CN115221298A - Question and answer matching method and device, electronic equipment and storage medium - Google Patents

Abstract

The present application is applicable to the field of computer technology, and provides a question-and-answer matching method, device, electronic device and storage medium, including: obtaining question text; determining a question entity according to the question text; The target triplet corresponding to the question entity; the target triplet is converted into a natural language form to obtain candidate text; the question text is matched with the candidate text to determine the target answer information. The embodiments of the present application can efficiently and accurately implement question-and-answer matching.

Description

Question-answer matching method, device, electronic device and storage medium

technical field

The present application belongs to the field of computer technology, and in particular, relates to a question-answer matching method, device, electronic device and storage medium.

Background technique

A question answering system is an advanced form of information retrieval system that can accurately answer questions posed by users in natural language. The method running in the question answering system, that is, the question and answer matching method that determines the corresponding answer information according to the input question sentence, is a research direction that has attracted much attention and has broad development prospects in the field of artificial intelligence and natural language processing. The current question-answer matching method usually needs to match the relationship between the natural question and the triple to determine the answer information corresponding to the question, which is difficult to match, with poor accuracy and low efficiency.

SUMMARY OF THE INVENTION

In view of this, the embodiments of the present application provide a question-answer matching method, apparatus, electronic device, and storage medium, so as to solve the problem of how to efficiently and accurately implement question-answer matching in the prior art.

A first aspect of the embodiments of the present application provides a question-and-answer matching method, including:

Get the question text;

Determine the question entity according to the question text;

According to the question entity, determine the target triplet corresponding to the question entity;

Converting the target triplet into a natural language form to obtain candidate text;

Matching the question text and the candidate text to determine target answer information.

Optionally, determining the question entity according to the question text includes:

The question text is input into the trained entity recognition model for processing, and the question entity is obtained.

Optionally, determining the target triplet corresponding to the question entity according to the question entity, including:

According to the question entity, determine a candidate entity from a preset entity library;

According to the candidate entity, the target triplet is determined from a preset triplet database.

Optionally, according to the question entity, the candidate entity is determined from a preset entity library, including:

According to the question entity, obtain a first entity matching the character of the question entity from the entity library;

Determine the entity similarity between the question entity and the first entity, and filter out candidate entities from the first entity according to the entity similarity.

Optionally, the determining the entity similarity between the question entity and the first entity includes:

Perform word segmentation and part-of-speech extraction processing on the question entity, and determine the question entity keyword corresponding to the question entity;

Perform word segmentation and part-of-speech extraction processing on the first entity to determine the first entity keyword corresponding to the first entity;

Perform word-level similarity calculation on the question sentence entity keyword and the first entity keyword to obtain the entity similarity.

Optionally, determining the target triplet from a preset triplet database according to the candidate entity includes:

According to the candidate entity, the triplet whose header information matches the candidate entity is obtained from the triplet database as a candidate triplet;

According to preset filter conditions, the target triplet is filtered out from the candidate triplet.

Optionally, the matching of the question text with the candidate text to determine target answer information includes:

The question text is processed with each statement formed by each of the candidate texts, and the input preset two-classification model is processed to obtain the target classification result corresponding to each candidate text and the corresponding confidence level;

The target answer information is determined according to the target classification results and corresponding confidence levels corresponding to each of the candidate texts respectively.

A second aspect of the embodiments of the present application provides a question and answer matching device, including:

Get unit, used to get the question text;

The question entity determination unit is used to determine the question entity according to the question text;

The triplet determining unit is configured to determine the target triplet corresponding to the question entity according to the question entity;

a conversion unit for converting the target triplet into a natural language form to obtain candidate text;

A matching unit, configured to match the question text with the candidate text to determine target answer information.

A third aspect of the embodiments of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, when the processor executes the computer program At the time, the electronic device is made to implement the steps of the question-answer matching method.

A fourth aspect of the embodiments of the present application provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, enables an electronic device to implement the question-and-answer matching as described above steps of the method.

A fifth aspect of the embodiments of the present application provides a computer program product, which, when the computer program product runs on an electronic device, enables the electronic device to execute the question-answer matching method described in any one of the above-mentioned first aspects.

Compared with the prior art, the embodiment of the present application has the following beneficial effects: in the embodiment of the present application, the question text is obtained, the question entity is determined according to the question text, and the corresponding question entity is determined according to the question entity. The target triplet of . After that, convert the target triplet into a natural language form to obtain candidate text, and then match the question text with the candidate text to determine the target answer information. Since the target triplet can be determined according to the entity of the question sentence, the search matching range can be narrowed and the efficiency of question and answer matching can be improved; after that, the determined target triplet can be further converted into candidate text in the form of natural language, and then the question The sentence text is matched with the candidate text, and the target answer information is determined, that is, the question text in the form of natural language can be accurately matched with the candidate text in the form of natural language, and the relationship between the question text and the triplet can be avoided. In the case of low matching degree and difficult question-and-answer matching, question-and-answer matching can be implemented efficiently and accurately.

Description of drawings

In order to illustrate the technical solutions in the embodiments of the present application more clearly, the following briefly introduces the accompanying drawings that are required to be used in the description of the embodiments or the prior art.

Fig. 1 is the implementation flow schematic diagram of a kind of question-answer matching method provided in the embodiment of the present application;

2 is a schematic diagram of a question and answer matching device provided by an embodiment of the present application;

FIG. 3 is a schematic diagram of an electronic device provided by an embodiment of the present application.

Detailed ways

In the following description, for the purpose of illustration rather than limitation, specific details such as a specific system structure and technology are set forth in order to provide a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be practiced in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present application with unnecessary detail.

In order to illustrate the technical solutions described in the present application, the following specific embodiments are used for description.

It is to be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described feature, integer, step, operation, element and/or component, but does not exclude one or more other features , whole, step, operation, element, component and/or the presence or addition of a collection thereof.

It should also be understood that the terminology used in the specification of the application herein is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural unless the context clearly dictates otherwise.

It should also be further understood that, as used in this specification and the appended claims, the term "and/or" refers to and including any and all possible combinations of one or more of the associated listed items .

As used in this specification and the appended claims, the term "if" may be contextually interpreted as "when" or "once" or "in response to determining" or "in response to detecting" . Similarly, the phrases "if it is determined" or "if the [described condition or event] is detected" may be interpreted, depending on the context, to mean "once it is determined" or "in response to the determination" or "once the [described condition or event] is detected. ]" or "in response to detection of the [described condition or event]".

In addition, in the description of the present application, the terms "first", "second", "third", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

For ease of understanding, first, some related concepts of the embodiments of the present application are explained as follows:

Entity: An entity is an abstraction of an objective individual. A person, a movie, or a phrase can be regarded as an entity. For example: Star A, "Journey to the West", "Beijing's Forbidden City", etc.

Triplet: A knowledge representation consisting of three elements, "head", "relation" and "tail", which usually correspond to the subject, predicate and predicate of a sentence, respectively. Object. For example: ["Qing Government", "Signed", "Tianjin Treaty"], ["Shenzhen City", "Area", "1997.47 square kilometers"]. Compared with two-dimensional table storage data, triplet storage data has the characteristics of less space occupation, good sparsity and flexibility, and is often used in question answering systems.

At present, triple-based question-answer matching methods usually include two main steps: entity recognition and relation extraction. Among them, entity recognition is to identify the entity in the question, that is, the subject content that needs to be understood in the question, while relation extraction is to identify which attribute the question asks about the entity in the question. For example, for the question "What is the area of Shenzhen?", the entity to be identified is "Shenzhen", and the relationship to be extracted is "area". However, due to the huge gap between the method of defining relations in triples and the methods of using relations in natural language, the extracted scheme needs to be mapped and matched, which cannot achieve good accuracy, resulting in questions in the form of natural language in triples. The matching degree is low and the matching difficulty is high when finding the answer of relationship matching in the group. Moreover, when the amount of data is large, the retrieval matching speed is slow. The current question-answer matching method usually needs to match the relationship between the natural question and the triple to determine the answer information corresponding to the question, which is difficult to match, with poor accuracy and low efficiency.

In order to solve the above-mentioned technical problems, embodiments of the present application provide a question-and-answer matching method, device, electronic device, and storage medium, including: acquiring question text; determining a question entity according to the question text; sentence entity, determine the target triplet corresponding to the question entity; convert the target triplet into a natural language form to obtain candidate text; match the question text with the candidate text to determine the target answer information.

Since the target triplet can be determined according to the entity of the question sentence, the search matching range can be narrowed and the efficiency of question and answer matching can be improved; after that, the determined target triplet can be further converted into candidate text in the form of natural language, and then the question The sentence text is matched with the candidate text, and the target answer information is determined, that is, the question text in the form of natural language can be accurately matched with the candidate text in the form of natural language, and the relationship between the question text and the triplet can be avoided. In the case of low matching degree and difficult question-and-answer matching, question-and-answer matching can be implemented efficiently and accurately.

Example 1:

1 shows a schematic flowchart of a question-and-answer matching method provided by an embodiment of the present application. The execution subject of the question-and-answer matching method may be an electronic device such as a computer, a mobile phone, or a robot, and the details are as follows:

In S101, the question text is acquired.

In this embodiment of the present application, the question text is a sentence in which the user sends a question. For example, "What is the area of Shenzhen?" "I want to know how tall is star A?" and so on.

The electronic device acquires the question text input by the user. In one embodiment, the character information input by the user through the keyboard of the electronic device can be obtained to obtain the question text. In another embodiment, the user's voice information may be acquired through a microphone of the electronic device, and text recognition may be performed on the voice information to obtain the question text.

In S102, the question entity is determined according to the question text.

In the embodiment of the present application, the question entity is an entity included in the question, and generally is the subject content to be understood in the question. For example, for the question text "What is the area of Shenzhen?", the question entity is "Shenzhen"; for the question text "I want to know how tall is the star A?", the question entity is " Star A".

After the question sentence is acquired, through the preset entity recognition algorithm, the entity information contained in the question sentence text is identified as the question sentence entity.

In S103, a target triplet corresponding to the question entity is determined according to the question entity.

After determining the question entity in the question text, search for a target triplet associated with the question entity from a preset triplet database according to the question entity, and the number of the target triplet can be one or more. In one embodiment, a triple containing the information of the question entity can be directly determined as a target triple; in another embodiment, a triple whose header information is consistent with the information of the question entity can be determined The group is determined as the target triplet.

In S104, the target triplet is converted into a natural language form to obtain candidate text.

After determining the target triplet associated with the question entity, convert the target triplet to text in natural language form. These texts converted from target triples are candidate texts that may match the question texts, and these texts are called candidate texts.

In one embodiment, the target triplet can be converted into corresponding natural language text according to a preset natural language text template, that is, the candidate text can be obtained. Exemplarily, the natural language text template may be: [head entity]++[relationship]+is+[tail entity]. If a current target triplet is ["Shenzhen City", "Area", "1997.47 square kilometers"], then according to the natural language text template, a corresponding candidate text is obtained: the area of Shenzhen City is 1997.47 square kilometers .

In S105, the question text and the candidate text are matched to determine target answer information.

In the embodiment of the present application, after the candidate text corresponding to the target triplet is determined, the question text is matched with the candidate text through a preset matching algorithm to obtain target answer information for answering the question text. Afterwards, the target answer information can be fed back to the user by means of screen display, voice playback, and information push.

In one embodiment, if there are multiple candidate texts determined in step S104, a preset matching algorithm is used to determine the text with the highest matching degree with the question text from the multiple candidate texts as the target text. After that, the target text can be directly used as the target answer information; for example, for the question text "What is the area of Shenzhen?", the corresponding target text "Shenzhen's area is 1997.47 square kilometers" can be directly output as the target answer information . Or, the triplet corresponding to the target text is used as the target answer information; for example, the triplet corresponding to the target text ["Shenzhen", "area", "1997.47 square kilometers"] is used as the target answer information, the target The answer information is transmitted to the next level system for processing. Or, after determining the triplet corresponding to the target text, the tail information of the triplet is used as the target answer information; The tail information "1997.47 square kilometers" in the "square kilometer"] is fed back to the user as the target answer information.

In the embodiment of the present application, since the target triplet can be determined according to the entity of the question sentence, the search matching scope can be narrowed, and the question-answer matching efficiency can be improved; after that, the determined target triplet can be further converted into candidates in the form of natural language After the text, the question text is matched with the candidate text to determine the target answer information, that is, the question text in the form of natural language can be accurately matched with the candidate text in the form of natural language, avoiding the question text and triples. In the case of direct matching, the relationship matching degree is low and the question and answer matching is difficult, so the answer information corresponding to the question can be determined efficiently and accurately.

Optionally, determining the question entity according to the question text includes:

The question text is input into the trained entity recognition model for processing, and the question entity is obtained.

In the embodiment of the present application, the trained entity recognition model is a neural network model obtained by offline training in advance and used for entity recognition. In one embodiment, the entity recognition model can be obtained by tuning based on an existing pre-trained language model. Exemplarily, a sentence marked with an entity position can be used as the entity recognition sample data, and the language model can be trained or tuned to obtain an entity recognition model. In some embodiments, the entity recognition sample data can be marked in the form of BIO, where B (Begin) is used to mark the start position of the entity, I (Inner) is used to mark the non-start position of the entity, and O (Other) Used to label non-solid locations. For example, for the entity recognition sample data "The height of star A is XX cm", the corresponding BIO is marked as "BIIOOOOOOOOO".

Input the question text into the above-mentioned trained entity recognition model, the entity recognition model can locate the location of the entity in the question text, and extract the entity at the position as the question entity.

In the embodiment of the present application, since the entity recognition of the question text can be accurately performed by the trained entity recognition model, the question entity can be accurately extracted, and the accuracy of the question-answer matching method can be ensured.

Optionally, determining the target triplet corresponding to the question entity according to the question entity, including:

According to the question entity, determine a candidate entity from a preset entity library;

According to the candidate entity, the target triplet is determined from a preset triplet database.

In the embodiment of the present application, the preset entity database stores a large number of entities that users may ask questions in advance, and the preset triplet database stores a large number of triples corresponding to these entities. The answer corresponding to the sentence text.

After the question entity is extracted from the question text, an entity that is consistent, similar and synonymous with the information of the question entity is searched from the preset entity database as a candidate entity. Afterwards, according to the candidate entity, a triplet related to the candidate entity is obtained from a preset triplet database as a target triplet.

In the embodiment of the present application, since the candidate entity matching the question entity can be screened from the entity database first, the entity information to be asked by the user can be obtained more accurately and comprehensively, so that the target three can be more accurately determined according to the candidate entity. Tuples to improve the accuracy of question answer matching.

Optionally, according to the question entity, the candidate entity is determined from a preset entity library, including:

According to the question entity, obtain a first entity matching the character of the question entity from the entity library;

Determine the entity similarity between the question entity and the first entity, and filter out candidate entities from the first entity according to the entity similarity.

In this embodiment of the present application, after determining the question entity, first, according to the character-level matching method, an entity matching the characters of the question entity is selected from the entity library as the first entity, and the number of the first entities may be multiple. In one embodiment, the first entity corresponding to the question entity may be determined by an inverted index retrieval engine.

After preliminary screening to obtain each first entity matching the characters of the question entity, the similarity between the question entity and the first entity is calculated (this similarity is called entity similarity), and according to the entity similarity , screen out the top K entities with the highest entity similarity with the question entity from the first entity as candidate entities (where K is a positive integer), or screen out the first entity whose similarity is greater than the preset similarity threshold. An entity serves as a candidate entity. In one embodiment, when calculating the similarity between the question entity and the first entity, the question entity may be segmented first and converted into a corresponding question entity vector according to the word level, and the first entity may be segmented according to the word level. The word level is converted into the corresponding first entity vector, and then the cosine similarity algorithm is used to calculate the word level cosine similarity between the question entity and the first entity as the entity similarity.

In the embodiment of the present application, since the first entity matching the entity character of the question sentence can be quickly screened from the entity library according to the word level, and then the candidate entity can be further accurately screened from the first entity through similarity calculation, Therefore, candidate entities related to question entities can be efficiently and accurately determined, and the efficiency and accuracy of question-answer matching can be improved.

Optionally, the determining the entity similarity between the question entity and the first entity includes:

Perform word segmentation and part-of-speech extraction processing on the question entity, and determine the question entity keyword corresponding to the question entity;

Perform word segmentation and part-of-speech extraction processing on the first entity to determine the first entity keyword corresponding to the first entity;

Perform word-level similarity calculation on the question sentence entity keyword and the first entity keyword to obtain the entity similarity.

In the embodiment of the present application, word segmentation processing and part-of-speech extraction processing may be performed for both the question entity and the first entity, respectively, and important words are reserved according to the part of speech to obtain the question entity keyword and the first entity keyword. The word segmentation process can be implemented by a preset word segmentation algorithm; the part-of-speech extraction process can be implemented by open source tools. Exemplarily, the part-of-speech tags and meanings of words contained in entities can be as shown in Table 1:

In one embodiment, after the word segmentation process and part of speech extraction process, words whose part of speech tags are n, nr, ns, nt, nw, nz, PER, and LOC may be determined as important words, and the words belonging to the question entity may be determined as important words. Such important words are reserved to obtain question sentence entity keywords; and such important words in the first entity are reserved to obtain first entity keywords. Exemplarily, assuming that the current question entity is "the Forbidden City in Beijing", after word segmentation and part-of-speech extraction processing, the corresponding question entity keywords are "Beijing" and "Forbidden City".

After obtaining the question entity keyword and the first entity keyword, determine the question keyword vector corresponding to the question entity and the first entity key corresponding to the first entity according to the question entity keyword and the first entity keyword word vector. After that, the cosine similarity between the question keyword vector and the first entity keyword can be obtained, so as to realize the word-level similarity calculation between the question entity and the first entity, and obtain the entity similarity.

In the embodiment of the present application, since the entity keyword and the first entity keyword of the question sentence can be obtained through word segmentation and part-of-speech extraction, and then the similarity calculation at the word level can be performed, it is possible to prevent irrelevant characters from affecting the accuracy of the similarity calculation. This ensures the accuracy of subsequent entity screening, thereby improving the accuracy of question-answer matching.

Optionally, determining the target triplet from a preset triplet database according to the candidate entity includes:

According to the candidate entity, the triplet whose header information matches the candidate entity is obtained from the triplet database as a candidate triplet;

According to preset filter conditions, the target triplet is filtered out from the candidate triplet.

In the embodiment of the present application, since the header information of the triplet is usually the subject content of the question to be asked, after screening the candidate entities according to the entity of the question, the header can be filtered from the preset triplet database The triples whose information matches the candidate entity are regarded as candidate triples related to the question text.

After that, according to the preset filter conditions, the target triplet is further screened from the candidate triplet, so as to narrow the subsequent matching range and improve the question-answer matching efficiency. The preset filter conditions may include conditions related to the character length, source information, and editing times of the triplet. Exemplarily, the preset filtering conditions include: filtering out triples whose character length is greater than a preset length; filtering out triples whose source is a specified source; filtering out triples whose number of edits is less than the preset number of times. Tuples are filtered out.

Optionally, the matching of the question text with the candidate text to determine target answer information includes:

The question text is processed with each statement formed by each of the candidate texts, and the input preset two-classification model is processed to obtain the target classification result corresponding to each candidate text and the corresponding confidence level;

The target answer information is determined according to the target classification results and corresponding confidence levels corresponding to each of the candidate texts respectively.

In the embodiment of the present application, the preset two-category model is a two-sentence two-category language model obtained by training in advance and capable of performing two-category discrimination on the matching result between the candidate text and the question text, wherein the target classification result of the two-category can be classified into There are two categories of "can answer the question" and "can't answer the question". In one embodiment, the pre-trained language model can be fine-tuned by using the sample data of the labeled classification categories, so as to obtain the binary classification model. Among them, each sample data is double-sentence data containing a question and an answer. In the label of the sample data, "1" is used to mark "can answer the question", indicating that the answer in the double-sentence data can answer the question. The question of the question; "0" indicates "cannot answer the question", indicating that the answer in the double sentence data cannot answer the question of the question. An example of some sample data is shown below:

1 How tall is star A, star A's height is xxxcm

0 How tall is star A? The weight of star A is xxxkg

1 How many people are there in Shenzhen? Shenzhen's population is xxx million

0 How many people are there in Shenzhen? The area of Shenzhen is xxx square kilometers

In the embodiment of the present application, after the current candidate text is determined, the current question text and each candidate text can be respectively formed into sentence pairs (also referred to as double sentence data); after that, the sentence pairs are input into the binary classification After the model is processed, the target classification result and confidence level corresponding to the candidate text in the sentence pair can be obtained. Among them, the confidence is the probability that the candidate text is classified into the current target classification result, and its value range is [0, 1].

In one embodiment, the above-mentioned pair of pre-trained language models can be a BERT model (a bidirectional language model), and the sentence pairs (also called double sentences) input into the BERT model can be processed to obtain a CLS latent vector, based on the CLS latent vector. The CLS latent vector can obtain the confidence corresponding to the candidate question in the sentence pair.

After determining the target classification results corresponding to each candidate text and the corresponding confidence level, a candidate text with the highest confidence level can be selected from the candidate texts whose target classification result is "can answer the question" as the target text. Then, according to the target text, target answer information is determined.

In the embodiment of the present application, since the target classification result and confidence level of the candidate text can be accurately determined based on the preset binary classification model, the target answer information can be accurately determined from the candidate text, and the accuracy of question-answer matching can be improved.

It should be understood that the size of the sequence numbers of the steps in the above embodiments does not mean the sequence of execution, and the execution sequence of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

Embodiment 2:

FIG. 2 shows a schematic structural diagram of a question and answer matching apparatus provided by an embodiment of the present application. For the convenience of description, only parts related to the embodiment of the present application are shown:

The question and answer matching apparatus includes: an acquisition unit 21 , a question entity determination unit 22 , a triple determination unit 23 , a conversion unit 24 and a matching unit 25 . in:

The obtaining unit 21 is used to obtain the question text.

The question entity determination unit 22 is configured to determine the question entity according to the question text.

The triplet determining unit 23 is configured to determine the target triplet corresponding to the question entity according to the question entity.

The conversion unit 24 is configured to convert the target triplet into a natural language form to obtain candidate texts.

The matching unit 25 is configured to match the question text with the candidate text to determine target answer information.

Optionally, the question entity determining unit 22 is specifically configured to input the question text into a trained entity recognition model for processing to obtain a question entity.

Optionally, the triple determination unit 23 includes:

a candidate entity determination module, configured to determine candidate entities from a preset entity library according to the question entity;

A target triplet determination module, configured to determine the target triplet from a preset triplet database according to the candidate entity.

Optionally, the candidate entity determination module includes:

a first entity determining unit, configured to obtain, according to the question entity, a first entity that matches the character of the question entity from the entity library;

A candidate entity determination unit, configured to determine the entity similarity between the question entity and the first entity, and filter out candidate entities from the first entity according to the entity similarity.

Optionally, in the candidate entity determining unit, the determining the entity similarity between the question entity and the first entity includes:

Perform word segmentation and part-of-speech extraction processing on the question entity, and determine the question entity keyword corresponding to the question entity;

Perform word segmentation and part-of-speech extraction processing on the first entity to determine the first entity keyword corresponding to the first entity;

Perform word-level similarity calculation on the question sentence entity keyword and the first entity keyword to obtain the entity similarity.

Optionally, the target triplet determination module is specifically configured to obtain, according to the candidate entity, a triplet whose header information matches the candidate entity from the triplet database as a candidate triplet ; Screen out the target triplet from the candidate triplet according to the preset filter conditions.

Optionally, the matching unit 25 is specifically configured to process each sentence composed of the question text and each of the candidate texts into a preset two-class model to obtain the target classification corresponding to each candidate text. The result and the corresponding confidence level; the target answer information is determined according to the target classification result and the corresponding confidence level corresponding to each of the candidate texts respectively.

It should be noted that the information exchange, execution process and other contents between the above-mentioned devices/units are based on the same concept as the method embodiments of the present application. For specific functions and technical effects, please refer to the method embodiments section. It is not repeated here.

Those skilled in the art can clearly understand that, for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned functions can be allocated to different functional units, Module completion, that is, dividing the internal structure of the device into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment may be integrated in one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit, and the above-mentioned integrated units may adopt hardware. It can also be realized in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing from each other, and are not used to limit the protection scope of the present application. For the specific working processes of the units and modules in the above-mentioned system, reference may be made to the corresponding processes in the foregoing method embodiments, which will not be repeated here.

Embodiment three:

FIG. 3 is a schematic diagram of an electronic device provided by an embodiment of the present application. As shown in FIG. 3 , the electronic device 3 of this embodiment includes: a processor 30 , a memory 31 , and a computer program 32 stored in the memory 31 and executable on the processor 30 , such as a question-answer matching program. When the processor 30 executes the computer program 32 , the steps in each of the above-mentioned question-answer matching method embodiments are implemented, for example, steps S101 to S105 shown in FIG. 1 . Alternatively, when the processor 30 executes the computer program 32, the functions of the modules/units in the above device embodiments, such as the functions of the acquisition unit 21 to the matching unit 25 shown in FIG. 2, are implemented.

Exemplarily, the computer program 32 can be divided into one or more modules/units, and the one or more modules/units are stored in the memory 31 and executed by the processor 30 to complete the this application. The one or more modules/units may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program 32 in the electronic device 3 .

The electronic device 3 may be a computing device such as a desktop computer, a notebook, a palmtop computer, and a cloud server. The electronic device may include, but is not limited to, the processor 30 and the memory 31 . Those skilled in the art can understand that FIG. 3 is only an example of the electronic device 3 , and does not constitute a limitation on the electronic device 3 , and may include more or less components than those shown in the figure, or combine some components, or different components For example, the electronic device may further include an input and output device, a network access device, a bus, and the like.

The so-called processor 30 may be a central processing unit (Central Processing Unit, CPU), and may also be other general-purpose processors, digital signal processors (Digital Signal Processors, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), Field-Programmable Gate Array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The memory 31 may be an internal storage unit of the electronic device 3 , such as a hard disk or a memory of the electronic device 3 . The memory 31 may also be an external storage device of the electronic device 3, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) equipped on the electronic device 3. card, flash card (Flash Card) and so on. Further, the memory 31 may also include both an internal storage unit of the electronic device 3 and an external storage device. The memory 31 is used to store the computer program and other programs and data required by the electronic device. The memory 31 can also be used to temporarily store data that has been output or will be output.

Those skilled in the art can clearly understand that, for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned functions can be allocated to different functional units, Module completion, that is, dividing the internal structure of the device into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment may be integrated in one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit, and the above-mentioned integrated units may adopt hardware. It can also be realized in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing from each other, and are not used to limit the protection scope of the present application. For the specific working processes of the units and modules in the above-mentioned system, reference may be made to the corresponding processes in the foregoing method embodiments, which will not be repeated here.

In the foregoing embodiments, the description of each embodiment has its own emphasis. For parts that are not described or described in detail in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled artisans may implement the described functionality using different methods for each particular application, but such implementations should not be considered beyond the scope of this application.

In the embodiments provided in this application, it should be understood that the disclosed apparatus/electronic device and method may be implemented in other manners. For example, the above-described embodiments of the apparatus/electronic device are only illustrative. For example, the division of the modules or units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units. Or components may be combined or may be integrated into another system, or some features may be omitted, or not implemented. On the other hand, the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of devices or units, and may be in electrical, mechanical or other forms.

The units described as separate components may or may not be physically separated, and components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution in this embodiment.

In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware, or may be implemented in the form of software functional units.

The integrated modules/units, if implemented in the form of software functional units and sold or used as independent products, may be stored in a computer-readable storage medium. Based on this understanding, the present application can implement all or part of the processes in the methods of the above embodiments, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer When the program is executed by the processor, the steps of the foregoing method embodiments can be implemented. Wherein, the computer program includes computer program code, and the computer program code may be in the form of source code, object code, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM, Read-Only Memory) , Random Access Memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the readable medium of the computer can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction area. For example, in certain jurisdictions, according to legislation and patent practice, computer readable medium Excluding electric carrier signals and telecommunications signals.

The above-mentioned embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the above-mentioned embodiments, those of ordinary skill in the art should understand that: it is still possible to implement the above-mentioned implementations. The technical solutions described in the examples are modified, or some technical features thereof are equivalently replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions in the embodiments of the application, and should be included in the within the scope of protection of this application.

Claims (10)

1. a question and answer matching method, is characterized in that, comprises: Get the question text; According to the question text, determine the question entity; According to the question entity, determine the target triplet corresponding to the question entity; Converting the target triplet into a natural language form to obtain candidate text; Matching the question text and the candidate text to determine target answer information. 2. The question-answer matching method as claimed in claim 1, wherein, determining the question entity according to the question text, comprising: The question text is input into the trained entity recognition model for processing, and the question entity is obtained. 3. The question-answer matching method according to claim 1, wherein determining the target triplet corresponding to the question entity according to the question entity, comprising: According to the question entity, determine a candidate entity from a preset entity library; According to the candidate entity, the target triplet is determined from a preset triplet database. 4. The question-and-answer matching method of claim 2, wherein determining a candidate entity from a preset entity library according to the question entity, comprising: According to the question entity, obtain a first entity matching the character of the question entity from the entity library; Determine the entity similarity between the question entity and the first entity, and filter out candidate entities from the first entity according to the entity similarity. 5. The question-answer matching method of claim 4, wherein the determining the entity similarity between the question entity and the first entity comprises: Perform word segmentation and part-of-speech extraction processing on the question entity, and determine the question entity keyword corresponding to the question entity; Perform word segmentation and part-of-speech extraction processing on the first entity to determine the first entity keyword corresponding to the first entity; Perform word-level similarity calculation on the question sentence entity keyword and the first entity keyword to obtain the entity similarity. 6. The question-answer matching method according to claim 3, wherein determining the target triplet from a preset triplet database according to the candidate entity, comprising: According to the candidate entity, the triplet whose header information matches the candidate entity is obtained from the triplet database as a candidate triplet; According to preset filter conditions, the target triplet is filtered out from the candidate triplet. 7. The question-and-answer matching method according to any one of claims 1 to 6, wherein the matching of the question text with the candidate text to determine target answer information comprises: The question text is processed with each statement formed by each of the candidate texts, and the input preset two-classification model is processed to obtain the target classification result corresponding to each candidate text and the corresponding confidence level; The target answer information is determined according to the target classification results and corresponding confidence levels corresponding to each of the candidate texts respectively. 8. A question and answer matching device, comprising: Get unit, used to get the question text; The question entity determination unit is used to determine the question entity according to the question text; The triplet determining unit is configured to determine the target triplet corresponding to the question entity according to the question entity; a conversion unit for converting the target triplet into a natural language form to obtain candidate text; A matching unit, configured to match the question text with the candidate text to determine target answer information. 9. An electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the electronic The device implements the steps of the method as claimed in any one of claims 1 to 7. 10. A computer-readable storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, the electronic device is made to implement any one of claims 1 to 7 the steps of the method.

Images