CN101789008A - Man-machine interface system knowledge base and construction method thereof - Google Patents

Abstract

The invention discloses a human-machine interface system knowledge base, wherein the first corpus is used to store the corpus of dialogue initiated by the user; the second corpus is used to store the corpus of returned dialogue in different fields; Extracting the word corpus of the corresponding field, and sending the extracted word corpus of the corresponding field to the second corpus; the matching processing unit is used to match the corpus of the user-initiated dialogue with the corpus of the first corpus, and obtain the matching dialogue initiation corpus, And matching the dialogue initiation corpus with the corpus in the second corpus to obtain the matched dialogue return corpus; the feedback unit is used to feed back the matched dialogue return corpus to the user. The invention can realize the specificity of the conversation between the user and the chat robot, control the conversation topic in a relatively special field, and the two corpus jointly form a knowledge base, so as to achieve the separation of form and content. The invention also provides a method for constructing the knowledge base of the man-machine interface system.

Description

Human-machine interface system knowledge base and its construction method

technical field

The invention relates to the fields of man-machine interface technology and natural language processing, in particular to a man-machine interface system knowledge base and a construction method thereof.

Background technique

Human-machine interface systems such as Jabberwacky and ALICEBOT are mainly used in the field of human-computer dialogue, and are usually called chatbots. Chatbots are mainly designed to find ways to allow people to communicate with machines. The method for the chatbot to communicate with the user is to match the user's input with the knowledge base stored by the chatbot, and then return the matching result to the user immediately. Since the matching sentences in the knowledge base of the chatbot are quite extensive and the dialogue domain is not divided, the dialogue content returned to the user is also quite extensive, and it is easy to divert the user's attention to other topics.

Therefore, it is necessary to provide an improved human-machine interface system knowledge base and its construction method to overcome the defects of the prior art.

Contents of the invention

The purpose of the present invention is to provide a human-machine interface system knowledge base and its construction method, which can limit the dialogue field between the user and the chat robot.

In order to achieve the above object, the present invention provides a human-machine interface system knowledge base including a first corpus, a second corpus, a returned corpus extraction unit, a matching processing unit and a feedback unit. The first corpus is used to store the corpus of the dialogue initiated by the user; the second corpus is used to store the corpus of the returned dialogue by domain; the returned corpus extraction unit is connected to the second corpus for Extracting the word corpus of the corresponding field, and sending the extracted word corpus of the corresponding field to the second corpus; the matching processing unit is connected with the first corpus and the second corpus, and is used to connect the The corpus of the dialogue is matched with the corpus in the first corpus, the matching dialogue initiation corpus is obtained, and the dialogue initiation corpus is matched with the corpus in the second corpus, and the matching dialogue return corpus is obtained; The feedback unit is connected to the matching processing unit, and is configured to feed back the matched dialog return corpus to the user.

In one embodiment of the present invention, the man-machine interface system knowledge base further includes a dialogue material collection unit, which is connected to the first corpus and is used to conduct dialogue experiments on users and collect experimental dialogues Initiating corpus, formalizing and summarizing the dialogue initiating corpus whose usage frequency is higher than a predetermined threshold frequency, and sending the formalized and summarized dialogue initiating corpus to the first corpus.

In another embodiment of the present invention, the return corpus extraction unit includes a first-level return corpus extraction unit and a second-level return corpus extraction unit. The first-level return corpus extraction unit is used to extract sentences in corresponding fields according to the knowledge documents in each field; the second-level return corpus extraction unit is connected with the first-level return corpus extraction unit and the second corpus, using Extract the word corpus in the corresponding field according to the sentences in the corresponding field extracted by the first-level returned corpus extraction unit, and formally classify the extracted word corpus in the corresponding field, and send the formally classified word corpus to the The second corpus, the word corpus after the formal classification is the corpus of returned dialogue.

In yet another embodiment of the present invention, the categories of the formal classification are "name", "behavior and action", "modification", "position and time" and "pure grammar", and the second corpus classifies and saves The word corpus after the formal classification of the corresponding field.

In yet another embodiment of the present invention, the human-machine interface system knowledge base further includes a natural language generation system, and the natural language generation system is connected to the matching processing unit and the feedback unit for converting the matching The dialogue returned corpus is converted into natural language, and the result of the conversion is fed back to the user.

A method for constructing a knowledge base of a human-machine interface system, comprising the steps of: storing corpus of dialogue initiated by a user; extracting word corpus in a corresponding field according to knowledge documents in various fields; classifying and storing the extracted word corpus in a corresponding field, and storing the word corpus in the corresponding field The word corpus is used as the corpus of the returned dialogue; the corpus of the user-initiated dialogue is matched with the corpus of the stored user-initiated dialogue, the matching dialogue is initiated, and the dialogue is initiated with the stored returned dialogue corpus Matching, obtaining the matched dialogue return corpus; feeding back the matched dialogue return corpus to the user.

In one embodiment of the present invention, the method for constructing the knowledge base of the human-machine interface system further includes: conducting a dialogue experiment on users, collecting experimental dialogue initiation corpus, and performing a form test on dialogue initiation corpus whose use frequency is higher than a specified threshold frequency. Induction. The step of storing the corpus of dialogue initiated by the user is specifically: storing the formalized and summarized corpus of dialogue initiation.

In another embodiment of the present invention, the step of extracting the word corpus in the corresponding field according to the knowledge documents in each field is specifically: extracting the sentences in the corresponding field according to the knowledge documents in each field; extracting the words in the corresponding field according to the extracted sentences in the corresponding field Word corpus: formally classify the extracted word corpus in the corresponding field, and the formally classified word corpus is the corpus for returning the dialogue.

In yet another embodiment of the present invention, the step of formally classifying the extracted word corpus in the corresponding field is specifically: according to "title", "behavior and action", "modification", "location and time" and The "Pure Grammar" category formally classifies the extracted word corpus of the corresponding domain. The step of storing the extracted word corpus in the corresponding field specifically includes: classifying and storing the formally classified word corpus in the corresponding field.

In yet another embodiment of the present invention, the step of feeding back the matched dialogue return corpus to the user is specifically: converting the matched dialogue return corpus into natural language; feeding back the result of the conversion to the user .

Compared with the prior art, the second corpus of the human-machine interface system knowledge base of the present invention is domain-specific, so the user has specificity when talking with the chat robot, and can control the dialogue topic in a relatively specialized domain, thereby Pass the professional knowledge points in the field to users through dialogue as much as possible.

In addition, the knowledge base of the man-machine interface system of the present invention establishes the form of knowledge through the first corpus, and establishes the content of knowledge through the second corpus, and the two corpora jointly form the knowledge base to achieve the separation of form and content.

The present invention will become clearer through the following description in conjunction with the accompanying drawings, which are used to explain the embodiments of the present invention.

Description of drawings

Fig. 1 is a structural block diagram of the knowledge base of the man-machine interface system of the present invention.

Fig. 2 is a flow chart of the method for constructing the knowledge base of the human-machine interface system of the present invention.

Detailed ways

Embodiments of the present invention will now be described with reference to the drawings, in which like reference numerals represent like elements.

The human-machine interface system knowledge base in this embodiment includes a first corpus 20 , a dialogue corpus collection unit 10 , a second corpus 30 , a returned corpus extraction unit 40 , a matching processing unit 50 , a feedback unit 70 and a natural language generation system 60 .

The first corpus 20 is used to store the corpus of dialogue initiated by the user;

The dialogue material collection unit 10 is connected with the first corpus 20, and is used to carry out dialogue experiments to users through chat tools such as chat robot platforms, frequently asked questions (FAQ, Frequently asked question), user questionnaires, etc., and collect experiments. The dialogue initiation corpus is formalized and summarized for the dialogue initiation corpus whose use frequency is higher than the prescribed threshold frequency, and the formalized and summarized dialogue initiation corpus is sent to the first corpus 20 . Among them, when conducting experiments on users, the more people are tested and the more dialogue materials are retained, the higher the success rate of subsequent matching will be.

The second corpus 30 is used to store the corpus of the returned dialogue by domain.

The return corpus extracting unit 40 is connected to the second corpus 30, and is used to extract word corpus in the corresponding field according to the knowledge documents in each field, and send the extracted word corpus in the corresponding field to the second corpus 30 ;

Wherein, the return corpus extraction unit 40 includes a first-level return corpus extraction unit and a second-level return corpus extraction unit. The first-level return corpus extraction unit is used to extract sentences in corresponding fields according to the knowledge documents in each field; the second-level return corpus extraction unit is connected with the first-level return corpus extraction unit and the second corpus 30, and is used to extract sentences according to the The sentence in the corresponding field extracted by the first-level return corpus extraction unit extracts the word corpus in the corresponding field, and formally classifies the word corpus in the corresponding field extracted, and sends the word corpus after the formal classification to the second The corpus 30, the word corpus after the formal classification is the corpus of the returned dialogue. Wherein, the formal classification is to add additional information head characters to the extracted word corpus corresponding to the field.

As can be seen from the above, the returned corpus extracting unit 40 breaks up the entire descriptions of knowledge documents in various fields into sentences of conversations, and then breaks them into zeros, extracts the word corpus that meets the above classification in the sentences, and performs The formalized classification is then sent to the second corpus 30 for storage.

Wherein, the categories of the formal classification are "name", "behavior and action", "modification", "orientation and time" and "pure grammar", and the second corpus 30 classifies and saves the formalization of the corresponding field Classified word corpus.

The matching processing unit 50 is connected to the first corpus 20 and the second corpus 30, and is used to match the corpus of the user-initiated dialogue with the corpus in the first corpus 20, and obtain the matched dialogue initiation corpus, And matching the dialogue initiation corpus with the corpus in the second corpus 30 to obtain the matched dialogue return corpus. The matching processing unit 50 establishes matching rules through XML (Extensible Markup Language, Extensible Markup Language) and RegExp (Regular Expression, regular expression), and performs matching based on the established matching rules.

The natural language generating system 60 is connected to the matching processing unit 50 for converting the matched dialogue return corpus into natural language, and sending the converted result to the feedback unit 70 .

The feedback unit 70 is connected with the natural language generation system 60, and is used for feeding back the conversion result of the natural language generation system 60 to the user.

As can be seen from the above, the man-machine interface system knowledge base of the present invention adopts two separate corpora-the first corpus and the second corpus store respectively the formal corpus (dialogue initiation corpus) and the content of the knowledge contained in the dialogue process (dialogue return corpus) ). Specifically, the present invention establishes the form of knowledge expression through the first corpus 20 , and establishes the content of knowledge through the second corpus 30 , and the two corpora jointly form a knowledge base to achieve the separation of form and content.

In addition, the second corpus 30 of the knowledge base of this system is divided into domains, so the user has specificity when talking with the chat robot, and can control the conversation topic in a relatively specialized domain, thereby as much as possible. Knowledge points are delivered to users in the form of dialogue. It can be understood that the knowledge base established by the present invention can quickly develop various applications, such as: question and answer learning system, advertisement recommendation system and so on. Different from general chatbots, the knowledge base generated by this invention is only applicable to a specific field and only for specific topics. Therefore, users cannot disperse their attention to other places, preventing users from learning knowledge from becoming aimless small talk. Moreover, since the content of the knowledge base is collected in different fields, knowledge in different fields can be continuously added later. Therefore, the knowledge base model has scalability.

As shown in Figure 2, a method for constructing a human-machine interface system knowledge base includes the following steps:

Step S10, conduct dialogue experiments on users in the form of chat tools such as chat robot platforms, frequently asked questions (FAQ, Frequently Asked Question), user questionnaires, etc., collect experimental dialogue initiation corpus, and initiate dialogues with frequency of use higher than the specified threshold frequency The corpus is formally summarized;

Step S20, storing the dialogue initiation corpus after formalization and induction;

Step S30, extracting sentences corresponding to the domain according to the knowledge documents in each domain;

Step S40, extracting word corpus in the corresponding field according to the extracted sentences in the corresponding field;

Step S50, according to the category of "name", "behavior and action", "modification", "orientation and time" and "pure grammar", formally classify the extracted word corpus in the corresponding field, and save the formalization of the corresponding field The word corpus after classification, the word corpus after the formal classification of described corresponding field is used as the corpus of returning dialogue;

Step S60, matching the user-initiated dialogue corpus with the stored user-initiated dialogue corpus, obtaining the matched dialogue-initiating corpus, and matching the dialogue-initiating corpus with the stored returned dialogue corpus, obtaining the matched dialogue return corpus;

Step S70, converting the matched dialogue return corpus into natural language, that is, constructing a natural sentence based on the dialogue form in the first corpus and the matched knowledge content in the second corpus;

Step S80, feeding back the conversion result to the user.

The present invention has been described above in conjunction with the best embodiments, but the present invention is not limited to the above-disclosed embodiments, but should cover various modifications and equivalent combinations made according to the essence of the present invention.

Claims (10)

1. A human-machine interface system knowledge base, comprising: The first corpus is used to store the corpus of the conversation initiated by the user; The second corpus is used to store the corpus of the returned dialogue by domain; Returning to the corpus extraction unit, connected to the second corpus, for extracting the word corpus in the corresponding field according to the knowledge documents in each field, and sending the extracted word corpus in the corresponding field to the second corpus; A matching processing unit, connected to the first corpus and the second corpus, is used to match the corpus of the user-initiated dialogue with the corpus in the first corpus, obtain the matched dialogue initiation corpus, and send the dialogue matching the initiating corpus with the corpus in the second corpus, and obtaining the matched dialogue return corpus; and A feedback unit, connected to the matching processing unit, configured to feed back the matched dialog return corpus to the user. 2. The man-machine interface system knowledge base as claimed in claim 1, is characterized in that, also comprises: The dialogue material collection unit is connected with the first corpus, and is used for conducting dialogue experiments on users, collecting experimental dialogue initiation corpora, performing formal induction on dialogue initiation corpora whose use frequency is higher than the specified threshold frequency, and sending the formal induction The subsequent dialogue initiation corpus is transferred to the first corpus. 3. the man-machine interface system knowledge base as claimed in claim 1, is characterized in that, described return corpus extracting unit comprises: The first level returns the corpus extraction unit, which is used to extract sentences in the corresponding field according to the knowledge documents in each field; The second-level return corpus extraction unit is connected with the first-level return corpus extraction unit and the second corpus, and is used to extract the word corpus in the corresponding field according to the sentences in the corresponding field extracted by the first-level return corpus extraction unit , and formally classify the extracted word corpus corresponding to the field, and send the formally classified word corpus to the second corpus, where the formally classified word corpus is the corpus for returning the dialogue. 4. The human-machine interface system knowledge base according to claim 3, wherein the categories of the formal classification are "name", "behavior and action", "modification", "orientation and time" and "pure Grammar", the second corpus classifies and saves the formally classified word corpus of the corresponding field. 5. The human-machine interface system knowledge base as claimed in claim 1, further comprising: A natural language generation system, connected to the matching processing unit and the feedback unit, is used to convert the matched dialog return corpus into natural language, and feed back the conversion result to the user. 6. A method for constructing a human-machine interface system knowledge base, comprising the steps of: Store the corpus of conversations initiated by users; Extract the word corpus in the corresponding field according to the knowledge documents in each field; Classify and store the word corpus in the corresponding field extracted, and use the word corpus in the corresponding field as the corpus for returning the dialogue; Matching the user-initiated dialogue corpus with the stored user-initiated dialogue corpus, obtaining the matched dialogue initiation corpus, and matching the dialogue initiation corpus with the stored returned dialogue corpus, obtaining the matched dialogue return corpus ;as well as Feedback the matched dialog return corpus to the user. 7. The human-machine interface system knowledge base construction method as claimed in claim 6, is characterized in that, also comprises: Conduct dialogue experiments on users, collect experimental dialogue data, and formally summarize the dialogue initiation corpus whose frequency of use is higher than the specified threshold frequency. Then, the step of storing the corpus of the user-initiated dialogue is specifically: The dialog initiation corpus after formalization and induction is stored. 8. The human-machine interface system knowledge base construction method as claimed in claim 6, wherein the step of extracting the word corpus of the corresponding field according to the knowledge documents in each field is specifically: According to the knowledge documents in various fields, sentences in the corresponding fields are extracted; Extracting word corpus in the corresponding field according to the extracted sentences in the corresponding field; Perform formal classification on the extracted word corpus corresponding to the field, and the formally classified word corpus is returned dialogue corpus. 9. The human-machine interface system knowledge base construction method as claimed in claim 8, wherein the step of formally classifying the word corpus of the extracted corresponding field is specifically: According to the categories of "name", "behavior and action", "modification", "position and time" and "pure grammar", the extracted word corpus in the corresponding field is formally classified, Then, the step of storing and extracting the word corpus in the corresponding field is specifically: The classification saves the formally classified word corpus of the corresponding field. 10. The method for constructing a human-machine interface system knowledge base as claimed in claim 6, wherein the step of feeding back the matched dialogue corpus to the user is specifically: Convert the matched dialogue return corpus into natural language; The result of the conversion is fed back to the user.

Images