planbased dialog
US7398211B2 (en) 2000-09-14 2008-07-08 Microsoft Corporation Method and apparatus for performing plan-based dialog
WO2002027712A1 (en) * 2000-09-29 2002-04-04 Professorq, Inc. Natural-language voice-activated personal assistant
US7216080B2 (en) 2000-09-29 2007-05-08 Mindfabric Holdings Llc Natural-language voice-activated personal assistant
JP2002229591A (en) * 2001-02-02 2002-08-16 Hitachi Ltd Speech interpretation system, and speech interpretation method
GB2402507A (en) * 2003-06-03 2004-12-08 Canon Kk A user input interpreter and a method of interpreting user input
EP1589440A2 (en) 2004-04-23 2005-10-26 Microsoft Corporation Semantic programming language and linguistic object model
EP1589440A3 (en) * 2004-04-23 2008-08-13 Microsoft Corporation Semantic programming language and linguistic object model
US7681186B2 (en) 2004-04-23 2010-03-16 Microsoft Corporation Resolvable semantic type and resolvable semantic type resolution
US7689410B2 (en) 2004-04-23 2010-03-30 Microsoft Corporation Lexical semantic structure
US7761858B2 (en) 2004-04-23 2010-07-20 Microsoft Corporation Semantic programming language
US8201139B2 (en) 2004-04-23 2012-06-12 Microsoft Corporation Semantic framework for natural language programming
DE102010049869A1 (en) 2010-10-28 2012-05-03 Volkswagen Ag Method for providing voice interface in vehicle, involves determining hit list from stored data depending on assigned category and comparison result
US9323722B1 (en) * 2010-12-07 2016-04-26 Google Inc. Low-latency interactive user interface
WO2013102892A1 (en) * 2012-01-06 2013-07-11 Technologies Of Voice Interface Ltd A system and method for generating personalized sensor-based activation of software
US9229974B1 (en) 2012-06-01 2016-01-05 Google Inc. Classifying queries
US20130346992A1 (en) * 2012-06-13 2013-12-26 Casio Computer Co., Ltd. Computing system, method for controlling thereof, and computer-readable recording medium having computer program for controlling thereof
CN106406999A (en) * 2012-06-13 2017-02-15 卡西欧计算机株式会社 A computing system and a method for controlling thereof
US9614790B2 (en) 2012-07-25 2017-04-04 Casio Computer Co., Ltd. Apparatus for controlling execution of software, method for controlling thereof, and computer-readable recording medium having computer program for controlling thereof
CN102902664A (en) * 2012-08-15 2013-01-30 中山大学 Artificial intelligence natural language operation system on intelligent terminal
WO2014078747A1 (en) * 2012-11-15 2014-05-22 Maxwell Worthington, Llc Natural language command string for controlling target applications
EP3213224A4 (en) * 2014-10-27 2017-09-06 Identifor Inc. Identifying an individual's abilities, skills and interests through gaming data analytics
US10315118B2 (en) 2014-10-27 2019-06-11 Identifor, Inc. Identifying an individual's abilities, skills and interests through gaming data analytics
CN104360897A (en) * 2014-10-29 2015-02-18 百度在线网络技术(北京)有限公司 Conversation processing method and conversation management system
CN104360897B (en) * 2014-10-29 2017-09-22 百度在线网络技术(北京)有限公司 Dialogue approach and dialogue management system
WO2016196701A1 (en) * 2015-06-05 2016-12-08 Microsoft Technology Licensing, Llc Natural language engine for coding and debugging

Also Published As

Publication number Publication date
JP2002523828A (en) 2002-07-30
EP1116134A1 (en) 2001-07-18
CN1313972A (en) 2001-09-19

Similar Documents

Publication Publication Date Title
US5721938A (en) Method and device for parsing and analyzing natural language sentences and text
Shawar et al. Chatbots: are they really useful?
Allen Natural language processing
CN1156821C (en) Recognition engines with complementary language models
RU2352979C2 (en) Synchronous comprehension of semantic objects for highly active interface
KR100563365B1 (en) Hierarchical Language Model
US6088731A (en) Intelligent assistant for use with a local computer and with the internet
US7174507B2 (en) System method and computer program product for obtaining structured data from text
US5377103A (en) Constrained natural language interface for a computer that employs a browse function
US7983997B2 (en) Interactive complex task teaching system that allows for natural language input, recognizes a user's intent, and automatically performs tasks in document object model (DOM) nodes
Allen et al. An architecture for a generic dialogue shell
Seneff et al. Galaxy-II: A reference architecture for conversational system development
Katz From sentence processing to information access on the world wide web
US20110231182A1 (en) Mobile systems and methods of supporting natural language human-machine interactions
Bates Models of natural language understanding
US6665640B1 (en) Interactive speech based learning/training system formulating search queries based on natural language parsing of recognized user queries
US7873654B2 (en) Multimodal natural language query system for processing and analyzing voice and proximity-based queries
Shamsfard et al. The state of the art in ontology learning: a framework for comparison
JP4112171B2 (en) The system and method using a natural language understanding for voice control applications
US20130173519A1 (en) Method and computer program product for providing a response to a statement of a user
Oviatt et al. Designing the user interface for multimodal speech and pen-based gesture applications: state-of-the-art systems and future research directions
US6785651B1 (en) Method and apparatus for performing plan-based dialog
US20040148170A1 (en) Statistical classifiers for spoken language understanding and command/control scenarios
US4914590A (en) Natural language understanding system
US7580842B1 (en) System and method of providing a spoken dialog interface to a website

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 99810028.5

Country of ref document: CN

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE

AK Designated states

Kind code of ref document: A1

Designated state(s): CN JP RU

121 Ep: the epo has been informed by wipo that ep was designated in this application
DFPE Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101)
WWE Wipo information: entry into national phase

Ref document number: 1999943868

Country of ref document: EP

WWP Wipo information: published in national office

Ref document number: 1999943868

Country of ref document: EP

WWW Wipo information: withdrawn in national office

Ref document number: 1999943868

Country of ref document: EP