CN111159382A - Method and device for constructing and using session system knowledge model - Google Patents
Method and device for constructing and using session system knowledge model Download PDFInfo
- Publication number
- CN111159382A CN111159382A CN201911407443.1A CN201911407443A CN111159382A CN 111159382 A CN111159382 A CN 111159382A CN 201911407443 A CN201911407443 A CN 201911407443A CN 111159382 A CN111159382 A CN 111159382A
- Authority
- CN
- China
- Prior art keywords
- knowledge
- user
- model
- question
- tree
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
Abstract
The invention discloses a method and a device for constructing and using a knowledge model of a session system, wherein the method for constructing the knowledge model of the session system comprises the following steps: abstracting, by a knowledge point in a conversational system, at least one topic associated with the knowledge point; building a system topic tree based on the at least one topic; determining an incidence relation between knowledge points based on the system theme tree; and constructing a conversation system knowledge model at least based on the incidence relation between the system topic tree and the knowledge points. The method of the embodiment abstracts the knowledge points into at least one theme, then constructs a theme tree by using the theme, determines the incidence relation between the knowledge points according to the theme tree, and then can construct a session system knowledge model for recommending questions or answers to the user in the subsequent heuristic session process, thereby recommending the questions or contents more conforming to the user interests to the user.
Description
Technical Field
The invention belongs to the technical field of session systems, and particularly relates to a method and a device for constructing and using a session system knowledge model.
Background
In the related art, in the application scenario of a general conversation system, such as a smart speaker, a smart television, a device installed in a vehicle, etc., when a user says a question, the smart customer service automatically asks and answers and converses, the basic flow of such a system is that it receives a question from the user or a sentence spoken by the user, and then some processing is performed in the system to give a reply to the user.
Heuristic sessions allow sessions to persist by obtaining some connections behind the problem.
The inventor discovers that in the process of implementing the application: the existing products or technologies are mainly heuristic sessions based on themes or scenes, and the interest points of the users are not considered.
Disclosure of Invention
The embodiment of the invention provides a method and a device for constructing and using a session system knowledge model, which are used for solving at least one of the technical problems.
In a first aspect, an embodiment of the present invention provides a method for constructing a knowledge model of a session system, including: abstracting, by a knowledge point in a conversational system, at least one topic associated with the knowledge point; building a system topic tree based on the at least one topic; determining an incidence relation between knowledge points based on the system theme tree; and constructing a conversation system knowledge model at least based on the incidence relation between the system theme tree and the knowledge points.
In a second aspect, an embodiment of the present invention provides a method for using a session system knowledge model, including: constructing a user interest model based on basic information of a user and historical conversation data of the user, wherein the user interest model comprises at least one knowledge point; matching knowledge points in the user interest model with knowledge points in the session system knowledge model constructed according to the method of the first aspect based on a semantic graph; generating at least one matched knowledge point according to a matching result of the semantic graph, wherein the knowledge point is associated with at least one question; and recommending the question associated with the at least one matched knowledge point to the user.
In a third aspect, an embodiment of the present invention provides a device for constructing a knowledge model of a session system, including: the knowledge point abstraction module is configured to abstract at least one theme related to the knowledge points from the knowledge points in the session system; a subject tree construction module configured to construct a system subject tree based on the at least one subject; the incidence relation determining module is configured to determine incidence relations among the knowledge points based on the system theme tree; and the knowledge model building module is configured to build a session system knowledge model at least based on the incidence relation between the system topic tree and each knowledge point.
In a fourth aspect, an embodiment of the present invention provides an apparatus for using a session system knowledge model, including: the user interest model building module is configured to build a user interest model based on basic information of a user and historical conversation data of the user, wherein the user interest model comprises at least one knowledge point; a semantic graph matching module configured to perform semantic graph-based matching on the knowledge points in the user interest model and the knowledge points in the session system knowledge model constructed according to the method of the first aspect; the matching knowledge point generating module is configured to generate at least one matching knowledge point according to a matching result of the semantic graph, wherein the knowledge point is associated with at least one question; and a recommending module configured to recommend the question associated with the at least one matched knowledge point to the user.
In a fifth aspect, an electronic device is provided, comprising: at least one processor, and a memory communicatively coupled to the at least one processor, wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the steps of the method of constructing a conversational system knowledge model of any embodiment of the invention.
In a sixth aspect, the present invention further provides a computer program product, where the computer program product includes a computer program stored on a non-volatile computer-readable storage medium, and the computer program includes program instructions, which, when executed by a computer, make the computer execute the steps of the method for constructing a session system knowledge model according to any embodiment of the present invention.
The method of the embodiment abstracts the knowledge points into at least one theme, then constructs a theme tree by using the theme, determines the incidence relation between the knowledge points according to the theme tree, and then can construct a session system knowledge model for recommending questions or answers to the user in the subsequent heuristic session process, thereby recommending the questions or contents more conforming to the user interests to the user.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on the drawings without creative efforts.
Fig. 1 is a flowchart of a method for constructing a knowledge model of a session system according to an embodiment of the present invention;
FIG. 2 is a flowchart of another method for constructing a knowledge model of a session system according to an embodiment of the present invention;
FIG. 3 is a flow chart of a method for using a knowledge model of a session system according to an embodiment of the present invention;
FIG. 4 is a flow chart of another method for using a knowledge model of a conversational system according to an embodiment of the invention;
FIG. 5 is a flow chart of a method for using a knowledge model of a conversational system according to an embodiment of the invention;
FIG. 6 is a detailed flow chart of the system according to an embodiment of the present invention;
FIG. 7 is a diagram illustrating a specific example of a topic and knowledge point based user interest point model structure according to an embodiment of the present invention;
FIG. 8 is a semantic representation of a user question and a semantic representation of a dialog system knowledge system provided by an embodiment of the present invention;
fig. 9 is a block diagram of an apparatus for constructing a knowledge model of a session system according to an embodiment of the present invention;
FIG. 10 is a block diagram of an apparatus for using a knowledge model of a conversational system according to an embodiment of the invention;
fig. 11 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, which shows a flowchart of an embodiment of a method for constructing a knowledge model of a session system according to the present application, the method for constructing a knowledge model of a session system according to the present embodiment may be applied to a terminal with an intelligent voice conversation function, such as an intelligent children story machine, an intelligent conversation toy, a device including an intelligent story playing, and the like.
As shown in fig. 1, in step 101, abstracting at least one topic associated with a knowledge point by the knowledge point in the conversational system;
in step 102, a system topic tree is constructed based on at least one topic;
in step 103, determining the incidence relation between knowledge points based on the system theme tree;
in step 104, a conversational system knowledge model is constructed based at least on the association between the system topic tree and the knowledge points.
In this embodiment, for step 101, the construction apparatus of the knowledge model of the session system abstracts at least one topic associated with the knowledge point from the knowledge point in the session system, wherein the knowledge point may be, for example, pruni, which may be abstracted to be poetry or a game character, thereby constructing a topic, which is, for example, poetry or a game character.
Thereafter, for step 102, a system topic tree is constructed based on at least one topic, such as a topic tree constructed from poetry of Dopu et al derived from poetry Libai. Or a theme tree constructed by deriving all large game characters such as Direnjie from the Libai of the game characters.
Then, for step 103, the association relationship between the knowledge points is determined based on the system theme tree, for example, the association relationship between the poems such as liban and dupu is determined based on the theme tree of poems, or the association relationship between the poems such as liban and dironey is determined based on the theme tree of game characters.
Finally, for step 104, a knowledge model of the conversational system is constructed based on at least the associations between the system topic tree and the knowledge points, for example, the associations between the poetry topic tree and the knowledge points, wherein the knowledge model is an abstraction that formalizes and structures knowledge.
The method of the embodiment abstracts the knowledge points into at least one theme, then constructs a theme tree by using the theme, determines the incidence relation between the knowledge points according to the theme tree, and then can construct a session system knowledge model for recommending questions or answers to the user in the subsequent heuristic session process, thereby recommending the questions or contents more conforming to the user interests to the user.
With further reference to fig. 2, a flowchart of another method for constructing a knowledge model of a session system according to an embodiment of the present application is shown. The flow chart is primarily a flow chart for the steps further defined in step 103 of the flow chart 1. The knowledge points comprise question-answer pairs and a knowledge graph, the question-answer pairs comprise first entities, and the knowledge graph comprises second entities and incidence relations among the second entities.
As shown in fig. 2, in step 201, when a first entity and a second entity in a topic tree are the same, performing associative fusion on the first entity and the second entity;
in step 202, the association relationship between the knowledge points is determined based on the association-fused topic tree.
In this embodiment, for step 201, when the first entity and the second entity in the topic tree are the same, the constructing apparatus of the session system knowledge model performs association fusion on the first entity and the second entity, for example, a poetry living in the topic tree is taken as the first entity, a movie role living in the topic tree is taken as the second entity, and then performs association fusion on them, so as to form a fused topic tree;
finally, for step 202, the association relationship between the knowledge points is determined based on the association-fused topic tree, for example, the association relationship between the knowledge points is determined based on the poetry-video role-fused topic tree to construct a session system knowledge model.
In the scheme of the embodiment, the relationship fusion is performed on the entities with the same theme tree species, and the association relationship between the knowledge points is determined, so that the relationship between the knowledge points can be tighter by performing the fusion on the same entities, and the relationship between two original knowledge points without relationship is established.
Referring to fig. 3, a flowchart of an embodiment of a method for using a knowledge model of a session system provided in the present application is shown.
As shown in fig. 3, in step 301, a user interest model is constructed based on basic information of a user and historical conversation data of the user, wherein the user interest model includes at least one knowledge point;
in step 302, semantic graph-based matching is carried out on the knowledge points in the user interest model and the knowledge points in the session system knowledge model constructed by the method of the above embodiment;
in step 303, generating at least one matched knowledge point according to the matching result of the semantic graph, wherein the knowledge point is associated with at least one question;
in step 304, questions associated with at least one matching knowledge point are recommended to the user.
In this embodiment, for step 301, the construction device of the session system knowledge model collects basic information such as age and sex of the user and historical session data of the user to construct a user interest model based on the basic information of the user and the historical session data of the user, for example, to construct a Tang poetry interest model of the user based on data such as "Libai", "Dufu" and the like appearing in the historical session of the user, or to construct a Song word interest model of the user based on data such as "sushi", "Liqing" and the like in the historical session of the user.
Thereafter, semantic graph based matching is performed on the knowledge points in the user interest model and the knowledge points in the knowledge model of the conversational system constructed according to the method of claim 1 or 2 for step 302, then at least one matched knowledge point is generated from the matching result of the semantic graph for step 303, and finally, for step 304, the question associated with the at least one matched knowledge point is recommended to the user.
For example, according to the method for constructing a knowledge model in the above embodiment, a knowledge model is created for "poetry of the user" in the interest model, then the knowledge points in the "poetry of the user" are matched based on the semantic graph, and then, for example, "zhangjiu age" is generated according to the matching result of the semantic graph. And finally recommending the questions related to the at least one matched knowledge point of Zhang Jiu age, Du mu and the like to the user.
According to the method, the user interest model is built according to the existing data of the user, and then the problem matched with the knowledge point in the user interest model is recommended to the user, so that the problem which is more in line with the user interest can be recommended to the user according to the user interest model built based on the user data, the matching degree of the recommended problem and the user is higher, and the user experience is better.
With further reference to FIG. 4, shown is a flow chart of another embodiment of a method for using a knowledge model of a conversational system as provided herein. The flow chart is mainly a flow chart of the further defined steps of "building a user interest model based on the basic information of the user and the historical conversation data of the user" in step 301 of the flow chart 3. The historical conversation data of the user comprises active questions and answers of the user and system recommendation knowledge and answers accepted by the user.
As shown in fig. 4, in step 401, hierarchical clustering is performed on the user basic information and the historical conversation data of the user, and a user topic tree is generated;
in step 402, constructing the user's active questions and answers and the system recommended knowledge and answers the user accepts into knowledge triples containing questions, answers and associations between the questions and answers, wherein the questions and answers form question-answer pairs;
in step 403, associating the knowledge triples to the user topic tree;
in step 404, mapping the entities in the question-answer pairs in the triples to the knowledge triples of the knowledge model of the session system by using an entity linking technology, and adding the mapped knowledge triples to the knowledge points under the user topic tree to form a user interest model;
in step 405, semantic similarity between the question-answer pairs is calculated and added to the user interest model.
In this embodiment, for step 401, the constructing apparatus of the knowledge model of the session system performs hierarchical clustering on the basic information of the gender, the age, etc. of the user and the historical session data of the user to generate the user topic tree, wherein the hierarchical clustering attempts to divide the data sets at different levels, so as to form a tree-shaped clustering structure. The data set partitioning may employ a "bottom-up" aggregation strategy or a "top-down" splitting strategy. For example, "poetry of the Tang Dynasty" is a cluster, and the next-layer cluster includes clusters of "Libai", "Dufu", "Zhangjiu" and the like, and for example, the cluster of "poetry of the Tang Dynasty" includes clusters of "poetry of the Tang Dynasty", "Song dynasty words" and the like, and so on, and thus, the description thereof is omitted.
Then, for step 402, the construction device of the session system knowledge model constructs the questions and answers actively asked by the user and the system recommendation knowledge and answers received by the user into knowledge triplets containing the questions, answers and the association relations between the questions and answers, for example, the user actively asks the equipment about "what is the standing night of the plum leaves", then the equipment also asks the user whether to receive other poems of the plum leaves recommended by the system after giving the answers, if the user receives the system recommendation knowledge, the questions, answers and the system recommendation knowledge compose a knowledge triplet, wherein the questions and the answers compose question-answer pairs;
then, for step 403, the construction device of the knowledge model of the session system associates the knowledge triples to the user topic tree;
then, for step 404, the constructing device of the knowledge model of the session system maps the entities in the question-answer pairs in the triples to the knowledge triples of the knowledge model of the session system by using an entity linking technology, and adds the mapped knowledge triples to the knowledge points under the user topic tree to form a user interest model; among them, the entity linking technique is an important method for solving the ambiguity problem of named entities, and the method realizes the elimination of entity ambiguity by linking ambiguous entity reference chains to a given knowledge base.
Finally, for step 405, calculating semantic similarity between question-answer pairs and adding the semantic similarity into the user interest model.
The method of the embodiment forms a user topic tree by performing hierarchical clustering on some existing information of a user, then constructs a knowledge triple, associates the knowledge triple to the user topic tree, constructs a user interest model, and finally calculates the semantic similarity of question-answer pairs in the user interest model and adds the semantic similarity to the user interest model, so that the user interest model containing the semantic similarity can be constructed, and the problem or answer with higher voice similarity can be conveniently recommended to the user in a subsequent heuristic session.
With further reference to FIG. 5, a flow diagram of yet another embodiment of a method for using a knowledge model of a conversational system is provided. The flow chart is mainly a flow of steps further defined by the steps before step 401 "hierarchical clustering is performed on the user basic information and the historical conversation data of the user to generate the user topic tree" in the flow chart 4. The method is mainly suitable for the user with particularly little historical conversation data, for example, the historical conversation data of the user can be less than a preset threshold value. The preset threshold may be customized, and the application is not limited herein.
As shown in fig. 5, in step 501, users with the same basic information are searched according to the basic information of the users;
in step 502, constructing a question-answer pair set and a knowledge triple set of a user based on a question-answer pair set and a knowledge triple set of the user with the same basic information;
in step 503, calculating semantic similarity between the user active question and questions in the constructed question-answer pair set, and selecting question-answer pairs corresponding to the first N questions with the highest semantic similarity;
in step 504, calculating semantic similarity between an answer triple corresponding to the system recommended answer received by the user and the question in the constructed knowledge triple, and selecting the first M knowledge triples with the highest semantic similarity;
in step 505, the question-answer pairs corresponding to the first N questions and the first M knowledge triples are used as the initial historical session data of the user.
In this embodiment, for step 501, if the historical session data of the user is particularly small, first, the constructing device of the session system knowledge model searches for the user with the same basic information according to the basic information of the user; then, for step 502, constructing a question-answer pair set and a knowledge triple set of the user based on the question-answer pair set and the knowledge triple set of the user with the same basic information; then, for step 503, calculating semantic similarity between the user active question and questions in the constructed question-answer pair set, and selecting question-answer pairs corresponding to the first N questions with highest semantic similarity; then, for step 504, calculating semantic similarity between an answer triple corresponding to the system recommended answer received by the user and the question in the constructed knowledge triple, and selecting the first M knowledge triples with the highest semantic similarity; finally, for step 505, the question-answer pairs corresponding to the first N questions and the first M knowledge triples are used as the initial historical session data of the user.
The scheme provided by the embodiment of the application comprehensively considers the knowledge points concerned by the user and the association between the knowledge points, realizes the modeling of finer granularity level, finds the knowledge points matched with the interest points of the user by using a graph matching algorithm according to the association of the knowledge points, the similarity between the users and the unified modeling of the interest points and the knowledge points of the user, combines the time sensitivity, the location sensitivity and the event sensitivity, so that the user can more accurately depict the interest of the user and can quickly and accurately calculate the interest points of the user, construct an initial interest model of the user and recommend more accurate conversation content for the user.
In some optional embodiments, the semantic graph matching includes node semantic similarity matching and path semantic similarity matching.
The following description is provided to enable those skilled in the art to better understand the present disclosure by describing some of the problems encountered by the inventors in implementing the present disclosure and by describing one particular embodiment of the finally identified solution.
One simple flow of heuristic sessions is as follows: first actively guide the dialogue interaction according to the question of the user, the user asks a question, and the system lists some relevant questions according to the question or asks the user that he does not want to know. The user's question is connected to the knowledge point in many forms, of course, after the conversation, we call the knowledge point, and in a knowledge point way, connecting a knowledge point may be for a specific question, which may have various different questions, we all think it is a knowledge point.
The invention provides a personalized heuristic dialogue technology based on knowledge relevance and user interest. Firstly, the patent provides a combined modeling method based on topics and knowledge points, and based on historical conversation record data of a user, the user interest is modeled in a layering mode, namely a topic layer and a knowledge point layer are included, and the topic layer supports a hierarchical structure. Secondly, the knowledge points are modeled and organized, and the modeling method comprises topic abstraction of the knowledge points, topic tree construction and association of the knowledge points; and thirdly, carrying out predictive modeling on the interest points of the new user to construct an initial interest model. And fourthly, matching the user interest graph with the knowledge graph, and finding out the knowledge points which are potentially interested by the user by combining time sensitivity, place sensitivity and event sensitivity to recommend more accurate conversation content to the user.
According to the invention, through unified modeling of the topic association of the user interest points and knowledge and the association of the knowledge points, an individualized heuristic conversation method is realized, the conversation efficiency is improved, and the communication target of the user is efficiently achieved.
The invention has the technical innovation points that:
1. and jointly modeling based on the theme, the knowledge points and the user interest points.
The traditional user modeling method mostly uses subject modeling, the patent considers subject information, and also comprehensively considers knowledge points concerned by users and the association between the knowledge points, so that the modeling of finer granularity level is realized, and the user interest can be more accurately depicted.
2. Prediction of points of interest for new users.
The traditional user modeling method utilizes the history of the user, often has the cold start problem, and is not suitable for the scene with a small amount of user history. According to the method and the device, the interest points of the user can be rapidly and accurately calculated according to the association of the knowledge points and the similarity between the users, and an initial interest model of the user is constructed.
3. And recommending the user interest points and the knowledge points based on graph matching.
The user interest points and the knowledge points are modeled in a unified mode, the knowledge points matched with the user interest points can be found by using a graph matching algorithm, and more accurate conversation contents are recommended for the user by combining time sensitivity, place sensitivity and event sensitivity.
Fig. 6 shows a flow chart of a system provided by an embodiment of the present application.
As shown in FIG. 6, the system mainly comprises three parts, wherein the left part models the interest of users lacking in historical conversation data, and such users only have basic information or registration information of the users and have a small number of conversation records. For such users, the system predicts a user interest model based on a small number of session records and a session system knowledge model and provides an initialized user interest model.
The middle part is a user with rich historical conversation data, the user has rich conversation history, and basic information of the user, active questions and answers of the user, knowledge and answers recommended by the system are used, and the user receives three types of data to model the interest of the user.
The right part is a modeling process of a knowledge system in the session system, and comprises topic tree construction, knowledge graph construction of entities and relations, and association of knowledge points and topic trees; the arrangement of knowledge of question-answer pairs includes the corresponding relation between question and answer, which may be one-to-one or one-to-many, the semantic similarity between questions, the association between the entity involved in question-answer pairs and the entity in knowledge map, etc
And matching the user interest model and the system knowledge model based on the semantic graph, calculating the best matched knowledge point, and recommending the related problems to the user.
FIG. 7 illustrates a user point of interest model structure based on topics and knowledge points.
The data sources for modeling the user interest comprise basic information of the user, active questions and answers of the user, knowledge and answers recommended by the system and three types of data accepted by the user
The user interest model is represented in the form of a semantic graph, fig. 7 shows a model structure, which includes 5 elements of topics, knowledge points, question-answer pairs, entities and attributes, and a topic layer supports a hierarchical structure, that is, one topic may have a plurality of sub-topics, for example, topic a includes two sub-topics, topic B and topic C; below the topic are knowledge points, e.g., topic D includes knowledge point 1 and knowledge point 2; knowledge points may belong to multiple topics; knowledge points include question-answer pairs and knowledge in the knowledge graph, and are represented in the form of triples, including entities, relations and attributes, and attribute values, for example, knowledge point 1 includes question-answer pair 1 and entities E1, E2 and relations between them.
The first step is as follows: and performing hierarchical clustering according to active questions and answers of the user and the knowledge and answers recommended by the system to obtain the topic tree with a hierarchical structure. Meanwhile, questions actively asked by the user and knowledge recommended by the system are constructed into question-answer pairs or knowledge triples and are associated to the theme tree.
The second step is that: and mapping the entity mentions in the question-answer pair to a system knowledge model by utilizing an entity link technology, and adding the mapped knowledge triples to the knowledge points under the corresponding topic tree. Where entity reference means the representation of the entity in the text. For example, the entity, Beijing university, may be in the form of Beijing university in text.
The third step: and calculating the semantic similarity between the question and answer pairs, and adding the semantic similarity into the user interest model.
User interest prediction and interest model initialization based on semantic similarity propagation
For users with less user historical conversation data, a user interest propagation model based on semantic similarity is provided, the user interest is predicted by utilizing the model, and an interest model is initialized.
For purposes of illustration, assume that user A has only one active question Q1 resolved by a question-answer pair, and that a question Q2 that the system recommends and is accepted by A is resolved by a knowledge triple.
Firstly, similar users are found according to the user registration/basic information, such as a user set C with similar age, gender and geographic position, the question-answer pairs of the users in the set C form a set CQA, and the knowledge triples form a set CKB.
And secondly, calculating semantic similarity of the questions Q1 and CQA, and selecting the first N questions as candidates. The specific algorithm can be simple edit distance and set distance, and can also represent the problem into a vector and calculate the cosine distance of the vector.
And thirdly, calculating the semantic similarity of the question in the answer triplet < e1, r1, e2> corresponding to Q2 and CKB. Assuming the triples in CKB are < e1 x, r2, e2 x >, the calculation method may be the minimum distance from e1 and e2 to e1 x and e2 in the knowledge graph.
And fourthly, taking the first N question-answer pairs with the highest similarity from the result of the second step, taking the first M knowledge triples with the shortest distance from the result of the third step, taking N and M as the initial interest points of the user, and constructing the interest point model of the user by using the method for constructing the model in the embodiment.
Matching the system knowledge semantic graph and the user interest semantic graph: the semantic graph matching considers the node semantic similarity and the path semantic similarity at the same time.
FIG. 8 shows a semantic representation of a user question (top section), a dialog system knowledge system semantic representation (bottom section).
The upper graph shows semantic representations of user questions and semantic representations of the knowledge system of the dialog system. Suppose a user accesses the system, asks a question Q, and hits question 1, i.e., the questions in Q and question 1 are semantically equivalent.
First, a path T1 from question answer 1 to topic is constructed, as shown in the upper part of fig. 8, where the path includes knowledge point a of question answer 1, subtopic 2, primary topic 1, and also includes all question answers QA1 and all knowledge triples KB1 below the same knowledge point as question answer 1.
Second, a topic path T2 is selected from the system knowledge hierarchy, as shown in the lower part of FIG. 8, such as the primary topic 1 and subtopic 3, knowledge point B, question-answer set QA2, knowledge point triple KB2
And thirdly, calculating the similarity of the theme paths obtained in the first step and the second step, taking the weight of the theme into consideration, averaging the similarity of the KB1 and the KB2, and averaging the similarity of the QA1 and the QA 2. The similarity calculation method can be an edit distance or a cosine distance of a word vector after word embedding.
Sim (T1, T2) ═ w1 Sim (topic 1, topic 3) + w2 Sim (sub topic 2, sub topic 4) + average (Sim (QA1, QA2)) + average (Sim (KB1, KB2)), where Sim means similarity, i.e. similarity.
Fourth, the weight of the corresponding QA or knowledge triples is increased according to time, location, event sensitivity, e.g., if the current time is mid-autumn, the weight of the QA or knowledge triples about mid-autumn will be increased. The location and event handling methods are similar.
Updating a semantic graph of a knowledge system of a session system and a semantic graph of user interest: after the question is recommended to a certain user. When the user selects a question recommended by the system, the question is added to the user interest model. If the user does not select the recommended question, the recommendation index of the question with higher semantic similarity in the user interest model and the question is reduced by a certain value.
The method and the device recommend the knowledge points which are more interesting to the user by integrating the relevance of knowledge and the interest points of the user, so that the high-efficiency personalized heuristic conversation is completed. Personalized heuristic sessions have several benefits: firstly, the communication efficiency is improved, and convergence to topics which are interesting to the user is accelerated. Second, user satisfaction is improved.
Referring to fig. 9, a block diagram of a device for constructing a knowledge model of a conversational system according to an embodiment of the present application is shown.
As shown in fig. 9, an apparatus 900 for constructing a knowledge model of a session system includes: a knowledge point abstraction module 910, a topic tree construction module 920, an association determination module 930, and a knowledge model construction module 940.
The knowledge point abstraction module 910 is configured to abstract at least one topic associated with a knowledge point from the knowledge point in the session system; a topic tree construction module 920 configured to construct a system topic tree based on the at least one topic; an association relation determining module 930 configured to determine an association relation between knowledge points based on the system topic tree; and a knowledge model construction module 940 configured to construct a conversational system knowledge model based at least on the association between the system topic tree and the knowledge points.
Referring to fig. 10, a block diagram of a device for constructing a knowledge model of a conversational system according to an embodiment of the present application is shown.
As shown in fig. 10, an apparatus 1000 for using a knowledge model of a session system includes: the system comprises a user interest model building module 1010, a semantic graph matching module 1020, a matching knowledge point generating module 1030 and a recommending module 1040.
The user interest model building module 1010 is configured to build a user interest model based on basic information of a user and historical conversation data of the user, wherein the user interest model comprises at least one knowledge point; a semantic map matching module 1020 configured to perform semantic map-based matching on knowledge points in the user interest model and knowledge points in the knowledge model of the session system constructed according to the method of claim 1 or 2; a matching knowledge point generating module 1030 configured to generate at least one matching knowledge point according to a matching result of the semantic graph, wherein the knowledge point is associated with at least one question; and a recommending module 1040 configured to recommend the question associated with the at least one matched knowledge point to the user.
It should be understood that the modules recited in fig. 9 and 10 correspond to various steps in the methods described with reference to fig. 1, 2, 3, 4, and 5. Thus, the operations and features described above for the method and the corresponding technical effects are also applicable to the modules in fig. 5, and are not described again here.
It should be noted that the modules in the embodiments of the present disclosure are not intended to limit the solution of the present disclosure, for example, the attribute analysis module may be described as a module that analyzes the basic attribute of the user based on the obtained voiceprint information of the user. In addition, the related function module may also be implemented by a hardware processor, for example, the attribute analysis module may also be implemented by a processor, which is not described herein again.
In other embodiments, an embodiment of the present invention further provides a non-volatile computer storage medium, where the computer storage medium stores computer-executable instructions, and the computer-executable instructions may execute the method for constructing the session system knowledge model in any of the above method embodiments;
as one embodiment, a non-volatile computer storage medium of the present invention stores computer-executable instructions configured to:
abstracting, by a knowledge point in a conversational system, at least one topic associated with the knowledge point;
building a system topic tree based on the at least one topic;
determining an incidence relation between knowledge points based on the system theme tree;
and constructing a conversation system knowledge model at least based on the incidence relation between the system topic tree and the knowledge points.
The non-volatile computer-readable storage medium may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created from use of a construction apparatus of the session system knowledge model, and the like. Further, the non-volatile computer-readable storage medium may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid state storage device. In some embodiments, the non-transitory computer-readable storage medium optionally includes a memory remotely located from the processor, and the remote memory may be connected to the construction apparatus of the session system knowledge model via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
Embodiments of the present invention also provide a computer program product, which includes a computer program stored on a non-volatile computer-readable storage medium, where the computer program includes program instructions, and when the program instructions are executed by a computer, the computer executes any one of the above methods for constructing a knowledge model of a session system.
Fig. 11 is a schematic structural diagram of an electronic device according to an embodiment of the present invention, and as shown in fig. 11, the electronic device includes: one or more processors 1110 and a memory 1120, with one processor 1110 being an example in fig. 11. The device of the construction method of the session system knowledge model can also comprise: an input device 1130 and an output device 1140. The processor 1110, the memory 1120, the input device 1130, and the output device 1140 may be connected by a bus or other means, and the bus connection is exemplified in fig. 11. The memory 1120 is a non-volatile computer-readable storage medium as described above. The processor 1110 executes various functional applications of the server and data processing by running nonvolatile software programs, instructions and modules stored in the memory 1120, so as to implement the method for constructing the knowledge model of the session system according to the above method embodiment. The input device 1130 may receive input numeric or character information and generate key signal inputs related to user settings and function control of the above-described devices. The output device 1140 may include a display device such as a display screen.
The product can execute the method provided by the embodiment of the invention, and has corresponding functional modules and beneficial effects of the execution method. For technical details that are not described in detail in this embodiment, reference may be made to the method provided by the embodiment of the present invention.
As an embodiment, the electronic device is applied to a construction apparatus of a knowledge model of a session system, and is used for a client, and the construction apparatus includes: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor to cause the at least one processor to:
analyzing the basic attribute of the user based on the acquired voiceprint information of the user;
recommending a first set of stories to the user for selection based on the user's underlying attributes;
judging whether the user selects any story in the first story set and recording the selection condition of the user, wherein the any story has at least one story attribute and each story attribute corresponds to a weight value;
updating the weight value of each story attribute of the user based on the selection condition of the user;
recommending a second story set to the user for selection based on the base attributes of the user and the updated weight values of the story attributes of the user.
The electronic device of the embodiments of the present application exists in various forms, including but not limited to:
(1) a mobile communication device: such devices are characterized by mobile communications capabilities and are primarily targeted at providing voice, data communications. Such terminals include smart phones (e.g., iphones), multimedia phones, functional phones, and low-end phones, among others.
(2) Ultra mobile personal computer device: the equipment belongs to the category of personal computers, has calculation and processing functions and generally has the characteristic of mobile internet access. Such terminals include: PDA, MID, and UMPC devices, etc., such as ipads.
(3) A portable entertainment device: such devices can display and play multimedia content. Such devices include audio and video players (e.g., ipods), handheld game consoles, electronic books, as well as smart toys and portable car navigation devices.
(4) The server is similar to a general computer architecture, but has higher requirements on processing capability, stability, reliability, safety, expandability, manageability and the like because of the need of providing highly reliable services.
(5) And other electronic devices with data interaction functions.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods of the various embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A construction method of a session system knowledge model comprises the following steps:
abstracting, by a knowledge point in a conversational system, at least one topic associated with the knowledge point;
building a system topic tree based on the at least one topic;
determining an incidence relation between knowledge points based on the system theme tree;
and constructing a conversation system knowledge model at least based on the incidence relation between the system topic tree and the knowledge points.
2. The method of claim 1, wherein the knowledge points comprise question-answer pairs and a knowledge graph, the question-answer pairs comprising a first entity, the knowledge graph comprising a second entity and associations between the second entities, the determining associations between knowledge points based on the topic tree comprising:
when the first entity and the second entity in the theme tree are the same, performing association fusion on the first entity and the second entity;
and determining the association relation among the knowledge points based on the association fused topic tree.
3. A method of using a conversational system knowledge model, comprising:
constructing a user interest model based on basic information of a user and historical conversation data of the user, wherein the user interest model comprises at least one knowledge point;
performing semantic graph-based matching on knowledge points in the user interest model and knowledge points in the session system knowledge model constructed according to the method of claim 1 or 2;
generating at least one matched knowledge point according to a matching result of the semantic graph, wherein the knowledge point is associated with at least one question;
recommending the question associated with the at least one matched knowledge point to the user.
4. The method of claim 3, wherein the historical session data of the user comprises user active questions and answers and user accepted system recommendation knowledge and answers, and the constructing the user interest model based on the basic information of the user and the historical session data of the user comprises:
carrying out hierarchical clustering on the basic information of the user and the historical conversation data of the user to generate a user theme tree;
constructing the active questions and answers of the users and the system recommendation knowledge and answers received by the users into knowledge triples containing questions, answers and incidence relations among the questions and the answers, wherein the questions and the answers form question-answer pairs;
associating the knowledge triples to the user topic tree;
mapping the entities in the question-answer pairs in the triples to the knowledge triples of the session system knowledge model by using an entity linking technology, and adding the mapped knowledge triples to the knowledge points under the user topic tree to form a user interest model;
and calculating the semantic similarity between the question-answer pairs, and adding the semantic similarity into the user interest model.
5. The method according to claim 4, wherein if the historical conversation data of the user is less than the preset threshold, before the hierarchical clustering of the basic information of the user and the historical conversation data of the user to generate the user topic tree, the method further comprises:
searching users with the same basic information according to the basic information of the users;
constructing a question-answer pair set and a knowledge triple set of the user based on the question-answer pair set and the knowledge triple set of the user with the same basic information;
calculating the semantic similarity between the active questions of the user and the questions in the constructed question-answer pair set, and selecting question-answer pairs corresponding to the first N questions with the highest semantic similarity;
calculating semantic similarity between answer triples corresponding to the system recommended answers received by the user and the questions in the constructed knowledge triples, and selecting the first M knowledge triples with the highest semantic similarity;
and taking the question-answer pairs corresponding to the first N questions and the first M knowledge triples as initial historical conversation data of the user.
6. The method of any of claims 3-5, wherein the semantic graph matching includes node semantic similarity matching and path semantic similarity matching.
7. An apparatus for constructing a knowledge model of a session system, comprising:
the knowledge point abstraction module is configured to abstract at least one theme related to the knowledge points from the knowledge points in the session system;
a subject tree construction module configured to construct a system subject tree based on the at least one subject;
the incidence relation determining module is configured to determine incidence relations among the knowledge points based on the system theme tree;
and the knowledge model building module is configured to build a session system knowledge model at least based on the incidence relation between the system topic tree and each knowledge point.
8. An apparatus for using a knowledge model of a conversational system, comprising:
the user interest model building module is configured to build a user interest model based on basic information of a user and historical conversation data of the user, wherein the user interest model comprises at least one knowledge point;
a semantic map matching module configured to perform semantic map-based matching on knowledge points in the user interest model and knowledge points in the session system knowledge model constructed according to the method of claim 1 or 2;
the matching knowledge point generating module is configured to generate at least one matching knowledge point according to a matching result of the semantic graph, wherein the knowledge point is associated with at least one question;
a recommendation module configured to recommend a question associated with the at least one matched knowledge point to the user.
9. An electronic device, comprising: at least one processor, and a memory communicatively coupled to the at least one processor, wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the steps of the method of any one of claims 1 to 6.
10. A storage medium having stored thereon a computer program, characterized in that the program, when being executed by a processor, is adapted to carry out the steps of the method of any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911407443.1A CN111159382B (en) | 2019-12-27 | 2019-12-27 | Method and device for constructing and using session system knowledge model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911407443.1A CN111159382B (en) | 2019-12-27 | 2019-12-27 | Method and device for constructing and using session system knowledge model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111159382A true CN111159382A (en) | 2020-05-15 |
| CN111159382B CN111159382B (en) | 2022-07-12 |
Family
ID=70559851
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911407443.1A Active CN111159382B (en) | 2019-12-27 | 2019-12-27 | Method and device for constructing and using session system knowledge model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111159382B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111814091A (en) * | 2020-07-17 | 2020-10-23 | 北京达佳互联信息技术有限公司 | Message distribution method and related device |
| CN113704440A (en) * | 2021-09-06 | 2021-11-26 | 中国计量大学 | Conversation recommendation method based on path representation in article graph network |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109086434A (en) * | 2018-08-13 | 2018-12-25 | 华中师范大学 | A kind of knowledge polymerizing method and system based on thematic map |
| CN110008326A (en) * | 2019-04-01 | 2019-07-12 | 苏州思必驰信息科技有限公司 | Knowledge abstraction generating method and system in conversational system |
| CN110377715A (en) * | 2019-07-23 | 2019-10-25 | 天津汇智星源信息技术有限公司 | Reasoning type accurate intelligent answering method based on legal knowledge map |
-
2019
- 2019-12-27 CN CN201911407443.1A patent/CN111159382B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109086434A (en) * | 2018-08-13 | 2018-12-25 | 华中师范大学 | A kind of knowledge polymerizing method and system based on thematic map |
| CN110008326A (en) * | 2019-04-01 | 2019-07-12 | 苏州思必驰信息科技有限公司 | Knowledge abstraction generating method and system in conversational system |
| CN110377715A (en) * | 2019-07-23 | 2019-10-25 | 天津汇智星源信息技术有限公司 | Reasoning type accurate intelligent answering method based on legal knowledge map |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111814091A (en) * | 2020-07-17 | 2020-10-23 | 北京达佳互联信息技术有限公司 | Message distribution method and related device |
| CN111814091B (en) * | 2020-07-17 | 2023-11-28 | 北京达佳互联信息技术有限公司 | Message distribution method and related device |
| CN113704440A (en) * | 2021-09-06 | 2021-11-26 | 中国计量大学 | Conversation recommendation method based on path representation in article graph network |
| CN113704440B (en) * | 2021-09-06 | 2022-02-18 | 中国计量大学 | Conversation recommendation method based on path representation in article graph network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111159382B (en) | 2022-07-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11302337B2 (en) | Voiceprint recognition method and apparatus | |
| JP2020173462A (en) | Selection of synthetic speech for agent by computer | |
| US20160379106A1 (en) | Human-computer intelligence chatting method and device based on artificial intelligence | |
| CN112074857A (en) | Combining machine learning and social data to generate personalized recommendations | |
| CN108763494B (en) | Knowledge sharing method between conversation systems, conversation method and device | |
| US11188720B2 (en) | Computing system including virtual agent bot providing semantic topic model-based response | |
| CN112487173B (en) | Man-machine conversation method, device and storage medium | |
| CN110399470B (en) | Session message handling | |
| CN110597962A (en) | Search result display method, device, medium and electronic equipment | |
| CN110234018B (en) | Multimedia content description generation method, training method, device, equipment and medium | |
| CN110008326B (en) | Knowledge abstract generation method and system in session system | |
| CN109948151A (en) | The method for constructing voice assistant | |
| CN111798279A (en) | Dialog-based user portrait generation method and apparatus | |
| CN111159382B (en) | Method and device for constructing and using session system knowledge model | |
| US20230352012A1 (en) | Speech skill jumping method for man machine dialogue, electronic device and storage medium | |
| US20190163816A1 (en) | Electronic message text classification framework selection | |
| CN113641807A (en) | Training method, device, equipment and storage medium of dialogue recommendation model | |
| JP2023036574A (en) | Conversational recommendation method, method and device of training model, electronic apparatus, storage medium, and computer program | |
| CN108306813B (en) | Session message processing method, server and client | |
| KR20200096935A (en) | Method and system for providing multiple profiles | |
| CN111737449B (en) | Method and device for determining similar problems, storage medium and electronic device | |
| CN111427444B (en) | Control method and device of intelligent device | |
| US10872121B2 (en) | Systems and methods for matching users | |
| CN113360590B (en) | Method and device for updating interest point information, electronic equipment and storage medium | |
| CN112257781A (en) | Model training method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: 215123 building 14, Tengfei Innovation Park, 388 Xinping street, Suzhou Industrial Park, Suzhou City, Jiangsu Province Applicant after: Sipic Technology Co.,Ltd. Address before: 215123 building 14, Tengfei Innovation Park, 388 Xinping street, Suzhou Industrial Park, Suzhou City, Jiangsu Province Applicant before: AI SPEECH Co.,Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |