CN111782789A

CN111782789A - Intelligent question and answer method and system

Info

Publication number: CN111782789A
Application number: CN202010636798.4A
Authority: CN
Inventors: 郑鹏程; 张远; 林森鸣
Original assignee: Jiangsu Hantao Software Technology Co ltd
Current assignee: Jiangsu Hantao Software Technology Co ltd
Priority date: 2020-07-03
Filing date: 2020-07-03
Publication date: 2020-10-16

Abstract

The invention provides an intelligent question answering method and system, and relates to the field of machine learning. The intelligent question answering method comprises the following steps: the method comprises the steps of obtaining a first question in a first text, obtaining a plurality of candidate answers in a second text, wherein each candidate answer is a combination of continuous answers with the number smaller than a preset value, and according to the similarity between the first question and each candidate answer, the candidate answer with the highest similarity is used as the answer of the first question. In addition, the invention also provides an intelligent question-answering system, which comprises: the device comprises a first acquisition module, a second acquisition module and a calculation module. The answer corresponding to the best question can be obtained through similarity calculation.

Description

Intelligent question and answer method and system

Technical Field

The invention relates to the field of machine learning, in particular to an intelligent question answering method and system.

Background

A Question-Answering system (QA for short) can receive the Question input by the user in the form of natural language, i.e. a system capable of returning a simple and accurate answer, by comprehensively using the technologies of knowledge representation, information retrieval, natural language processing, etc. Compared with the traditional search engine, the automatic question-answering system has the advantages of being more convenient and more accurate, and is a research hotspot in the fields of current natural language processing and artificial intelligence.

The existing question-answering system can not well convert words in a question into database language elements, so that answers closest to correct answers can not be well matched.

Disclosure of Invention

The invention aims to provide an intelligent question-answering method which can calculate the answer corresponding to the best question through the similarity.

Another objective of the present invention is to provide an intelligent question answering system, which can execute a first obtaining module, a second obtaining module and a calculating module.

The embodiment of the invention is realized by the following steps:

in a first aspect, an embodiment of the present application provides an intelligent question-answering method, which includes obtaining a first question in a first text, obtaining a plurality of candidate answers in a second text, where each candidate answer is a combination of consecutive answers with a number smaller than a preset value, and according to a similarity between the first question and each candidate answer, using a candidate answer with a highest similarity as an answer to the first question.

In some embodiments of the present invention, the method for acquiring the similarity includes: based on the first question and the candidate answer, a source question vector and a plurality of candidate answer vectors are obtained through a text embedding model respectively.

In some embodiments of the present invention, the method for obtaining the similarity further includes: determining similarity through a similarity model based on the source question vector and each candidate answer vector; wherein, the similarity model is a machine learning model.

In some embodiments of the present invention, the above-mentioned, according to the similarity between the first question and each of the candidate answers, using the candidate answer with the highest similarity as the answer of the first question includes: and taking one or more candidate answers with the similarity meeting the preset conditions as candidate answers, and acquiring user input to determine a first answer.

In some embodiments of the invention, the above includes: a first question is looked up in a plurality of source text units, wherein the source text units use a language corresponding to the first text and contain content different from the first text.

In some embodiments of the invention, the above includes: and for each source text unit containing the first question, taking the corresponding target text unit as a third text, and acquiring a second question in the target text unit.

In some embodiments of the invention, the above includes: and acquiring a plurality of candidate answers in the fourth text, wherein each candidate answer is a combination of continuous answers with the number smaller than a preset value, acquiring the similarity between the second question and each candidate answer, and determining the similarity as the candidate answer of the second question.

In a second aspect, an embodiment of the present application provides an intelligent question-answering system, which includes a first obtaining module, configured to obtain a first question in a first text; the second obtaining module is used for obtaining a plurality of candidate answers in the second text, wherein each candidate answer is a combination of continuous answers of which the number is smaller than a preset value; and the calculation module is used for determining the candidate answer with the highest similarity as the answer of the first question according to the similarity between the first question and each candidate answer.

Compared with the prior art, the embodiment of the invention has at least the following advantages or beneficial effects:

through training of the similarity model, the original text pair confirmed by the user can be obtained as a positive sample, words in the original text are replaced by similar words to be used as negative samples, and parameters of the model are updated through a gradient descent method based on training data. By the method, dependence on the automatic question-answering model can be avoided, selection deviation caused by interference of the automatic question-answering model is reduced, and the similarity model with better effect is obtained through sufficient corpus selection, so that similarity calculation is better completed. And calculating to obtain an answer corresponding to the optimal question through the similarity.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.

Fig. 1 is an exemplary flowchart of an intelligent question answering method according to an embodiment of the present invention;

FIG. 2 is an exemplary flow chart of a method for intelligent question answering according to an embodiment of the present invention;

fig. 3 is a schematic diagram of an intelligent question answering system according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some embodiments of the present application, but not all embodiments. The components of the embodiments of the present application, generally described and illustrated in the figures herein, can be arranged and designed in a wide variety of different configurations.

Thus, the following detailed description of the embodiments of the present application, presented in the accompanying drawings, is not intended to limit the scope of the claimed application, but is merely representative of selected embodiments of the application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined and explained in subsequent figures. Meanwhile, in the description of the present application, the terms "first", "second", and the like are used only for distinguishing the description, and are not to be construed as indicating or implying relative importance.

Some embodiments of the present application will be described in detail below with reference to the accompanying drawings. The embodiments described below and the individual features of the embodiments can be combined with one another without conflict.

Example 1

Referring to fig. 1, fig. 1 is a flowchart illustrating an exemplary intelligent question answering method according to an embodiment of the present application, including the following steps:

step S110, acquiring a first question in a first text;

specifically, the first text may refer to text that is asked a question. The first text may include text in various languages. For example, english, chinese, or japanese.

In some embodiments, the first text may be a sentence or a paragraph of the total text that is asked a question. The first text may be selected by the user among all texts, for example, the user selects a sentence or a paragraph in succession as the first text by mouse. The first text may also be automatically selected by the auxiliary software, for example, by using the original text corresponding to the paragraph currently processed by the user as the first text.

In some embodiments, the first question is a word or combination of words that need to be solved. A problem generally refers to words in the original text that express concepts in a specific area or subject area, with obvious specialized and standardized features, and also words that appear in the original text at high frequency. In particular, the first question may be a non-standard question, there is no uniform answer, and it is difficult to automatically obtain an accurate answer thereof.

In some embodiments, the first question may be a keyword or phrase in the first text. The first question may be selected in the first text by the user through the user terminal or manually entered.

Step S120, a plurality of candidate answers in the second text are obtained, and each candidate answer is a combination of continuous answers of which the number is smaller than a preset value;

specifically, the second text refers to the solved text corresponding to the first text. The second text may come from a server, storage device, other interface, or content uploaded by the user, entered, etc., as well as obtained by other means.

In some embodiments, the second text may be obtained by an automatic solution by a solution program, or by obtaining input after a manual solution, or derived from stored pre-solved content. For example, the first text may be a sentence containing the first question "traffic data": the processor may be a macro object of the traffic data related to the road segment, and the corresponding second text may be: the processor may obtain traffic data related to the road segment.

In some embodiments, the processing device may obtain a plurality of candidate answers in the second text. The candidate answer is a combination of continuous words with the number smaller than a preset value in the second text. The candidate answer may include a word, words, or a combination of words. The preset value refers to an upper limit of the number of words that can be included in a candidate answer. The preset value can be preset by a user or can be set by default by the processing equipment. For example, the preset value may be 2, 3, 4, 5, 6, etc., preferably, the preset value may be 3 or 4. For example, the second text is: the processor may obtain traffic data related to the road segment. If the preset value is 3, then the following are all candidate answers: processor, may, processor/may/obtaining, related/related, related/traffic, traffic/data, etc. Taking "processor/available/obtained" as an example, this is a candidate answer consisting of three consecutive words that are available and obtained by the processor. While the following phrases are not candidate answers due to discontinuities: processor/get, related/traffic/data. The candidate answers may be obtained by sequentially obtaining combinations of consecutive words after word segmentation. The grouping can be realized by word segmentation tools such as ansj, jieba, hand and the like, and the description is not limited.

Step S130, according to the similarity between the first question and each candidate answer, the candidate answer with the highest similarity is used as the answer to the first question.

Specifically, the similarity of the first question to each candidate answer reflects the semantic proximity between the first question and each candidate answer. The similarity may be obtained in various ways.

In some embodiments, the similarity model may be a machine learning model. Such as neural network models. In some embodiments, the first question vector and each candidate answer vector are input to the neural network model separately to determine the similarity. The model has the input of the first question vector and each candidate answer vector, and the output of the model is the similarity of the first question and each candidate answer.

The similarity model may be obtained by training. During the training of the similarity model, the original text pair confirmed by the user can be obtained as a positive sample, the words in the original text are replaced by the similar words as negative samples, and the parameters of the model are updated through a gradient descent method based on the training data.

By the method, dependence on the automatic question-answering model can be avoided, selection deviation caused by interference of the automatic question-answering model is reduced, and the similarity model with better effect is obtained through sufficient corpus selection, so that similarity calculation is better completed.

Example 2

Referring to fig. 2, fig. 2 is a flowchart illustrating an exemplary intelligent question answering method according to an embodiment of the present application, including the following steps:

step S200, acquiring a first question in a first text;

Step S210, a first question is searched for in a plurality of source text units, wherein the source text units use a language corresponding to the first text and contain contents different from the first text.

Specifically, the source text unit uses a language corresponding to the first text, and the source text unit is the remaining text units in the original text, such as sentences, paragraphs, and the like, except for the first text. The source text units can be divided in advance according to preset conditions or can be divided by a user in a user-defined mode.

In some embodiments, after obtaining the first question from the first text, the processing device may look up the first question in a plurality of source text units, and may obtain a source text unit containing the first question. For example, the first question is "how to differentiate? "by looking up the first question in the aforementioned m source text units, the source text unit containing the first question can be obtained, e.g., the source text unit containing the first question can be source text unit 1, source text unit 2, etc.

In step S220, a plurality of candidate answers in the second text are obtained, where each candidate answer is a combination of consecutive answers whose number is smaller than a preset value.

In step S230, according to the similarity between the first question and each candidate answer, the candidate answer with the highest similarity is used as the answer to the first question.

Specifically, based on the first question vector and each candidate answer vector, determining similarity through a similarity model; wherein, the similarity model is a machine learning model.

In some embodiments, the similarity may be determined based on a vector distance of the first question vector from each candidate answer vector. For example, the similarity between the source question vector and each candidate answer vector may be calculated using a cosine method, an euclidean distance method, a mahalanobis distance method, a minkowski distance method, a hamming distance method, or the like.

By the method, similarity calculation can be completed by directly applying a common translation model and a word embedding model without depending on special training. The method is convenient to implement, avoids the influence on the accuracy when the training sample is insufficient, and improves the quality of the auxiliary translation work.

Step S230, regarding each source text unit containing the first question, taking the corresponding target text unit as the third text.

Specifically, the target text unit is a translation of the corresponding source text unit, and the target text unit uses a language corresponding to the third text. For example, the language used for the second text is Chinese, and the language used for the target text unit is also Chinese.

In some embodiments, for each source text unit that contains the first question, the content of its corresponding target text unit is taken as the third text. For example, source text element 1, source text element 2 contain the first question "how to differentiate? ", the corresponding target text unit 1 may be the third text of source text unit 1, and the target text unit 2 may be the third text of source text unit 2.

In some embodiments, the target text unit corresponding to each source text unit containing the first question is used as the third text, so that the processing range is reduced, and the efficiency is improved.

In step S240, a second question in the target text unit is obtained.

Specifically, the target text unit is a translation of the corresponding source text unit, and the target text unit uses a language corresponding to the fourth text. For example, the language used for the fourth text is english, and the language used for the target text unit is also english.

Step S250, a plurality of candidate answers in the fourth text are obtained, and each candidate answer is a combination of continuous answers of which the number is smaller than a preset value;

specifically, the fourth text refers to the solved text corresponding to the target text unit. The fourth text may come from a server, storage device, other interface, or content uploaded, entered by a user, etc., and obtained by other means.

In some embodiments, the fourth text may be obtained by an automatic solution by a solution program, or by obtaining input after a manual solution, or derived from stored pre-solved content. For example, the fourth text may be a sentence containing the second question "traffic data": the process may be a text object whose traffic data is related to the road segment, and the corresponding target text unit may be: the processor may obtain traffic data related to the road segment.

In some embodiments, the processing device may obtain a plurality of candidate answers in the target text unit. The candidate answer is a combination of continuous words with the number smaller than a preset value in the target text unit. The candidate answer may include a word, words, or a combination of words. The preset value refers to an upper limit of the number of words that can be included in a candidate answer. The preset value can be preset by a user or can be set by default by the processing equipment. For example, the preset value may be 2, 3, 4, 5, 6, etc., preferably, the preset value may be 3 or 4. For example, the target text unit is: the processor may obtain traffic data related to the road segment. If the preset value is 3, then the following are all candidate answers: processor, may, processor/may/obtaining, related/related, related/traffic, traffic/data, etc. Taking "processor/available/obtained" as an example, this is a candidate answer consisting of three consecutive words that are available and obtained by the processor. While the following phrases are not candidate answers due to discontinuities: processor/get, related/traffic/data. The candidate answers may be obtained by sequentially obtaining combinations of consecutive words after word segmentation. The grouping can be realized by word segmentation tools such as ansj, jieba, hand and the like, and the description is not limited.

Step S260, obtaining the similarity between the second question and each candidate answer, and determining the similarity as the candidate answer of the second question.

Specifically, based on the second question vector and each candidate answer vector, determining similarity through a similarity model; wherein, the similarity model is a machine learning model.

In some embodiments, the similarity may be determined based on the vector distance of the second question vector from each candidate answer vector. For example, the similarity between the source question vector and each candidate answer vector may be calculated using a cosine method, an euclidean distance method, a mahalanobis distance method, a minkowski distance method, a hamming distance method, or the like.

Step S270, obtaining the similarity between the second question and each candidate answer, and determining the similarity as the candidate answer of the second question.

Specifically, based on the source question vector and each candidate answer vector, determining similarity through a similarity model; wherein, the similarity model is a machine learning model.

Example 3

Referring to fig. 3, fig. 3 is a schematic diagram of an intelligent question answering system according to an embodiment of the present application. The intelligent question-answering system comprises a first obtaining module, a second obtaining module and a question-answering module, wherein the first obtaining module is used for obtaining a first question in a first text; the second obtaining module is used for obtaining a plurality of candidate answers in the second text, wherein each candidate answer is a combination of continuous answers of which the number is smaller than a preset value; and the calculation module is used for determining the candidate answer with the highest similarity as the answer of the first question according to the similarity between the first question and each candidate answer.

Also included is at least one memory for storing computer instructions; at least one processor in communication with the memory, wherein the at least one processor, when executing the computer instructions, causes the system to perform: the device comprises a first acquisition module, a second acquisition module and a calculation module.

The Memory may be, but is not limited to, a Random Access Memory (RAM), a Read Only Memory (ROM), a Programmable Read-Only Memory (PROM), an Erasable Read-Only Memory (EPROM), an electrically Erasable Read-Only Memory (EEPROM), and the like.

In the embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. The apparatus embodiments described above are merely illustrative, and for example, the flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

In addition, functional modules in the embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

The functions, if implemented in the form of software functional modules and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application or portions thereof that substantially contribute to the prior art may be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.

In summary, the embodiment of the present application provides an intelligent question answering method and system,

The above description is only a preferred embodiment of the present application and is not intended to limit the present application, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present application shall be included in the protection scope of the present application.

It will be evident to those skilled in the art that the present application is not limited to the details of the foregoing illustrative embodiments, and that the present application may be embodied in other specific forms without departing from the spirit or essential attributes thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the application being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference sign in a claim should not be construed as limiting the claim concerned.

Claims

1. An intelligent question answering method is characterized by comprising the following steps:

acquiring a first question in a first text;

obtaining a plurality of candidate answers in a second text, wherein each candidate answer is a combination of continuous answers of which the number is smaller than a preset value;

and according to the similarity between the first question and each candidate answer, taking the candidate answer with the highest similarity as the answer of the first question.

2. The intelligent question answering method according to claim 1, wherein the method for obtaining the similarity comprises:

and respectively obtaining a source question vector and a plurality of candidate answer vectors through a text embedding model based on the first question and the candidate answers.

3. The intelligent question answering method according to claim 2, wherein the method for obtaining the similarity further comprises:

determining similarity through a similarity model based on the source question vector and each candidate answer vector; wherein the similarity model is a machine learning model.

4. The method as claimed in claim 1, wherein said step of selecting the candidate answer with the highest similarity as the answer to the first question according to the similarity between the first question and each candidate answer comprises:

and taking one or more candidate answers with the similarity meeting a preset condition as candidate answers, and acquiring user input to determine a first answer.

5. The intelligent question answering method according to claim 1, characterized by comprising:

the first question is looked up in a plurality of source text units, wherein the source text units use a language corresponding to a first text and contain content different from the first text.

6. The intelligent question answering method according to claim 5, characterized by comprising:

regarding each source text unit containing the first question, taking a corresponding target text unit as a third text;

and acquiring a second question in the target text unit.

7. The intelligent question answering method according to claim 6, characterized by comprising:

and acquiring a plurality of candidate answers in a fourth text, wherein each candidate answer is a combination of continuous answers of which the number is smaller than a preset value.

8. The intelligent question answering method according to claim 6, characterized by comprising:

and acquiring the similarity between the second question and each candidate answer, and determining the similarity as the candidate answer of the second question.

9. An intelligent question-answering system, comprising:

the first obtaining module is used for obtaining a first question in the first text;

the second obtaining module is used for obtaining a plurality of candidate answers in a second text, wherein each candidate answer is a combination of continuous answers of which the number is smaller than a preset value;

and the calculation module is used for determining the candidate answer with the highest similarity as the answer of the first question according to the similarity between the first question and each candidate answer.

10. The intelligent question-answering system according to claim 9, comprising:

at least one memory for storing computer instructions;

at least one processor in communication with the memory, wherein the at least one processor, when executing the computer instructions, causes the system to perform: the device comprises a first acquisition module, a second acquisition module and a calculation module.