CN111782781A - A semantic analysis method, device, computer equipment and storage medium - Google Patents

Abstract

The invention relates to artificial intelligence, and provides a semantic analysis method, a semantic analysis device, computer equipment and a storage medium. The method includes: receiving a query request sent by a user terminal for an intelligent question answering system, where the query request at least carries query text information; reading a local entity library, and translating the query text information based on the local entity library and preset logic rules to obtain logic Text information; convert logical text information into target semantic tree based on semantic tree conversion rules; use the target semantic tree as the input content of the intelligent question answering system to perform intelligent question answering operations. In addition, the present invention also relates to the blockchain technology, and the user's private information can be stored in the blockchain. The present application can identify the user's real emotion in combination with the contextual logical relationship of the text input by the user, and can provide the user with answer content that is more suitable for the user's needs.

Description

A semantic analysis method, device, computer equipment and storage medium

technical field

The present application relates to the technical field of artificial intelligence, and in particular, to a semantic analysis method, apparatus, computer equipment and storage medium.

Background technique

The intelligent question answering system is an important position for artificial intelligence to be implemented in the industry. With the advent of the era of big data and AI, more and more traditional BI systems are facing the pressure and challenge of integrating data dimension and business function dimension.

There is a customer service robot question answering method, which communicates with the user through a preset customer service robot, and extracts the corresponding reply content in the database based on the keywords asked during the user communication process, so as to realize the process of intelligent question answering.

However, the traditional question and answer method of customer service robots is generally unintelligent. The customer service robot can only return fixed texts to answer questions by configuring keywords. It does not consider the combination of contexts and does not recognize the real emotions of users, but only mechanically. Answer, the user experience is poor.

SUMMARY OF THE INVENTION

The purpose of the embodiments of the present application is to solve the problem that traditional customer service robot question answering methods generally only return fixed text to answer questions by configuring keywords, do not consider the combination of contexts, and do not recognize the real emotions of users, but only mechanically The answer to the problem of poor user experience.

In order to solve the above technical problems, the embodiment of the present application provides a semantic analysis method, which adopts the following technical solutions:

receiving a query request sent by the user terminal for the intelligent question answering system, where the query request at least carries query text information;

reading a local entity library, and performing a translation operation on the query text information based on the local entity library and preset logic rules to obtain logical text information;

Converting the logical text information into a target semantic tree based on a semantic tree conversion rule;

The intelligent question answering operation is performed using the target semantic tree as the input content of the intelligent question answering system.

In order to solve the above technical problems, the embodiment of the present application also provides a semantic analysis device, which adopts the following technical solutions:

a request receiving module, configured to receive a query request sent by the user terminal for the intelligent question answering system, where the query request at least carries query text information;

a text acquisition module, configured to read a local entity library, perform a translation operation on the query text information based on the local entity library and preset logic rules, and acquire logical text information;

a semantic conversion module for converting the logical text information into a target semantic tree based on a semantic tree conversion rule;

The semantic input module is configured to use the target semantic tree as the input content of the intelligent question answering system to perform intelligent question answering operations.

In order to solve the above-mentioned technical problems, the embodiment of the present application also provides a computer device, which adopts the following technical solutions:

A memory and a processor are included, and a computer program is stored on the computer-readable storage medium, and when the computer program is executed by the processor, the steps of the semantic analysis method as described above are implemented.

In order to solve the above technical problems, the embodiments of the present application also provide a computer-readable storage medium, which adopts the following technical solutions:

A computer program is stored on the computer-readable storage medium, and when the computer program is executed by the processor, implements the steps of the semantic analysis method as described above.

Compared with the prior art, the semantic analysis method and the semantic analysis device provided by the embodiments of the present application have at least the following beneficial effects:

By translating the query text input by the user into logical text, it effectively helps us convert natural language questions into machine-recognizable expressions, clarifies the vague expressions in questions, and presents questions with more standardized and clearer rules. It lays a solid foundation for the subsequent realization of functions based on natural language; and converts the logical text into a semantic tree, which can organize semantics in a rigorous logical order, and the conversion of logical rules and semantic trees can make understandable language information into The computer-identifiable data structure provides the basis for subsequent query operations. The verification of logic rules provides the basis for semantic inspection and repair. The semantic tree is used as the input content for question-and-answer operations, which can be combined with the contextual logic relationship of the text input by the user. Recognizing the real emotions of users can provide users with answer content that is more in line with user needs.

Description of drawings

In order to illustrate the solutions in the present application more clearly, the following will briefly introduce the accompanying drawings used in the description of the embodiments of the present application. For those of ordinary skill, other drawings can also be obtained from these drawings without any creative effort.

Fig. 1 is the realization flow chart of the semantic analysis method provided by the first embodiment of the present invention;

2 is a schematic diagram of a target semantic tree provided by Embodiment 1 of the present invention;

Fig. 3 is the realization flow chart of step S102 in Fig. 1;

Fig. 4 is the realization flow chart of step S103 in Fig. 1;

Fig. 5 is the realization flow chart of the logic verification method provided by the first embodiment of the present invention;

6 is a schematic structural diagram of a semantic analysis device provided in Embodiment 2 of the present invention;

Fig. 7 is the structural representation of the text acquisition module in Fig. 6;

Fig. 8 is the structural representation of the semantic conversion module in Fig. 6;

FIG. 9 is a schematic structural diagram of an embodiment of a computer device according to the present application.

Detailed ways

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the technical field of this application; the terms used herein in the specification of the application are for the purpose of describing specific embodiments only It is not intended to limit the application; the terms "comprising" and "having" and any variations thereof in the description and claims of this application and the above description of the drawings are intended to cover non-exclusive inclusion. The terms "first", "second" and the like in the description and claims of the present application or the above drawings are used to distinguish different objects, rather than to describe a specific order.

Reference herein to an "embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the present application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor a separate or alternative embodiment that is mutually exclusive of other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

In order to make those skilled in the art better understand the solutions of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings.

Example 1

Referring to FIG. 1 , an implementation flowchart of the semantic analysis method provided by Embodiment 1 of the present invention is shown. For convenience of description, only parts related to the present invention are shown.

In step S101, a query request sent by the user terminal for the intelligent question answering system is received, where the query request at least carries query text information.

In the embodiment of the present invention, the user terminal may be such as a mobile phone, a smart phone, a notebook computer, a digital broadcast receiver, a PDA (personal digital assistant), a PAD (tablet computer), a PMP (portable multimedia player), a navigation device, etc. such as mobile terminals and fixed terminals such as digital TVs, desktop computers, etc., it should be understood that the examples of user terminals here are only for convenience of understanding, and are not intended to limit the present invention.

In the embodiment of the present invention, the intelligent question answering system is mainly used to accurately locate the questioning knowledge required by website users, and provide personalized information services for website users by interacting with website users.

In this embodiment of the present invention, the query request refers to the data stream information that the user sends to the system through the user terminal and contains the content that the user needs to ask. The query request may be text data input by text, or audio input by voice. data. When the query request is audio data, it is necessary to perform speech recognition on the audio data and convert it into text data that the system can perform semantic analysis. invention.

In this embodiment of the present invention, the query text information refers to text data input by the user, or text data converted from audio data input by the user, and the query text information is the most original semantic content of the user.

In step S102, the local entity library is read, and the query text information is translated based on the local entity library and preset logic rules to obtain logical text information.

In this embodiment of the present invention, the local entity library is mainly used to pre-store a knowledge base of entity content.

In the embodiment of the present invention, the preset logic rules are mainly used to accurately express natural language information and facilitate machine identification, and the preset logic rules can be correspondingly designed according to actual needs.

In the embodiment of the present invention, the translation operation refers to, after identifying the above query text information, by means of entity connection, comparing the query text information with the entity content in the above knowledge base, obtaining the text with the same entity content, The semantically represented logical rule symbol set translates the text, so as to obtain the above-mentioned logical text information.

In the embodiment of the present invention, the largest ranking entity that appears in the question is taken as the subject entity word of the question, and the entity finally queried by the user is taken as the end of the query graph link. By traversing the entities, relations and attributes in the question, the core reasoning link from the subject entity word to the ending entity word is obtained. For each core derivation link, it is converted into a tree structure for storage, and the core derivation link that is most similar to the question is obtained by comparing the similarity between the tree structure and the known intent. Finally, the link information of the core derivation link is parsed into query parameters and query process.

In step S103, the logical text information is converted into a target semantic tree based on the semantic tree conversion rule.

In this embodiment of the present invention, the semantic tree conversion rule refers to a logical rule expressing minimal semantics with "<>". We first divide the logic rules into a series of minimum semantic units, and then convert the minimum semantic units into nodes of the semantic tree (semantic combination and assembly), and build the semantic tree according to the reference relationship between the semantic units. This reference relationship represents the The sequence of queries and the way of association between knowledge graph entities.

In practical applications, leaf nodes correspond to entities of triples, non-leaf nodes correspond to relationships between entities of triples or attributes of entities, and answer nodes are located at the root node. Taking the exposure information of the counterparty of XX company as an example, XX company and the counterparty are entities, the transaction relationship is an inter-entity relationship, the exposure information is an attribute, and the final answer node asks for the attribute value, as shown in Figure 2.

In step S104, the intelligent question answering operation is performed using the target semantic tree as the input content of the intelligent question answering system.

By translating the query text input by the user into logical text, it effectively helps us convert natural language questions into machine-recognizable expressions, clarifies the vague expressions in questions, and presents questions with more standardized and clearer rules. It lays a solid foundation for the subsequent realization of functions based on natural language; and converts the logical text into a semantic tree, which can organize semantics in a rigorous logical order, and the conversion of logical rules and semantic trees can make understandable language information into The computer-identifiable data structure provides the basis for subsequent query operations. The verification of logic rules provides the basis for semantic inspection and repair. The semantic tree is used as the input content for question-and-answer operations, which can be combined with the contextual logic relationship of the text input by the user. Recognizing the real emotions of users can provide users with answer content that is more in line with user needs.

As some optional implementation manners of the first embodiment, the preset logic rules specifically include Unary unary rules, Binary binary rules, and Aggregation aggregation rules.

In the embodiment of the present invention, in order to accurately express natural language information and display it with machine-recognizable logical rules, three semantic unit templates are designed, namely Unary unary rule, Binary binary rule and Aggregation aggregation rule. The specific rules are as follows :

(1) Unary unary rule

For the entity e in the question, it needs to be clear whether the entity e represents a class of entities or an instance Instance. If only the category of the entity appears in the question, it is expressed as <Unary(class='E')>; if the instance value of the entity appears in the question, the corresponding entity category needs to be found, expressed as <Unary( class='E', value='Instance')>. For example, define 'organization' as an entity type. When 'company' appears in the question, it is expressed as <Unary(class='organization')>, and if 'XXX company' appears in the question, it is expressed as <Unary(class='organization', value='XXX company')>.

(2) Binary binary rule

Binary is mainly used to describe the relationship of triples, and triples can be understood as (head entity, relationship, tail entity), (head entity, attribute, attribute value) or (relation, relationship attribute, relationship attribute value). The Binary rule is to know two of the elements, to find the third element, use '? ' to mark the element to query.

Relationship class problem, S and E represent entities, rel represents relationships

Binary(S,rel?,E) returns the relationship between S and E;

Binary(S,rel,E?) returns what the rel of S is;

Binary(S?,rel,E) answers which entities have a rel relationship with E;

b) Entity attribute class problem, S means entity, PRO means attribute, value means attribute value

Binary(S, PRO, value?) returns the property PRO of S;

Binary(S?,PRO,value) seeks the entity S whose attribute PRO is value;

Binary(S, PRO?, value) asks what property of S is value;

c) Relational attribute class problem, R represents the relationship, RPRO represents the attribute of the relationship, and value represents the value of the attribute

Binary(R,RPRO,value?) What is the attribute RPRO of the relationship;

(3) Aggregation aggregation rules

A series of aggregation operators are quoted here to express the operational logic and qualification logic in the question, mainly including the following seven types:

Rank operation

Parameter settings: by—sorted property; obj—operated object; range—sorted return range

Purpose: Sorting, used to answer 'Top N', 'Nth' and other questions

Count operation

Parameter setting: obj—counted object; con—filter condition when counting

Purpose: Statistics, used to answer 'how many' similar questions

Sum operation

Parameter setting: obj—the object to be summed; con—the filter condition when adding up

Purpose: Summarize numeric attributes

Opr operation

Parameter setting: '='--calculation method (+-*/and or);

S—value 1; E—value 2; con—filter condition during calculation

Purpose: Binary computing, used to answer various computing problems

Filter operation

Parameter settings: obj--filtered object; con-filter condition

Purpose: to filter the properties of the relationship

Time operation

Parameter setting: con—time limit range

Purpose: used to express the time span of a question

Trend&Distribution operation

Parameter setting: none

Purpose: Distribution and trend calculation

The entire logic rule consists of multiple semantic units, and the syntax specification of a single semantic unit is as follows:

<unit_code:semantic_unit>

The construction of logical rules should follow the word order of the original question, and the arrangement order of the semantic units should be consistent with the semantic location in the original question.

unit_code represents the encoding of the semantic unit, and the logic rule should have only one main unit. The main unit indicates the content of the answer to be returned in the final question sentence, which is represented by A. The main unit must be executed last. The principle of semantic unit coding is to name extensions from A to the front and then back, for example: V5 V4 V3 V2 V1 AV6 V7 V8.

Semantic_unit is the triple representation mentioned in the previous unit, and will not be repeated.

Continuing to refer to FIG. 3 , a flowchart of the implementation of step S102 in FIG. 1 is shown. For the convenience of description, only the parts related to the present invention are shown.

As some optional implementation manners of the first embodiment, the above step S102 specifically includes: step S201 , step S202 , step S203 , step S204 , step S205 and step S206 .

In step S201, entity data corresponding to the query text information is acquired in the local entity library.

In the embodiment of the present invention, the keyword can be obtained by identifying the content in the question, and compared with the entity existing in the local entity database, and finally the data consistent with the keyword and the entity is obtained as the entity data.

In step S202, a topic entity word corresponding to the entity data is determined based on a preset weight value.

In this embodiment of the present invention, one of the multiple entities found in step S201 is selected as the subject entity word. The weight value of each entity is preset through the pointing relationship between entities in the local entity. Entities with weights greater than 0 are used as candidate sets of topic entity words, and subsequent operations are attempted on each candidate topic word in order.

In step S203, the end entity data of the query text information is used as the end entity word.

In this embodiment of the present invention, the entity to be inquired at the end of the question sentence is taken as the ending entity word. When the ending entity word is a relation, try to fix it.

In step S204, the subject entity word and the ending entity word are respectively used as the first entity word of the initial logical text to obtain an intermediate logical text.

In the embodiment of the present invention, the subject entity word obtained in step S202 and the ending entity word obtained in step S203 are used as the first and last ends of the link to traverse in the graph. Get the candidate subgraph set.

In step S205, the optimal intent with the highest similarity to the intermediate logical text is obtained in the intent database.

In the embodiment of the present invention, the similarity judgment is performed between the candidate subgraph and the existing intent in the intent library. Similarity is judged by whether entities, relationships, locations, etc. are consistent. Get the most similar one intent to return.

In step S206, logical text information is generated based on the optimal intention.

In this embodiment of the present invention, a logical form is generated according to the graph structure of the intent subgraph.

In practical applications, the node (entity type) corresponds to the uniary representation rule, the class corresponds to the entity type, and the value corresponds to the specific instance value, such as Ping An Life. If there is no specific entity value, uniform indicates information that can add value. Edges (relationships, attribute entities) correspond to binary representation rules. The first position of Binary is the node information corresponding to the edge, and the second position is the representation information of the edge, such as: attribute entity risk rate. The third position is filled with corresponding node information according to whether the edge is a relationship or an attribute entity. For example: There is only one Binary representation in this problem, and the second position is the risk rate, that is, the attribute entity. At this time, if there is an attribute value of the accident rate, for example, whether the value of the accident rate is greater than 20%, the attribute value of the accident rate is greater than 20%. Then the third position is filled with >20%. If there is no corresponding attribute value, the filling refers to entity A according to the definition of the logical form.

Finally, the following logical form is obtained:

<V1:Uniary(class=professional company, value=Ping An Life)>

<A:Binary(V1, risk rate, A?)>

Continuing to refer to FIG. 4 , a flow chart of the implementation of step S103 in FIG. 1 is shown. For convenience of description, only the parts related to the present invention are shown.

As some optional implementation manners of the first embodiment, the foregoing step S103 specifically includes: step S301 , step S302 and step S303 .

In step S301, a semantic segmentation operation is performed on the logical text information to obtain a semantic unit.

In step S302, a node conversion operation is performed on the semantic unit to obtain a semantic node.

In step S303, a semantic tree is constructed based on the semantic nodes to obtain a target semantic tree.

In practical applications, the program will use <> as the node unit to split the above logic rules into five nodes. Each node creates a tree node object (each tree node object has the following attributes: name, type, value, left child node, right child node, parent node, etc.), for example, with <V3:Binary(V2?, transaction relationship, V1)> node as an example, the name of the tree node object is V3, the type is Binary, the value is (V2?, transaction relationship, V1), and the left child node, right child node, and parent node are initialized to empty. In the subsequent description, we use the name of the node object to represent the node name. For example, <V1:Uniary(class='organization', value='XX company')> is described as a V1 node.

Build a tree based on five tree node objects, mainly update the left child node, right child node, and parent node of each tree node, for example:

1) For the V3 node, update the following values: the left child node of V3 is V2, the right child node is V1, the parent node of V2 is V3, and the parent node of V1 is V3.

2) For the V1 node, update the following values: the left child node of V1 is empty, and the right child node is empty.

3) For the V4 node, update the following values: the left child node of V4 is V3, the right child node is empty, and the parent node of V3 is V4.

4) For node A, update the following values: the left child node of node A is V4, the right child node is empty, and the parent-child node of node V4 is A.

In this way, a reference relationship between tree nodes is established, and at the same time, the root node is obtained as node A, and the finally formed semantic tree is shown in Figure 2 above.

Continuing to refer to FIG. 5 , a flow chart of the implementation of the logic verification method provided by Embodiment 1 of the present invention is shown. For convenience of description, only parts related to the present invention are shown.

As some optional implementation manners of the first embodiment, after the above step S103, it further includes: step S401, step S402, step S403 and step S404.

In step S401, the node type of the target semantic tree is acquired.

In step S402, it is judged whether there is an entity triple corresponding to the node type in the local entity library.

In step S403, if there is an entity triple corresponding to the node type in the local entity library, execute the intelligent question answering operation using the target semantic tree as the input content of the intelligent question answering system step.

In step S404, if the entity triple corresponding to the node type does not exist in the local entity library, a node error signal is output.

In the embodiment of the present invention, the process of judging whether there is an entity triple is to use the subsequent tree traversal search algorithm to search the entire tree from the bottom up, and at the same time, take a predefined concept map triple from the database, and compare the nodes by type. Yes, make sure that each node conforms to the defined triple, if so, continue to search upwards, if not, return the error node. The purpose of logic verification is to not only find the lack, redundancy and dislocation of semantics (providing a basis for the revision of the semantic tree), but also to provide error information for the semantic parsing module to make corrections.

To sum up, the semantic analysis method provided by the embodiment of the present invention has at least the following beneficial effects:

By translating the query text input by the user into logical text, it effectively helps us convert natural language questions into machine-recognizable expressions, clarifies the vague expressions in questions, and presents questions with more standardized and clearer rules. It lays a solid foundation for the subsequent realization of functions based on natural language; and converts the logical text into a semantic tree, which can organize semantics in a rigorous logical order, and the conversion of logical rules and semantic trees can make understandable language information into The computer-identifiable data structure provides the basis for subsequent query operations. The verification of logic rules provides the basis for semantic inspection and repair. The semantic tree is used as the input content for question-and-answer operations, which can be combined with the contextual logic relationship of the text input by the user. Recognizing the real emotions of users can provide users with answer content that is more in line with user needs.

Those of ordinary skill in the art can understand that the realization of all or part of the processes in the methods of the above embodiments can be accomplished by instructing relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium, and the program is During execution, it may include the processes of the embodiments of the above-mentioned methods. The aforementioned storage medium may be a non-volatile storage medium such as a magnetic disk, an optical disk, a read-only memory (Read-Only Memory, ROM), or a random access memory (Random Access Memory, RAM).

It should be understood that although the various steps in the flowchart of the accompanying drawings are sequentially shown in the order indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless explicitly stated herein, the execution of these steps is not strictly limited to the order and may be performed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but may be executed at different times, and the execution sequence is also It does not have to be performed sequentially, but may be performed alternately or alternately with other steps or at least a portion of sub-steps or stages of other steps.

Embodiment 2

With further reference to FIG. 6 , as an implementation of the method shown in FIG. 1 , the present application provides an embodiment of a semantic recognition device. The device embodiment corresponds to the method embodiment shown in FIG. 1 . Specifically, the device may Used in various electronic devices.

As shown in FIG. 6 , the semantic recognition apparatus 100 provided by the second embodiment of the present invention includes: a request receiving module 110 , a text obtaining module 120 , a semantic conversion module 130 , and a semantic input module 140 . in:

The request receiving module 110 is configured to receive a query request sent by the user terminal for the intelligent question answering system, where the query request at least carries query text information;

A text acquisition module 120, configured to read a local entity library, perform a translation operation on the query text information based on the local entity library and preset logic rules, and acquire logical text information;

a semantic conversion module 130, configured to convert the logical text information into a target semantic tree based on a semantic tree conversion rule;

The semantic input module 140 is configured to use the target semantic tree as the input content of the intelligent question answering system to perform intelligent question answering operation.

In the embodiment of the present invention, the user terminal may be such as a mobile phone, a smart phone, a notebook computer, a digital broadcast receiver, a PDA (personal digital assistant), a PAD (tablet computer), a PMP (portable multimedia player), a navigation device, etc. such as mobile terminals and fixed terminals such as digital TVs, desktop computers, etc., it should be understood that the examples of user terminals here are only for convenience of understanding, and are not intended to limit the present invention.

In the embodiment of the present invention, the intelligent question answering system is mainly used to accurately locate the questioning knowledge required by website users, and provide personalized information services for website users by interacting with website users.

In this embodiment of the present invention, the query request refers to the data stream information that the user sends to the system through the user terminal and contains the content that the user needs to ask. The query request may be text data input by text, or audio input by voice. data. When the query request is audio data, it is necessary to perform speech recognition on the audio data and convert it into text data that the system can perform semantic analysis. invention.

In this embodiment of the present invention, the query text information refers to text data input by the user, or text data converted from audio data input by the user, and the query text information is the most original semantic content of the user.

In this embodiment of the present invention, the local entity library is mainly used to pre-store a knowledge base of entity content.

In the embodiment of the present invention, the preset logic rules are mainly used to accurately express natural language information and facilitate machine identification, and the preset logic rules can be correspondingly designed according to actual needs.

In the embodiment of the present invention, the translation operation refers to, after identifying the above query text information, by means of entity connection, comparing the query text information with the entity content in the above knowledge base, obtaining the text with the same entity content, The semantically represented logical rule symbol set translates the text, so as to obtain the above-mentioned logical text information.

In the embodiment of the present invention, the largest ranking entity that appears in the question is taken as the subject entity word of the question, and the entity finally queried by the user is taken as the end of the query graph link. By traversing the entities, relations and attributes in the question, the core reasoning link from the subject entity word to the ending entity word is obtained. For each core derivation link, it is converted into a tree structure for storage, and the core derivation link that is most similar to the question is obtained by comparing the similarity between the tree structure and the known intent. Finally, the link information of the core derivation link is parsed into query parameters and query process.

In this embodiment of the present invention, the semantic tree conversion rule refers to a logical rule expressing minimal semantics with "<>". We first divide the logic rules into a series of minimum semantic units, and then convert the minimum semantic units into nodes of the semantic tree (semantic combination and assembly), and build the semantic tree according to the reference relationship between the semantic units. This reference relationship represents the The sequence of queries and the way of association between knowledge graph entities.

In practical applications, leaf nodes correspond to entities of triples, non-leaf nodes correspond to relationships between entities of triples or attributes of entities, and answer nodes are located at the root node. Taking the exposure information of the counterparty of XX company as an example, XX company and the counterparty are entities, the transaction relationship is an inter-entity relationship, the exposure information is an attribute, and the final answer node asks for the attribute value, as shown in Figure 2.

By translating the query text input by the user into logical text, it effectively helps us convert natural language questions into machine-recognizable expressions, clarifies the vague expressions in questions, and presents questions with more standardized and clearer rules. It lays a solid foundation for the subsequent realization of functions based on natural language; and converts the logical text into a semantic tree, which can organize semantics in a rigorous logical order, and the conversion of logical rules and semantic trees can make understandable language information into The computer-identifiable data structure provides the basis for subsequent query operations. The verification of logic rules provides the basis for semantic inspection and repair. The semantic tree is used as the input content for question-and-answer operations, which can be combined with the contextual logic relationship of the text input by the user. Recognizing the real emotions of users can provide users with answer content that is more in line with user needs.

As some optional implementation manners of the first embodiment, the preset logic rules specifically include Unary unary rules, Binary binary rules, and Aggregation aggregation rules.

In the embodiment of the present invention, in order to accurately express natural language information and display it with machine-recognizable logical rules, three semantic unit templates are designed, namely Unary unary rule, Binary binary rule and Aggregation aggregation rule. The specific rules are as follows :

(1) Unary unary rule

For the entity e in the question, it needs to be clear whether the entity e represents a class of entities or an instance Instance. If only the category of the entity appears in the question, it is expressed as <Unary(class='E')>; if the instance value of the entity appears in the question, the corresponding entity category needs to be found, expressed as <Unary( class='E', value='Instance')>. For example, define 'organization' as an entity type. When 'company' appears in the question, it is expressed as <Unary(class='organization')>, and if 'XXX company' appears in the question, it is expressed as <Unary(class='organization', value='XXX company')>.

(2) Binary binary rule

Binary is mainly used to describe the relationship of triples, and triples can be understood as (head entity, relationship, tail entity), (head entity, attribute, attribute value) or (relation, relationship attribute, relationship attribute value). The Binary rule is to know two of the elements, to find the third element, use '? ' to mark the element to query.

Relationship class problem, S and E represent entities, rel represents relationships

Binary(S,rel?,E) returns the relationship between S and E;

Binary(S,rel,E?) returns what the rel of S is;

Binary(S?,rel,E) answers which entities have a rel relationship with E;

b) Entity attribute class problem, S means entity, PRO means attribute, value means attribute value

Binary(S, PRO, value?) returns the property PRO of S;

Binary(S?,PRO,value) seeks the entity S whose attribute PRO is value;

Binary(S, PRO?, value) asks what property of S is value;

c) Relational attribute class problem, R represents the relationship, RPRO represents the attribute of the relationship, and value represents the value of the attribute

Binary(R,RPRO,value?) What is the attribute RPRO of the relationship;

(3) Aggregation aggregation rules

A series of aggregation operators are quoted here to express the operational logic and qualification logic in the question, mainly including the following seven types:

Rank operation

Parameter settings: by—sorted property; obj—operated object; range—sorted return range

Purpose: Sorting, used to answer 'Top N', 'Nth' and other questions

Count operation

Parameter setting: obj—counted object; con—filter condition when counting

Purpose: Statistics, used to answer 'how many' similar questions

Sum operation

Parameter setting: obj—the object to be summed; con—the filter condition when adding up

Purpose: Summarize numeric attributes

Opr operation

Parameter setting: '='--calculation method (+-*/and or);

S—value 1; E—value 2; con—filter condition during calculation

Purpose: Binary computing, used to answer various computing problems

Filter operation

Parameter settings: obj--filtered object; con-filter condition

Purpose: to filter the properties of the relationship

Time operation

Parameter setting: con—time limit range

Purpose: used to express the time span of a question

Trend&Distribution operation

Parameter setting: none

Purpose: Distribution and trend calculation

The entire logic rule consists of multiple semantic units, and the syntax specification of a single semantic unit is as follows:

<unit_code:semantic_unit>

The construction of logical rules should follow the word order of the original question, and the arrangement order of the semantic units should be consistent with the semantic location in the original question.

unit_code represents the encoding of the semantic unit, and the logic rule should have only one main unit. The main unit indicates the content of the answer to be returned in the final question sentence, which is represented by A. The main unit must be executed last. The principle of semantic unit coding is to name extensions from A to the front and then back, for example: V5 V4 V3 V2 V1 AV6 V7 V8.

Semantic_unit is the triple representation mentioned in the previous unit, and will not be repeated.

Continuing to refer to FIG. 7 , a schematic structural diagram of the text acquisition module 120 in FIG. 6 is shown. For convenience of description, only the parts related to the present invention are shown.

In some optional implementations of Embodiment 2 of the present invention, the text acquisition module 120 includes: a data acquisition submodule 121, a topic determination submodule 122, an end determination submodule 123, an intermediate text acquisition submodule 124, and an optimal intent submodule module 125 and text generation sub-module 126. in:

A data acquisition sub-module 121, configured to acquire entity data corresponding to the query text information in the local entity library;

a topic determination submodule 122, configured to determine a topic entity word corresponding to the entity data based on a preset weight value;

The end determination submodule 123 is used for taking the end entity data of the query text information as the end entity word;

The intermediate text obtaining submodule 124 is used to obtain the intermediate logical text by using the subject entity word and the ending entity word as the first entity word of the initial logical text respectively;

The optimal intent sub-module 125 is used to obtain the optimal intent with the highest similarity with the intermediate logical text in the intent database;

The text generation sub-module 126 is configured to generate the logical text information based on the optimal intention.

In the embodiment of the present invention, the keyword can be obtained by identifying the content in the question, and compared with the entity existing in the local entity database, and finally the data consistent with the keyword and the entity is obtained as the entity data.

In this embodiment of the present invention, one of the multiple entities found in the topic determination sub-module 122 is selected as the topic entity word. The weight value of each entity is preset through the pointing relationship between entities in the local entity. Entities with weights greater than 0 are used as candidate sets of topic entity words, and subsequent operations are attempted on each candidate topic word in order.

In this embodiment of the present invention, the entity to be inquired at the end of the question sentence is taken as the ending entity word. When the ending entity word is a relation, try to fix it.

In the embodiment of the present invention, the subject entity words obtained in the subject determination sub-module 122 and the ending entity words obtained in the ending determination sub-module 123 are used as the first and last ends of the link to be traversed in the graph. Get the candidate subgraph set.

In the embodiment of the present invention, the similarity judgment is performed between the candidate subgraph and the existing intent in the intent library. Similarity is judged by whether entities, relationships, locations, etc. are consistent. Get the most similar one intent to return.

In this embodiment of the present invention, a logical form is generated according to the graph structure of the intent subgraph.

In practical applications, the node (entity type) corresponds to the uniary representation rule, the class corresponds to the entity type, and the value corresponds to the specific instance value, such as Ping An Life. If there is no specific entity value, uniform indicates information that can add value. Edges (relationships, attribute entities) correspond to binary representation rules. The first position of Binary is the node information corresponding to the edge, and the second position is the representation information of the edge, such as: attribute entity risk rate. The third position is filled with corresponding node information according to whether the edge is a relationship or an attribute entity. For example: There is only one Binary representation in this problem, and the second position is the risk rate, that is, the attribute entity. At this time, if there is an attribute value of the accident rate, for example, whether the value of the accident rate is greater than 20%, the attribute value of the accident rate is greater than 20%. Then the third position is filled with >20%. If there is no corresponding attribute value, the filling refers to entity A according to the definition of the logical form.

Finally, the following logical form is obtained:

<V1:Uniary(class=professional company, value=Ping An Life)>

<A:Binary(V1, risk rate, A?)>

Continuing to refer to FIG. 8 , a schematic structural diagram of the semantic conversion module 130 in FIG. 6 is shown. For convenience of description, only the parts related to the present invention are shown.

In some optional implementations of Embodiment 2 of the present invention, the above-mentioned semantic conversion module 130 includes: a semantic segmentation sub-module 131 , a node conversion sub-module 132 , and a semantic tree construction sub-module 133 . in:

The semantic segmentation sub-module 131 is used to perform a semantic segmentation operation on the logical text information to obtain a semantic unit;

A node conversion submodule 132, configured to perform a node conversion operation on the semantic unit to obtain a semantic node;

The semantic tree construction sub-module 133 is configured to construct a semantic tree based on the semantic nodes to obtain the target semantic tree.

In practical applications, the program will use <> as the node unit to split the above logic rules into five nodes. Each node creates a tree node object (each tree node object has the following attributes: name, type, value, left child node, right child node, parent node, etc.), for example, with <V3:Binary(V2?, transaction relationship, V1)> node as an example, the name of the tree node object is V3, the type is Binary, the value is (V2?, transaction relationship, V1), and the left child node, right child node, and parent node are initialized to empty. In the subsequent description, we use the name of the node object to represent the node name. For example, <V1:Uniary(class='organization', value='XX company')> is described as a V1 node.

Build a tree based on five tree node objects, mainly update the left child node, right child node, and parent node of each tree node, for example:

1) For the V3 node, update the following values: the left child node of V3 is V2, the right child node is V1, the parent node of V2 is V3, and the parent node of V1 is V3.

2) For the V1 node, update the following values: the left child node of V1 is empty, and the right child node is empty.

3) For the V4 node, update the following values: the left child node of V4 is V3, the right child node is empty, and the parent node of V3 is V4.

4) For node A, update the following values: the left child node of node A is V4, the right child node is empty, and the parent-child node of node V4 is A.

In this way, a reference relationship between tree nodes is established, and at the same time, the root node is obtained as node A, and the finally formed semantic tree is shown in Figure 2 above.

In some optional implementations of Embodiment 2 of the present invention, the above semantic analysis apparatus 100 further includes: a type acquisition submodule, an entity judgment submodule, a semantic input submodule, and an error signal output submodule. in:

a type acquisition submodule for acquiring the node type of the target semantic tree;

an entity judging submodule for judging whether an entity triple corresponding to the node type exists in the local entity library;

A semantic input sub-module, configured to execute the intelligent question answering using the target semantic tree as the input content of the intelligent question answering system if there is an entity triple corresponding to the node type in the local entity library the steps of the operation;

The error signal output sub-module is configured to output a node error signal if the entity triple corresponding to the node type does not exist in the local entity library.

To solve the above technical problems, the embodiments of the present application also provide computer equipment. Please refer to FIG. 9 for details. FIG. 9 is a block diagram of the basic structure of a computer device according to this embodiment.

The computer device 9 includes a memory 91 , a processor 92 , and a network interface 93 that communicate with each other through a system bus. It should be noted that only the computer device 9 with components 91-93 is shown in the figures, but it should be understood that implementation of all shown components is not required, and more or less components may be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculation and/or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, special-purpose Integrated circuit (ApplicationSpecific Integrated Circuit, ASIC), programmable gate array (Field-Programmable GateArray, FPGA), digital processor (Digital Signal Processor, DSP), embedded equipment, etc.

The computer equipment may be a desktop computer, a notebook computer, a palmtop computer, a cloud server and other computing equipment. The computer device can perform human-computer interaction with the user through a keyboard, a mouse, a remote control, a touch pad or a voice control device.

The memory 91 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static Random Access Memory (SRAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), Programmable Read Only Memory (PROM), Magnetic Memory, Magnetic Disk, Optical Disk, etc. In some embodiments, the memory 91 may be an internal storage unit of the computer device 9 , such as a hard disk or a memory of the computer device 9 . In other embodiments, the memory 91 may also be an external storage device of the computer device 9, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, flash memory card (FlashCard) and so on. Of course, the memory 91 may also include both the internal storage unit of the computer device 9 and its external storage device. In this embodiment, the memory 91 is generally used to store the operating system and various application software installed on the computer device 9, such as program codes of the X method and the like. In addition, the memory 91 can also be used to temporarily store various types of data that have been output or will be output.

In some embodiments, the processor 92 may be a central processing unit (Central Processing Unit, CPU), a controller, a microcontroller, a microprocessor, or other data processing chips. This processor 92 is typically used to control the overall operation of the computer device 9 . In this embodiment, the processor 92 is configured to run the program code or process data stored in the memory 91, for example, the program code for executing the X method.

The network interface 93 may include a wireless network interface or a wired network interface, and the network interface 93 is generally used to establish a communication connection between the computer device 9 and other electronic devices.

The present application also provides another implementation manner, which is to provide a computer-readable storage medium, where an X program is stored in the computer-readable storage medium, and the X program can be executed by at least one processor, so that the at least one processor A processor executes the steps of method X as described above.

From the description of the above embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus a necessary general hardware platform, and of course hardware can also be used, but in many cases the former is better implementation. Based on this understanding, the technical solution of the present application can be embodied in the form of a software product in essence or in a part that contributes to the prior art, and the computer software product is stored in a storage medium (such as ROM/RAM, magnetic disk, CD-ROM), including several instructions to make a terminal device (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) execute the methods described in the various embodiments of this application.

Obviously, the above-described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The accompanying drawings show the preferred embodiments of the present application, but do not limit the scope of the patent of the present application. This application may be embodied in many different forms, rather these embodiments are provided so that a thorough and complete understanding of the disclosure of this application is provided. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or perform equivalent replacements for some of the technical features. . Any equivalent structures made by using the contents of the description and drawings of the present application, which are directly or indirectly used in other related technical fields, are all within the scope of the patent protection of the present application.

Claims (10)

1. a semantic analysis method, is characterized in that, comprises the following steps: receiving a query request sent by the user terminal for the intelligent question answering system, where the query request at least carries query text information; reading a local entity library, and performing a translation operation on the query text information based on the local entity library and preset logic rules to obtain logical text information; Converting the logical text information into a target semantic tree based on a semantic tree conversion rule; The intelligent question answering operation is performed using the target semantic tree as the input content of the intelligent question answering system. 2 . The semantic analysis method according to claim 1 , wherein the reading of the local entity library, the translation operation is performed on the query text information based on the local entity library and preset logic rules, and the logical text information is obtained. 3 . The steps specifically include: Acquiring entity data corresponding to the query text information in the local entity library; Determine the subject entity word corresponding to the entity data based on the preset weight value; Taking the end entity data of the query text information as the end entity word; Taking the subject entity word and the ending entity word as the first entity word of the initial logical text, respectively, to obtain an intermediate logical text; Obtain the optimal intent with the highest similarity with the intermediate logical text in the intent database; The logical text information is generated based on the optimal intent. 3. The semantic analysis method according to claim 1, wherein the step of converting the logical text information into a target semantic tree based on a semantic tree conversion rule specifically comprises: performing a semantic segmentation operation on the logical text information to obtain a semantic unit; Perform a node conversion operation on the semantic unit to obtain a semantic node; A semantic tree is constructed based on the semantic nodes to obtain the target semantic tree. 4. The semantic analysis method according to claim 1, wherein after the step of converting the logical text information into a target semantic tree based on the semantic tree conversion rule, the method further comprises: obtain the node type of the target semantic tree; Judging whether there is an entity triple corresponding to the node type in the local entity library; If there is an entity triple corresponding to the node type in the local entity library, the step of performing the intelligent question answering operation using the target semantic tree as the input content of the intelligent question answering system is performed; If the entity triple corresponding to the node type does not exist in the local entity library, a node error signal is output. 5 . The semantic analysis method according to claim 1 , wherein the preset logic rules specifically include Unary unary rules, Binary binary rules, and Aggregation aggregation rules. 6 . 6. A semantic analysis device, wherein the device comprises: a request receiving module, configured to receive a query request sent by the user terminal for the intelligent question answering system, where the query request at least carries query text information; a text acquisition module, configured to read a local entity library, perform a translation operation on the query text information based on the local entity library and preset logic rules, and acquire logical text information; a semantic conversion module for converting the logical text information into a target semantic tree based on a semantic tree conversion rule; The semantic input module is configured to use the target semantic tree as the input content of the intelligent question answering system to perform intelligent question answering operations. 7. The semantic analysis device according to claim 6, wherein the text acquisition module comprises: a data acquisition sub-module for acquiring entity data corresponding to the query text information in the local entity library; a topic determination submodule for determining topic entity words corresponding to the entity data based on a preset weight value; The end determination submodule is used to use the end entity data of the query text information as the end entity word; an intermediate text acquisition submodule, which is used to respectively use the subject entity word and the ending entity word as the first entity word of the initial logical text to obtain the intermediate logical text; The optimal intent sub-module is used to obtain the optimal intent with the highest similarity with the intermediate logical text in the intent database; A text generation submodule, configured to generate the logical text information based on the optimal intent. 8. The semantic analysis device according to claim 6, wherein the semantic conversion module comprises: Semantic segmentation sub-module, used to perform semantic segmentation operation on the logical text information to obtain semantic units; a node conversion submodule, used for performing a node conversion operation on the semantic unit to obtain a semantic node; The semantic tree construction sub-module is configured to construct a semantic tree based on the semantic nodes to obtain the target semantic tree. 9. A computer device, comprising a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the method for implementing the semantic analysis method according to any one of claims 1 to 5 is realized. step. 10. A computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the computer program according to any one of claims 1 to 5 is implemented. The steps of the semantic analysis method.

Images