WO2019041517A1

WO2019041517A1 - Electronic device, question recognition and confirmation method, and computer-readable storage medium

Info

Publication number: WO2019041517A1
Application number: PCT/CN2017/108763
Authority: WO
Inventors: 王健宗; 韩茂琨; 肖京
Original assignee: 平安科技（深圳）有限公司
Priority date: 2017-08-29
Filing date: 2017-10-31
Publication date: 2019-03-07
Also published as: CN107644012A; CN107644012B

Abstract

An electronic device, a question recognition and confirmation method and a-computer-readable storage medium. The method comprises: receiving a question speech pronounced by a user, and carrying out voice recognition on the received question speech to generate a question text (S1); carrying out word segmentation processing on the generated question text according to a predetermined word segmentation rule to obtain segmented words corresponding to the question text (S2); if the obtained segmented words include a predetermined feature word, determining, according to the probability distribution between a feature word and a question, the most likely question corresponding to the predetermined feature word, and determining, according to a predetermined mapping relationship between a question and an answer, an answer corresponding to the most likely question (S3); and feeding the determined answer back to the user (S4). By means of the method, the accuracy of an answer fed back to a user by an intelligent customer service robot and an intelligent customer service response system is improved.

Description

Electronic device, problem identification confirmation method, and computer readable storage medium

The present application is based on the priority of the Chinese Patent Application entitled "Electronic Device, Problem Identification Confirmation Method, and Computer Readable Storage Medium", filed on August 29, 2017, with the application number of CN 201710754550.6, which is filed on August 29, 2017. The entire content is incorporated herein by reference.

Technical field

The present application relates to the field of intelligent voice technologies, and in particular, to an electronic device, a problem recognition confirmation method, and a computer readable storage medium.

Background technique

At present, in order to effectively reduce the waiting conditions of customer service, improve service quality, and improve the convenience of customer service, many service companies (for example, mobile operators, insurance companies, financial institutions, etc.) have adopted offline intelligent customer service robots (for example The intelligent customer service robot set in the physical office area and/or an online intelligent customer service response system (for example, an intelligent voice response system) serves the customer. The existing schemes for such offline intelligent customer service robots and/or online intelligent customer service response systems are: pre-configure mapping relationship data between standard questions and standard answers; when receiving standard questions raised by customers, according to pre-configured The mapping relationship between the standard question and the standard answer determines the standard answer corresponding to the received standard question and feeds back the determined standard answer to the customer. For existing non-standard questions, this existing solution will be difficult to give feedback.

Although there is currently an improvement solution on the market to solve non-standard problems: when the answer corresponding to the non-standard question cannot be found, the non-standard question is calculated similarly to each standard question, and the standard question of the maximum similarity is corresponding. The standard answer is fed back as the answer to the non-standard question. However, since in most cases, the similarity between a non-standard problem and a standard question is caused by words that are not related to the meaning of the statement (for example, "", "?"), therefore, the improvement The accuracy is very low, often making mistakes, causing answers.

Summary of the invention

The main purpose of the present application is to provide a method for identifying and identifying problems, which aims to improve the accuracy of the intelligent customer service system for identifying and confirming non-standard problems, thereby improving the accuracy of feedback answers to non-standard questions.

A first aspect of the present application provides an electronic device including a memory, a processor, and a memory identification confirmation system operable on the processor, where the problem recognition confirmation system is executed by the processor Implement the following steps:

Receiving a question voice sent by the user, performing voice recognition on the received question voice, and generating a question text;

Performing word segmentation on the generated question text according to a predetermined word segmentation rule to obtain a word segment corresponding to the question text;

If the obtained participle contains a predetermined feature word, the problem of determining the maximum probability corresponding to the predetermined feature word is determined according to the probability distribution between the feature word and the question, and according to the mapping between the predetermined question and the answer Relationship, the answer corresponding to the question that determines the maximum probability;

The determined answer is fed back to the user.

A second aspect of the present application provides a method for identifying a problem, the method comprising the steps of:

The determined answer is fed back to the user.

A third aspect of the present application provides a computer readable storage medium storing a problem identification confirmation system, the problem identification confirmation system being executable by at least one processor to cause the at least one processor Perform the following steps:

The determined answer is fed back to the user.

After the user's question speech is recognized as a question text, the technical solution of the present application performs word segmentation on the question text, and obtains a feature word contained in the word segmentation result that reflects the subject or semantic direction of the user question, and is between the feature word and the question. Probabilistic distribution to find the problem with the greatest probability (ie, the most probable problem), and then determine the answer corresponding to the question with the highest probability to feed back to the user; because in this technical solution, the feature word can reflect the subject or semantic direction of the user problem Corresponding to the problem of the maximum probability corresponding to the feature word The answer, therefore, compared to the prior art, the accuracy of the answer to the user is significantly improved in the way of comparing the similarity of the whole problem with the standard problem to obtain the answer corresponding to the most similar question.

DRAWINGS

1 is a schematic flowchart of an embodiment of a method for identifying and confirming a problem of the present application;

2 is a schematic flowchart of an embodiment of a method for identifying and confirming a problem of the present application;

FIG. 3 is a schematic flowchart of determining a probability distribution between a feature word and a question in the problem identification and confirmation method of the present application;

4 is a schematic diagram of an operating environment of a preferred embodiment of the problem identification and confirmation system of the present application;

FIG. 5 is a block diagram of a program of an embodiment of the problem identification confirmation system of the present application; FIG.

FIG. 6 is a program block diagram of a second embodiment of the problem identification and confirmation system of the present application.

Detailed ways

The principles and features of the present application are described in the following with reference to the accompanying drawings, which are only used to explain the present application and are not intended to limit the scope of the application.

The present application proposes a method for identifying and identifying problems, which is mainly used for intelligent customer service products such as an intelligent customer service response system or an intelligent customer service robot.

As shown in FIG. 1 , FIG. 1 is a schematic flowchart of an embodiment of a method for identifying and confirming a problem of the present application.

In this embodiment, the method for identifying and identifying the problem includes:

Step S1: receiving a question voice sent by the user, performing voice recognition on the received question voice, and generating a question text;

When the user asks a question to the intelligent customer service voice system or the intelligent customer service robot, the problem recognition confirmation system receives the question voice issued by the user when questioning, identifies the received question voice, and generates the corresponding question text of the recognized question voice.

Step S2, performing segmentation processing on the generated problem text according to a predetermined word segmentation rule, and obtaining a word segment corresponding to the problem text;

After converting the received question speech recognition into the question text, the question recognition confirming system performs word segmentation processing on the question text according to the predetermined word segmentation rule, and after the word segmentation process, the word segment corresponding to the question text is obtained. In this embodiment, the word segmentation includes words and words. For example, the question text may be “Ping An has launched a product of Zunhong Life”, and after the word segmentation, the result is “Ping An”, “Pushing”, “Yes”, "Zhonghong life", "product", "?".

Step S3, if the obtained participle contains a predetermined feature word, the problem of determining the maximum probability corresponding to the predetermined feature word is determined according to the probability distribution between the feature word and the question, and according to the predetermined question and answer The mapping relationship between the two, determining the answer corresponding to the question of the maximum probability;

The system has preset feature words (for example, “Zhonghong Life”, “Peace”, etc.), the feature words can reflect the subject or semantic direction of the question corresponding to the question text; the system also has predetermined feature words and questions. The probability distribution between each feature word has a probability value corresponding to each pre-existing problem, and the problem text containing each feature word may be the probability of each problem; the system is also set between the preset question and the answer. Mapping relational tables. After obtaining the word segment corresponding to the question text, the system analyzes whether the obtained word segment contains a predetermined feature word; when the analyzed word segment does not contain the predetermined feature word, the user is prompted to re-question or the prompt is unrecognizable. Ask questions and other treatments. When it is analyzed that the obtained participle contains a predetermined feature word, the problem of determining the maximum probability corresponding to the predetermined feature word is determined according to the probability distribution between the feature word and the question, that is, the most probable problem After the problem is determined, the answer corresponding to the question of the maximum probability is obtained according to the mapping relationship between the predetermined question and the answer.

In step S4, the determined answer is fed back to the user.

After obtaining the determined answer, the system feeds the determined answer to the user by means of voice broadcast or sending to the display device for display or sending to the user's preset terminal.

After the user's question speech is recognized as the question text, the technical solution of the present invention performs word segmentation on the question text, and obtains a feature word contained in the word segmentation result that reflects the subject or semantic direction of the user question, and is between the feature word and the question. The probability distribution, in order to find the problem with the greatest probability (ie the most probable problem), and then determine the answer corresponding to the question of the maximum probability to feed back to the user; because in this technical solution, the feature word can reflect the subject or semantic of the user problem Direction, the corresponding answer found by the problem of the maximum probability corresponding to the feature word. Therefore, compared with the prior art, the similarity between the whole problem and the standard question is compared to obtain the answer corresponding to the most similar question. The accuracy of the answers to the feedback users in this case has improved significantly.

Preferably, in this embodiment, the predetermined word segmentation rule is a long word priority word segmentation rule. The long word priority word segmentation rule refers to: for a phrase T1 requiring a word segmentation, starting with the first word A, finding a longest word X1 starting from A from the pre-stored vocabulary, and then culling from T1 X1 has T2 left, and then the same segmentation principle is applied to T2. The result after segmentation is “X1/X2/, ,,,,,”; for example, including “Peace” and “Push” in the pre-stored thesaurus. When "Ye", "Zhonghong Life", "Product", "?", the phrase "Ping An has introduced the product of Zunhong Life" is the result of "Peace" / "Push" / "Y" / "Zun" Macro life" / "product" / "?".

As shown in FIG. 2, FIG. 2 is a schematic flowchart of the second embodiment of the method for identifying and confirming the problem of the present application. The solution of the embodiment is replaced by the following steps on the basis of the first embodiment:

Step S301, if the obtained word segment contains a predetermined feature word, determining, according to the probability distribution between the feature word and the question, determining the predetermined feature word corresponding to each problem The probability;

After analyzing the obtained participle containing the predetermined feature words, according to the probability distribution between the predetermined feature words and the problem in the system, it is determined that the predetermined feature words contained in the obtained word segment respectively correspond to each problem The probability.

Step S302, sorting each problem according to the order of probability from large to small, determining a preset number of questions in the prior order as candidate questions, and providing or broadcasting the determined candidate questions to the user for selection;

After the predetermined feature words contained in the obtained word segment respectively correspond to the probabilities of the respective questions, the problems are sorted in descending order according to the obtained probabilities, and then the pre-ordered number in the sorted problem sequence is extracted. The questions (for example, 3 or 4) are used as candidate questions, and the extracted candidate questions are fed back to the user for selection by the user. The manner in which the candidate question is fed back to the user may be: 1. voice broadcast; 2. providing a selection interface, and the candidate question is displayed on the selection interface (for example, generating a question selection interface for the user to select, the selection interface may include a candidate question list, Each candidate question in the list corresponds to an "OK" button, the user can click the button to select a corresponding question);

Step S303, after the user selects a question, the answer corresponding to the question is determined according to a mapping relationship between the predetermined question and the answer.

After the user makes a selection based on the candidate question of the system feedback, the system receives the question selected by the user, and determines the answer corresponding to the question selected by the user according to the mapping relationship between the predetermined question and the answer in the system. .

As shown in FIG. 3, the probability distribution between the feature words and the problem is determined as follows:

Step S51, adding a preset number of implicit topics between the feature words and the question;

First, an implicit number of predicted numbers (for example, 50) is added between the feature words and the problem layer as an intermediate layer to constitute a problem selection model; wherein the implicit theme is virtual and there is no real Meaning; each implied topic usually contains multiple feature words, each of which usually contains multiple implied topics.

Step S52: acquiring a problem text to be trained, and performing word segmentation processing on the obtained problem texts respectively, to obtain word segments corresponding to the respective problem texts;

After the problem selection model is formed, the problem text to be trained is obtained (the problem text is prepared in advance), and the obtained problem texts are separately subjected to word segmentation processing, thereby obtaining the word segmentation results corresponding to the respective problem texts.

Step S53: determining, according to a predetermined mapping relationship between the implicit subject and the feature word, a first quantity of the feature words included in each hidden topic, and respectively determining a second quantity of the hidden topic to which each feature word belongs, according to Corresponding first quantity and second quantity determining a first selection probability of each feature word for each implicit topic;

Determining, according to a predetermined mapping relationship between the implicit subject and the feature words in the system, the first quantity of the feature words contained in each hidden topic and the second quantity of the hidden topic to which each feature word belongs, and then according to Corresponding first quantity and second quantity respectively obtain a first selection probability of each feature word for each implicit topic; for example, the second quantity of the implicit topic to which the feature word Y belongs is X2, and the feature of an implicit topic The first number of words is X1, and the probability of selection of the feature word Y for the hidden subject is: 1/(X1*X2).

Step S54: Determine a third quantity of the hidden topic included in each question text according to a predetermined mapping relationship between the implicit topic and the question text, and determine a fourth quantity of the problem text to which each hidden topic belongs, according to Corresponding third and fourth quantities determine a second selection probability of each implicit topic for each question text;

According to the mapping relationship between the predetermined implicit topic and the problem text in the system, respectively determine the third quantity of the implicit topic contained in each question text and the fourth quantity of the problem text to which each implicit topic belongs, and then according to the corresponding The third quantity and the fourth quantity respectively obtain a second selection probability of each implicit topic for each question text; for example, the fourth quantity of the problem text to which the implicit theme K belongs is J2, and an implicit subject contained in a question text The third quantity is J1, then the probability of selection of the implicit subject K for the question text is: 1/(J1*J2). In this embodiment, the sequence of the step S54 and the step S53 can be changed.

Step S55: Substituting the corresponding first selection probability and the second selection probability into a predetermined calculation formula to calculate, calculating a third selection probability of each feature word for each problem text, and calculating each feature word separately for each problem The third selection probability of the text is the probability distribution between the feature words and the problem.

According to the first selection probability distribution of the feature word to the implicit topic, and the second selection probability distribution of the implicit topic to the question text, the third selection probability distribution of the feature word to the problem text can be further obtained. Specifically, by substituting the corresponding first selection probability and the second selection probability into a predetermined calculation formula, the selection probability of each feature word for each problem text is obtained, that is, the probability distribution between the feature word and the problem is obtained. In this embodiment, the predetermined calculation formula is: P3=P1*P2, where P1 represents a first selection probability, P2 represents a second selection probability, and P3 represents a third selection probability. For example, the first selection probability of the feature word Y for the implicit topic K is: 1/(X1*X2), and the second selection probability of the implicit topic K for the question text W is: 1/(J1*J2), then the feature The probability of selection of the word Y for the question text W is 1/(X1*X2)*(J1*J2).

The present application also proposes a problem identification confirmation system.

Please refer to FIG. 4 , which is a schematic diagram of an operating environment of a preferred embodiment of the problem identification and confirmation system 10 of the present application.

In the present embodiment, the problem recognition confirmation system 10 is installed and operated in the electronic device 1. The electronic device 1 may be a computing device such as a desktop computer, a notebook, a palmtop computer, and a server. The electronic device 1 may include, but is not limited to, a memory 11, a processor 12 and display 13. Figure 4 shows only the electronic device 1 with components 11-13, but it should be understood that not all illustrated components may be implemented, and more or fewer components may be implemented instead.

The memory 11 is a computer storage medium, which in some embodiments may be an internal storage unit of the electronic device 1, such as a hard disk or memory of the electronic device 1. The memory 11 may also be an external storage device of the electronic device 1 in other embodiments, such as a plug-in hard disk equipped on the electronic device 1, a smart memory card (SMC), and a secure digital (SD). Card, flash card, etc. Further, the memory 11 may also include both an internal storage unit of the electronic device 1 and an external storage device. The memory 11 is used to store application software and various types of data installed in the electronic device 1, such as program codes of the problem recognition confirmation system 10. The memory 11 can also be used to temporarily store data that has been output or is about to be output.

The processor 12, in some embodiments, may be a Central Processing Unit (CPU), microprocessor or other data processing chip for running program code or processing data stored in the memory 11, such as performing problem identification confirmation. System 10 and so on.

The display 13 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch sensor, or the like in some embodiments. The display 13 is for displaying information processed in the electronic device 1 and a user interface for displaying visualization, such as a business customization interface or the like. The components 11-13 of the electronic device 1 communicate with one another via a system bus.

Please refer to FIG. 5, which is a program module diagram of an embodiment of the problem identification confirmation system 10 of the present application. In the present embodiment, the problem identification confirmation system 10 can be divided into one or more modules, one or more modules are stored in the memory 11, and by one or more processors (the processor 12 in this embodiment) Executed to complete the application. For example, in FIG. 5, the problem identification confirmation system 10 can be divided into an identification module 101, a word segmentation module 102, a determination module 103, and a feedback module 104. A module referred to in this application refers to a series of computer program instruction segments capable of performing a specific function, and is more suitable than the program for describing the execution process of the problem recognition confirmation system 10 in the electronic device 1, wherein:

The identification module 101 is configured to receive a question voice sent by the user, perform voice recognition on the received question voice, and generate a question text.

The word segmentation module 102 is configured to perform word segmentation processing on the generated problem text according to a predetermined word segmentation rule to obtain a word segment corresponding to the problem text;

After converting the received speech recognition into a question text, the problem recognition confirmation system performs word segmentation on the problem text according to a predetermined word segmentation rule, through the word segmentation After the rule, the word segment corresponding to the question text is obtained. In this embodiment, the word segmentation includes words and words. For example, the question text may be “Ping An has launched a product of Zunhong Life”, and after the word segmentation, the result is “Ping An”, “Pushing”, “Yes”, "Zhonghong life", "product", "?".

a determining module 103, configured to determine a maximum probability corresponding to the predetermined feature word according to a probability distribution between the feature word and the question, after the obtained word segment includes the predetermined feature word, and according to the predetermined problem The mapping relationship with the answer, determining the answer corresponding to the question of the maximum probability;

The system has preset feature words (for example, “Zhonghong Life”, “Peace”, etc.), the feature words can reflect the subject or semantic direction of the question corresponding to the question text; the system also has predetermined feature words and questions. The probability distribution between each feature word has a probability value corresponding to each pre-existing problem, and the problem text containing each feature word may be the probability of each problem; the system is also set between the preset question and the answer. Mapping relational tables. After obtaining the word segment corresponding to the question text, the system analyzes whether the obtained word segment contains a predetermined feature word; when the analyzed word segment contains the predetermined feature word, the probability between the feature word and the question is determined. Distribution, the problem of determining the maximum probability corresponding to the predetermined feature word, that is, the most probable problem, after the problem is determined, the problem of the maximum probability is obtained according to the mapping relationship between the predetermined question and the answer. The corresponding answer. In addition, the determining module 103 also prompts the user to re-question or prompts that the proposed question cannot be recognized after the obtained participle does not contain the predetermined feature word.

The feedback module 104 is configured to feed back the determined answer to the user.

Preferably, in this embodiment, the predetermined word segmentation rule is a long word priority word segmentation rule. The long word priority word segmentation rule refers to: for a phrase T1 requiring a word segmentation, starting with the first word A, finding a longest word X1 starting from A from the pre-stored vocabulary, and then culling from T1 X1 has T2 left, and then the same segmentation principle is applied to T2. The result after segmentation is "X1/X2/, ,,,,,"; for example, included in the pre-stored thesaurus When “Peace”, “Publish”, “Yes”, “Zhonghong Life”, “Product”, “?”, the phrase “Ping An Hung Hsing Life Product” is the result of “Peace”/“Push” / "了" / "尊宏人生" / "product" / "?".

As shown in FIG. 6, FIG. 6 is a program module diagram of a second embodiment of the problem identification and confirmation system of the present application. The solution of the embodiment is replaced by the following module on the basis of the first embodiment:

The first determining sub-module 105 is configured to determine a probability that the predetermined feature word corresponds to each question according to a probability distribution between the feature word and the question, after the obtained word segment includes the predetermined feature word;

The second determining sub-module 106 is configured to sort each problem according to a descending order of probabilities, determine a preset number of questions in the prior order as candidate questions, and provide or broadcast the determined candidate questions to the candidate questions. The user makes a selection;

The third determining sub-module 107 is configured to determine an answer corresponding to the question according to a mapping relationship between the predetermined question and the answer after the user selects a question.

Preferably, in this embodiment, the probability distribution between the feature words and the problem is determined according to the following steps:

1. Add a preset number of implied topics between feature words and questions;

2. Obtaining the problem text to be trained, and separately performing word segmentation on the obtained problem text, and obtaining the word segment corresponding to each question text;

3. According to the mapping relationship between the predetermined implicit topic and the feature word, respectively determine the first quantity of the feature words contained in each hidden topic, and determine the second quantity of the hidden topic to which each feature word belongs, according to the corresponding The first quantity and the second quantity determine a first selection probability of each feature word for each implicit topic;

4. According to the mapping relationship between the predetermined implicit topic and the problem text, respectively determine the third quantity of the implicit topic contained in each question text, and determine the fourth quantity of the problem text to which each hidden topic belongs, according to the corresponding The third quantity and the fourth quantity determine a second selection probability of each implicit topic for each question text;

According to the mapping relationship between the predetermined implicit topic and the problem text in the system, respectively determine the third quantity of the implicit topic contained in each question text and the fourth quantity of the problem text to which each implicit topic belongs, and then according to the corresponding The third quantity and the fourth quantity respectively obtain a second selection probability of each implicit topic for each question text; for example, the fourth quantity of the problem text to which the implicit theme K belongs is J2, and an implicit subject contained in a question text The third quantity is J1, then the probability of selection of the implicit subject K for the question text is: 1/(J1*J2).

5. Substituting the corresponding first selection probability and the second selection probability into a predetermined calculation formula to calculate, and calculating a third selection probability of each feature word for each problem text, and calculating each feature word separately for each problem text The third selection probability is the probability distribution between the feature word and the problem.

According to the first selection probability distribution of the feature word to the implicit topic, and the second selection probability distribution of the implicit topic to the question text, the third selection probability distribution of the feature word to the problem text can be further obtained. Specifically, by substituting the corresponding first selection probability and the second selection probability into a predetermined calculation formula, the selection probability of each feature word for each problem text is obtained, that is, the probability distribution between the feature word and the problem is obtained. In this embodiment, the predetermined calculation formula is: P3=P1*P2, where P1 represents a first selection probability, P2 represents a second selection probability, and P3 represents a third selection probability. For example, the first selection probability of the feature word Y for the implicit topic K is: 1/(X1*X2), and the implicit topic K is asked. The second selection probability of the question text W is: 1/(J1*J2), then the selection probability of the feature word Y for the question text W is 1/(X1*X2)*(J1*J2).

The present application also provides a computer readable storage medium storing a problem identification confirmation system, the problem identification confirmation system being executable by at least one processor to cause the at least one processor to perform the above-described The problem identification confirmation method described in an embodiment.

The above description is only a preferred embodiment of the present invention, and is not intended to limit the scope of the invention, and the equivalent structural transformation, or direct/indirect use, of the present invention and the contents of the drawings are used in the inventive concept of the present invention. It is included in the scope of the patent protection of the present invention in other related technical fields.

Claims

An electronic device, comprising: a memory, a processor, wherein the memory stores a problem identification confirmation system operable on the processor, the problem identification confirmation system being the processor The following steps are implemented during execution:

S1: receiving a question voice sent by the user, performing voice recognition on the received question voice, and generating a question text;

S2. Performing word segmentation on the generated problem text according to a predetermined word segmentation rule to obtain a word segment corresponding to the question text;

S3. If the obtained participle contains a predetermined feature word, the problem of determining the maximum probability corresponding to the predetermined feature word is determined according to the probability distribution between the feature word and the question, and according to the predetermined question and answer Mapping relationship, determining the answer corresponding to the question of the maximum probability;

S4. The determined answer is fed back to the user.
The electronic device according to claim 1, wherein the probability distribution between the feature word and the question is determined as follows:

Add a preset number of implied topics between feature words and questions;

Obtaining the problem text to be trained, and separately performing word segmentation on the obtained problem text, and obtaining the word segment corresponding to each question text;

Determining, according to a predetermined mapping relationship between the implicit subject and the feature word, a first quantity of the feature words contained in each hidden topic, and respectively determining a second quantity of the hidden subject to which each feature word belongs, according to the corresponding number A quantity and a second quantity determine a first selection probability of each feature word for each implicit topic;

Determining, according to a predetermined mapping relationship between the implicit topic and the problem text, respectively determining a third quantity of the implicit topic contained in each question text, and respectively determining a fourth quantity of the problem text to which each implicit topic belongs, according to the corresponding number The third quantity and the fourth quantity determine a second selection probability of each implicit topic for each question text;

Substituting the corresponding first selection probability and the second selection probability into a predetermined calculation formula to calculate, calculating a third selection probability of each feature word for each problem text, and calculating each feature word separately for each question text The three-choice probability is the probability distribution between the feature word and the problem.
The electronic device according to claim 2, wherein said predetermined calculation formula is:

P3 = P1 * P2, where P1 represents the first selection probability, P2 represents the second selection probability, and P3 represents the third selection probability.
The electronic device according to claim 1, wherein said step S3 is replaced Change to the following steps:

If the obtained participle contains a predetermined feature word, determining a probability that the predetermined feature word corresponds to each question according to a probability distribution between the feature word and the question;

Sorting each problem according to the order of probability, determining the pre-sorted number of questions as candidate questions, and providing or broadcasting the determined candidate questions to the user for selection;

After the user selects a question, the answer corresponding to the question is determined according to the mapping relationship between the predetermined question and the answer.
The electronic device according to claim 4, wherein the probability distribution between the feature word and the question is determined as follows:

Add a preset number of implied topics between feature words and questions;

Obtaining the problem text to be trained, and separately performing word segmentation on the obtained problem text, and obtaining the word segment corresponding to each question text;

Determining, according to a predetermined mapping relationship between the implicit subject and the feature word, a first quantity of the feature words contained in each hidden topic, and respectively determining a second quantity of the hidden subject to which each feature word belongs, according to the corresponding number A quantity and a second quantity determine a first selection probability of each feature word for each implicit topic;

Determining, according to a predetermined mapping relationship between the implicit topic and the problem text, respectively determining a third quantity of the implicit topic contained in each question text, and respectively determining a fourth quantity of the problem text to which each implicit topic belongs, according to the corresponding number The third quantity and the fourth quantity determine a second selection probability of each implicit topic for each question text;

Substituting the corresponding first selection probability and the second selection probability into a predetermined calculation formula to calculate, calculating a third selection probability of each feature word for each problem text, and calculating each feature word separately for each question text The three-choice probability is the probability distribution between the feature word and the problem.
The electronic device according to claim 5, wherein said predetermined calculation formula is:

P3 = P1 * P2, where P1 represents the first selection probability, P2 represents the second selection probability, and P3 represents the third selection probability.
The electronic device according to claim 1, wherein said predetermined word segmentation rule is a long word priority word segmentation rule.
The electronic device according to claim 7, wherein the probability distribution between the feature word and the question is determined as follows:

Add a preset number of implied topics between feature words and questions;

Obtaining the problem text to be trained, and separately performing word segmentation on the obtained problem text, and obtaining the word segment corresponding to each question text;

Determining, according to a predetermined mapping relationship between the implicit subject and the feature word, a first quantity of the feature words contained in each hidden topic, and respectively determining a second quantity of the hidden subject to which each feature word belongs, according to the corresponding number A quantity and a second quantity determine a first selection probability of each feature word for each implicit topic;

Determining, according to a predetermined mapping relationship between the implicit topic and the problem text, respectively determining a third quantity of the implicit topic contained in each question text, and respectively determining a fourth quantity of the problem text to which each implicit topic belongs, according to the corresponding number The third quantity and the fourth quantity determine a second selection probability of each implicit topic for each question text;

Substituting the corresponding first selection probability and the second selection probability into a predetermined calculation formula to calculate, calculating a third selection probability of each feature word for each problem text, and calculating each feature word separately for each question text The three-choice probability is the probability distribution between the feature word and the problem.
The electronic device according to claim 8, wherein said predetermined calculation formula is:

P3 = P1 * P2, where P1 represents the first selection probability, P2 represents the second selection probability, and P3 represents the third selection probability.
A method for identifying and identifying a problem, characterized in that the method comprises the steps of:

A1. Receiving a question voice sent by the user, performing voice recognition on the received question voice, and generating a question text;

A2. Performing word segmentation on the generated problem text according to a predetermined word segmentation rule to obtain a word segment corresponding to the question text;

A3. If the obtained participle contains a predetermined feature word, the problem of determining the maximum probability corresponding to the predetermined feature word is determined according to the probability distribution between the feature word and the question, and according to the predetermined question and answer Mapping relationship, determining the answer corresponding to the question of the maximum probability;

A4. The determined answer is fed back to the user.
The problem recognition confirmation method according to claim 10, wherein the probability distribution between the feature word and the question is determined as follows:

Add a preset number of implied topics between feature words and questions;

Obtaining the problem text to be trained, and separately performing word segmentation on the obtained problem text, and obtaining the word segment corresponding to each question text;

Determining the first quantity of the feature words contained in each hidden topic according to the predetermined mapping relationship between the implicit topic and the feature words, respectively determining the implicit master to which each feature word belongs a second quantity of the problem, determining, according to the corresponding first quantity and the second quantity, a first selection probability of each feature word for each implied topic;

Determining, according to a predetermined mapping relationship between the implicit topic and the problem text, respectively determining a third quantity of the implicit topic contained in each question text, and respectively determining a fourth quantity of the problem text to which each implicit topic belongs, according to the corresponding number The third quantity and the fourth quantity determine a second selection probability of each implicit topic for each question text;

Substituting the corresponding first selection probability and the second selection probability into a predetermined calculation formula to calculate, calculating a third selection probability of each feature word for each problem text, and calculating each feature word separately for each question text The three-choice probability is the probability distribution between the feature word and the problem.
The problem recognition confirmation method according to claim 11, wherein said predetermined calculation formula is:

P3 = P1 * P2, where P1 represents the first selection probability, P2 represents the second selection probability, and P3 represents the third selection probability.
The problem identification confirmation method according to claim 10, wherein the step A3 is replaced by the following steps:

If the obtained participle contains a predetermined feature word, determining a probability that the predetermined feature word corresponds to each question according to a probability distribution between the feature word and the question;

Sorting each problem according to the order of probability, determining the pre-sorted number of questions as candidate questions, and providing or broadcasting the determined candidate questions to the user for selection;

After the user selects a question, the answer corresponding to the question is determined according to the mapping relationship between the predetermined question and the answer.
The problem recognition confirmation method according to claim 13, wherein the probability distribution between the feature word and the question is determined as follows:

Add a preset number of implied topics between feature words and questions;

Obtaining the problem text to be trained, and separately performing word segmentation on the obtained problem text, and obtaining the word segment corresponding to each question text;

Determining, according to a predetermined mapping relationship between the implicit subject and the feature word, a first quantity of the feature words contained in each hidden topic, and respectively determining a second quantity of the hidden subject to which each feature word belongs, according to the corresponding number A quantity and a second quantity determine a first selection probability of each feature word for each implicit topic;

Determining, according to a predetermined mapping relationship between the implicit topic and the problem text, respectively determining a third quantity of the implicit topic contained in each question text, and respectively determining a fourth quantity of the problem text to which each implicit topic belongs, according to the corresponding number Three quantities and a fourth quantity determine each implied The second selection probability of the subject's text for each question;

Substituting the corresponding first selection probability and the second selection probability into a predetermined calculation formula to calculate, calculating a third selection probability of each feature word for each problem text, and calculating each feature word separately for each question text The three-choice probability is the probability distribution between the feature word and the problem.
The problem recognition confirmation method according to claim 14, wherein said predetermined calculation formula is:

P3 = P1 * P2, where P1 represents the first selection probability, P2 represents the second selection probability, and P3 represents the third selection probability.
The problem recognition confirmation method according to claim 10, wherein said predetermined word segmentation rule is a long word priority word segmentation rule.
The problem recognition confirmation method according to claim 16, wherein the probability distribution between the feature word and the question is determined as follows:

Add a preset number of implied topics between feature words and questions;

Obtaining the problem text to be trained, and separately performing word segmentation on the obtained problem text, and obtaining the word segment corresponding to each question text;

Determining, according to a predetermined mapping relationship between the implicit subject and the feature word, a first quantity of the feature words contained in each hidden topic, and respectively determining a second quantity of the hidden subject to which each feature word belongs, according to the corresponding number A quantity and a second quantity determine a first selection probability of each feature word for each implicit topic;

Determining, according to a predetermined mapping relationship between the implicit topic and the problem text, respectively determining a third quantity of the implicit topic contained in each question text, and respectively determining a fourth quantity of the problem text to which each implicit topic belongs, according to the corresponding number The third quantity and the fourth quantity determine a second selection probability of each implicit topic for each question text;

Substituting the corresponding first selection probability and the second selection probability into a predetermined calculation formula to calculate, calculating a third selection probability of each feature word for each problem text, and calculating each feature word separately for each question text The three-choice probability is the probability distribution between the feature word and the problem.
The problem recognition confirmation method according to claim 17, wherein said predetermined calculation formula is:

P3 = P1 * P2, where P1 represents the first selection probability, P2 represents the second selection probability, and P3 represents the third selection probability.
A computer readable storage medium, wherein the computer readable storage The storage medium stores a problem identification confirmation system that is executable by at least one processor to cause the at least one processor to perform the following steps:

B1. Receiving a question voice sent by the user, performing voice recognition on the received question voice, and generating a question text;

B2. Performing word segmentation on the generated problem text according to a predetermined word segmentation rule to obtain a word segment corresponding to the question text;

B3. If the obtained participle contains a predetermined feature word, the problem of determining the maximum probability corresponding to the predetermined feature word is determined according to the probability distribution between the feature word and the question, and according to the predetermined question and answer Mapping relationship, determining the answer corresponding to the question of the maximum probability;

B4. The determined answer is fed back to the user.
The computer readable storage medium of claim 19, wherein said step B3 is replaced by the following steps:

If the obtained participle contains a predetermined feature word, determining a probability that the predetermined feature word corresponds to each question according to a probability distribution between the feature word and the question;

Sorting each problem according to the order of probability, determining the pre-sorted number of questions as candidate questions, and providing or broadcasting the determined candidate questions to the user for selection;

After the user selects a question, the answer corresponding to the question is determined according to the mapping relationship between the predetermined question and the answer.