Answer retrieval method and device of FAQ system, electronic equipment and storage medium

Abstract

The application provides an answer retrieval method and device of an FAQ system, an electronic device and a computer readable storage medium. The method comprises the following steps: determining a plurality of standard question groups serving as specified standard question groups and a first group score of each specified standard question group based on a first similarity between a user question and each question in an FAQ corpus; determining a second group score corresponding to each specified standard question group based on a second similarity between a standard question in each specified standard question group and the user question; performing weighted summation on the first group score and the second group score of each specified standard question group to obtain an integrated score of each specified standard question group; and determining a standard answer corresponding to a specified standard question group with the highest integrated score as an answer to the user question. According to the application, the integrated score determined based on the first group score and the second group score can take into account the breadth and precision of the retrieval, so that the answer to the user question can be obtained more accurately.

Description

Technical Field

[0001] This application relates to the field of natural language processing technology, and in particular to a method and apparatus for answer retrieval in a FAQ system, an electronic device, and a computer-readable storage medium. Background Technology

[0002] A FAQ (Frequently Asked Questions) system can provide definitive answers to user-submitted questions. Building an FAQ system begins with creating a FAQ corpus. The FAQ corpus consists of three parts: standard questions, standard answers, and similar questions. Standard questions are formal and standardized in language. The FAQ corpus can include multiple standard questions with high differentiation between them, covering a variety of different questions users might ask. Standard answers are responses to standard questions. Generally, there is a one-to-one correspondence between standard questions and standard answers, excluding complexities such as dimensionality. For example, the standard question "Is there an expiration date after the STAR Market is opened?" corresponds to the standard answer "Hello, once the STAR Market is successfully opened, it will remain valid indefinitely." Similar questions are extensions of standard questions. One standard question can correspond to multiple similar questions. Similar questions express the same meaning as their corresponding standard questions, but the language of similar questions is often less standardized and tries to cover various ways of asking the standard question. For example, the standard question is "Does the Science and Technology Innovation Board have an expiration date after it is opened?" Similar questions include "How long does it take for the Science and Technology Innovation Board to expire?" and "How long can the Science and Technology Innovation Board be used?"

[0003] See Figure 1 This is a schematic diagram illustrating the organizational structure of a FAQ corpus provided in an embodiment of this application, as shown below. Figure 1 As shown, each standard question maps to a standard answer, and each standard question corresponds to multiple similar questions.

[0004] FAQ retrieval and question-answering systems are a type of FAQ system. After receiving a user's question, the FAQ retrieval and question-answering system performs similarity matching between the user's question and standard questions and similar questions in the FAQ corpus. When the most similar question is a standard question, the standard answer mapped to the standard question is used as the system's response; when the most similar question is a similar question, the standard answer mapped to the standard question of the similar question is used as the system's response.

[0005] However, standard questions have formal language and high discriminative power, making standard question retrieval precise. When the user's question is relatively common, the probability of finding a standard question incorrectly is lower. Similar questions are numerous and cover a wide range of question types, giving similar question retrieval breadth. When the user's question is relatively rare, finding similar questions that address the user's question can provide a correct response. The response methods of related solutions do not effectively distinguish the different characteristics of standard and similar questions. Summary of the Invention

[0006] The purpose of this application is to provide a method, apparatus, electronic device, and computer-readable storage medium for retrieving answers to FAQ systems, which can determine the optimal question answer while taking into account the different characteristics of standard questions and similar questions.

[0007] On the one hand, this application provides a method for retrieving answers to an FAQ system, including:

[0008] Based on the first similarity between the user's question and each question in the FAQ corpus, several standard question groups are determined as designated standard question groups, and a first group score is determined for each designated standard question group; wherein, the questions in the FAQ corpus include standard questions and similar questions, and the standard question group includes one standard question and multiple similar questions;

[0009] Based on the second similarity between the standard question and the user question in each specified standard question group, the second group score corresponding to each specified standard question group is determined;

[0010] The scores of the first group and the second group for each specified standard question group are weighted and summed to obtain the integrated score for each specified standard question group;

[0011] The standard answer corresponding to the specified standard question group with the highest integration score is determined as the answer to the user question.

[0012] In one embodiment, determining several standard question groups as designated standard question groups, and a first group score for each designated standard question group, based on the first similarity between the user's question and each question in the FAQ corpus, includes:

[0013] Based on the first similarity between the user's question and each question, a first similarity score is determined for each question relative to the user's question;

[0014] Based on the first similarity score of all questions, determine several specified standard question groups;

[0015] Based on the first similarity score corresponding to each question within each specified standard question group, determine the first group score corresponding to each specified standard question group.

[0016] In one embodiment, determining a plurality of specified standard question groups based on the first similarity score of all questions includes:

[0017] Select a specified number of questions with the highest first similarity score as the designated questions;

[0018] Determine several standard question groups to which the specified question belongs, and use them as the specified standard question groups.

[0019] In one embodiment, determining the first group score corresponding to each specified standard question group based on the first similarity score corresponding to each question within each specified standard question group includes:

[0020] For each specified standard question group, the maximum, average, or median value of multiple first similarity scores in the specified standard question group is determined as the first group score of the specified standard question group.

[0021] In one embodiment, before determining a plurality of standard question groups as designated standard question groups, and a first group score for each designated standard question group, the method further includes:

[0022] For each question in the FAQ corpus, a first similarity is determined between the question and the user's question according to a first similarity algorithm.

[0023] In one embodiment, determining the second group score corresponding to each specified standard question group based on the second similarity between the standard question and the user question in each specified standard question group includes:

[0024] Based on the second similarity between the standard question and the user question in each specified standard question group, a second similarity score is determined for each standard question relative to the user question;

[0025] For each specified standard question group, the second similarity score of the standard questions within the specified standard question group is used as the second group score of the specified standard question group.

[0026] In one embodiment, before determining the second group score corresponding to each specified criterion group, the method further includes:

[0027] For each standard question in a group of specified standard questions, a second similarity between the standard question and the user question is determined according to a second similarity algorithm.

[0028] On the other hand, this application also provides an answer retrieval device for an FAQ system, comprising:

[0029] The first determining module is used to determine several standard question groups as specified standard question groups and a first group score for each specified standard question group based on the first similarity between the user's question and each question in the FAQ corpus; wherein, the questions in the FAQ corpus include standard questions and similar questions, and the standard question group includes one standard question and multiple similar questions;

[0030] The second determining module is used to determine the second group score corresponding to each specified standard question group based on the second similarity between the standard question and the user question in each specified standard question group;

[0031] The synthesis module is used to obtain a weighted sum of the first group score and the second group score for each specified standard question group, so as to obtain the integrated score for each specified standard question group.

[0032] The third determining module is used to determine the standard answer corresponding to the specified standard question group with the highest integrated score, which is the answer to the user question.

[0033] Furthermore, this application also provides an electronic device, the electronic device comprising:

[0034] processor;

[0035] Memory used to store processor-executable instructions;

[0036] The processor is configured to execute the answer retrieval method of the FAQ system described above.

[0037] In addition, this application also provides a computer-readable storage medium storing a computer program that can be executed by a processor to complete the answer retrieval method of the above-described FAQ system.

[0038] This application's solution uses the first similarity between the user's question and all questions in the FAQ corpus to filter out several designated standard question groups that are relatively similar to the user's question, and determines the first group score of each designated standard question group relative to the user's question; based on the second similarity between the standard questions and the user's question in each designated standard question group, the second group score corresponding to each designated standard question group can be determined; after weighted summation of the first group score and the second group score, the integrated score of each designated standard question can be obtained, and the standard answer corresponding to the designated standard question with the highest integrated score is taken as the answer to the user's question;

[0039] In the process of retrieving answers, firstly, a group of standard questions that are similar to the user's question is determined from the scope of all questions including standard questions and similar questions. Then, a second similarity score between the standard questions and the user's question is determined from the group of standard questions, which is used as the second group score. The integrated score determined by the first group score and the second group score can take into account both the breadth and precision characteristics of the retrieval, so as to obtain the answer to the user's question more accurately. Attached Figure Description

[0040] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly described below.

[0041] Figure 1 A schematic diagram illustrating the organization of a FAQ corpus provided in an embodiment of this application;

[0042] Figure 2This is a schematic diagram illustrating an application scenario of the answer retrieval method for the FAQ system provided in an embodiment of this application;

[0043] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application;

[0044] Figure 4 A flowchart illustrating the answer retrieval method of a FAQ system provided in an embodiment of this application;

[0045] Figure 5 A detailed flowchart of step 410 provided in one embodiment of this application;

[0046] Figure 6 A detailed flowchart of step 420 provided in one embodiment of this application;

[0047] Figure 7 A schematic diagram illustrating the determination of an integration score according to an embodiment of this application;

[0048] Figure 8 A block diagram of an answer retrieval device for a FAQ system provided in an embodiment of this application. Detailed Implementation

[0049] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings.

[0050] Similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0051] Figure 2 This is a schematic diagram illustrating an application scenario of the answer retrieval method for the FAQ system provided in this application embodiment. For example... Figure 2 As shown, the application scenario includes a client 20 and a server 30. The client 20 can be a user terminal such as a host, mobile phone, or tablet computer, used to send user questions to the server 30. The server 30 can be a server, server cluster, or cloud computing center equipped with an FAQ system, which can detect the answer corresponding to the user question in the FAQ system and return it to the client 20.

[0052] like Figure 3 As shown, this embodiment provides an electronic device 1, including: at least one processor 11 and a memory 12. Figure 3Taking a processor 11 as an example, the processor 11 and the memory 12 are connected via a bus 10. The memory 12 stores instructions that can be executed by the processor 11. The instructions are executed by the processor 11 to enable the electronic device 1 to perform all or part of the process of the method in the following embodiments. In one embodiment, the electronic device 1 can be the server 30 described above, used to execute the answer retrieval method of the FAQ system.

[0053] The memory 12 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable red-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0054] This application also provides a computer-readable storage medium storing a computer program that can be executed by a processor 11 to perform the answer retrieval method of the FAQ system provided in this application.

[0055] See Figure 4 This is a flowchart illustrating the answer retrieval method of a FAQ system provided in an embodiment of this application, as shown below. Figure 4 As shown, the method may include steps 410-440.

[0056] Step 410: Based on the first similarity between the user's question and each question in the FAQ corpus, determine several standard question groups as designated standard question groups, and the first group score of each designated standard question group; wherein, the questions in the FAQ corpus include standard questions and similar questions, and the standard question group includes one standard question and multiple similar questions.

[0057] User-generated questions are questions that the server, which runs a FAQ system, needs to answer. The server can receive user-generated questions from the user terminals it connects to. User-generated questions may use formal and standardized language, or they may use non-standard language.

[0058] A standard question group consists of one standard question and multiple similar questions. The multiple similar questions within the standard question group correspond to the standard question and express the same meaning. Figure 1For example, standard question 1 and similar questions 1.1, 1.2, 1.3, 1.4, and 1.5 constitute a standard question group; standard question 2 and similar questions 2.1, 2.2, 2.3, and 2.4 constitute a standard question group; standard question 3 and similar questions 3.1 and 3.2 constitute a standard question group.

[0059] The first similarity score, determined by a similarity algorithm, represents the similarity between the user's question text and the text of a standard or similar question. The first group score represents the overall similarity between the user's question text and the text of questions within the standard question group.

[0060] The server can filter out several standard question groups that are sufficiently similar to the user's question based on the similarity between the user's question and each question in the FAQ corpus, and use them as designated standard question groups. The server can also determine the first group score corresponding to the designated standard question group.

[0061] Step 420: Based on the second similarity between the standard question and the user question in each specified standard question group, determine the second group score corresponding to each specified standard question group.

[0062] The second similarity score, determined by a similarity algorithm, represents the similarity between the user's question text and the text of the standard questions in each specified standard question group. The second group score represents the overall similarity between the user's question text and the text of the questions within the standard question group.

[0063] In one embodiment, for each standard question in a specified standard question group, the first similarity corresponding to that standard question can be used as the second similarity, thereby reducing the amount of computation and improving retrieval efficiency.

[0064] The score for the second group is only related to the standard question of the specified standard question group. Therefore, in general, the scores for the first group and the second group of the same specified standard question group will be different.

[0065] Step 430: For each specified standard question group, the first group score and the second group score are weighted and summed to obtain the integrated score of each specified standard question group.

[0066] For each specified standard question group, the server can perform a weighted sum of the scores of the first group and the second group based on pre-configured weight coefficients to obtain the integrated score for that specified standard question group. This integrated score represents the overall similarity between the text of the user's question and the text of the questions within the specified standard question group.

[0067] Step 440: Determine the standard answer corresponding to the specified standard question group with the highest integration score, which is the answer to the user's question.

[0068] After determining the integration score for each specified standard question group, the server can identify the highest integration score from multiple integration scores and use the standard answer corresponding to the specified standard question group with the highest integration score as the answer to the user's question. The server can then return this standard answer to the source of the user's question.

[0069] Through the above measures, the first group score determined based on the first similarity can reflect the relative similarity between the user's question and all questions; the second group score determined based on the second similarity can reflect the relative similarity between the user's question and the standard questions within the specified standard question group. By determining the integrated score using the first and second group scores, and then using the integrated score to determine the answer to the user's question, both precision and breadth characteristics can be considered when retrieving answers, thereby improving the overall retrieval accuracy of the FAQ system.

[0070] In one embodiment, see Figure 5 The following is a detailed flowchart of step 410 provided in an embodiment of this application, as shown below. Figure 5 As shown, when performing step 410, steps 411-413 can be performed.

[0071] Step 411: Determine the first similarity score of each question relative to the user's question based on the first similarity between the user's question and each question.

[0072] The first similarity score represents the relative similarity between the text of a standard or similar question and the text of a user's question.

[0073] The server can use the softmax function to process the first similarity of each question, thereby obtaining the first similarity score of each question relative to the user's question. The processing can be represented by the following formula (1):

[0074]

[0075] Here, x i The first similarity between the i-th question and the user's question is represented by e, where e is the natural base; N is the total number of questions in the FAQ corpus; x j Generally refers to the first similarity between any question and a user's question.

[0076] The sum of the first similarity scores of all questions determined by the above formula (1) is 1.

[0077] Step 412: Based on the first similarity score of all questions, determine several specified standard question groups.

[0078] After obtaining the first similarity scores for all questions, several standard question groups can be selected based on multiple first similarity scores to identify the overall text that is most similar to the text asked by the user, and these groups can be designated as standard question groups.

[0079] Step 413: Based on the first similarity score corresponding to each question within each specified standard question group, determine the first group score corresponding to each specified standard question group.

[0080] After filtering out the specified standard question groups, for each specified standard question group, the server can obtain multiple first similarity scores within that group and determine the first group score corresponding to that specified standard question group based on these multiple first similarity scores. Since the first group score is related to all first similarity scores within the group, it can characterize the overall similarity between the text of the specified standard question group and the text of the user's question, from the dimension of the retrieval scope including similar questions and standard questions.

[0081] In one embodiment, when the server determines several specified standard question groups based on the first similarity score of all questions, it can filter out a specified number of questions with the highest first similarity score as specified questions. Here, the specified number can be configured based on empirical values. For example, the specified number can be 10% of the total number of standard question groups; for instance, if the total number of standard question groups is 100, the specified number can be 10.

[0082] The selected questions are those with the highest relative similarity to the user's question. Generally, the selected questions can include standard questions and similar questions; if the user's question is not in standard language, the selected questions may only include similar questions.

[0083] When multiple specified questions are identified, the server can determine several standard question groups containing the specified questions. The standard question groups containing the specified questions are generally similar to the user questions and can be used as the specified standard question groups.

[0084] In one embodiment, when determining the first group score corresponding to each specified standard question group based on the first similarity score corresponding to each question within each specified standard question group, for each specified standard question group, in one case, the server can determine the maximum value of multiple first similarity scores in the specified standard question group as the first group score of the specified standard question group; in another case, the server can determine the average value of multiple first similarity scores in the specified standard question group as the first group score of the specified standard question group; in yet another case, the server can determine the median value of multiple first similarity scores in the specified standard question group as the first group score of the specified standard question group.

[0085] When using the median as the score for the first group, if the number of questions in the specified standard question group is even, the server can calculate the average of the two first similarity scores at the middle position and use this average as the score for the first group.

[0086] In one embodiment, before performing step 410, the server may determine the first similarity between each question in the FAQ corpus and the user's question based on a first similarity algorithm.

[0087] The first similarity algorithm can be any one of the following: bm25 (Best Match 25) algorithm, edit distance algorithm, or encoded vector distance algorithm.

[0088] In one embodiment, see Figure 6 The following is a detailed flowchart of step 420 provided in an embodiment of this application, as shown below. Figure 6 As shown, when performing step 420, steps 421-422 can be performed.

[0089] Step 421: Determine the second similarity score of each standard question relative to the user question based on the second similarity between the standard question and the user question in each specified standard question group.

[0090] The second similarity score represents the relative similarity between the text of the standard question and the text of the user question.

[0091] The server can process the second similarity of each standard question using the softmax function to obtain the second similarity score of each question relative to the user's question. The processing can be represented by the following formula (2):

[0092]

[0093] Here, x i This represents the second similarity between the standard question and the user question within the i-th specified standard question group; e is the natural base; N is the total number of specified standard question groups; x j Generally refers to the second similarity between standard questions and user questions within any specified standard question group.

[0094] The sum of the second similarity scores corresponding to the standard questions within all specified standard question groups determined by the above formula (2) is 1.

[0095] Step 422: For each specified standard question group, take the second similarity score of the standard questions within the specified standard question group as the second group score of the specified standard question group.

[0096] After obtaining the second similarity score corresponding to the standard questions within each specified standard question group, the server can use this second similarity score as the second grouping score for the specified standard question group to which the standard question belongs. Since the second grouping score is related to the standard questions within the group, the second grouping score can characterize the similarity between the text of the specified standard question group and the text of the user question from the dimension of the retrieval scope of the standard questions.

[0097] In one embodiment, before performing step 420, the server may determine the second similarity between the standard question and the user question for each standard question in a group of several specified standard questions, based on the second similarity algorithm.

[0098] The second similarity algorithm can be any of the following: bm25 algorithm, edit distance algorithm, or encoded vector distance algorithm. The second similarity algorithm selected by the server can be the same as or different from the first similarity algorithm.

[0099] The following explains the overall process of the FAQ system's answer retrieval method:

[0100] For example, the FAQ corpus of the FAQ system includes 10 standard question groups. Each standard question group includes one standard question and 10 similar questions corresponding to the standard question. Each standard question group corresponds to a unique standard answer. In this case, there are a total of 110 questions in the FAQ corpus.

[0101] After obtaining the user's question, the server can calculate the first similarity between the user's question and each of the 110 questions using the encoded vector distance algorithm, resulting in 110 first similarity scores. Further, the softmax function is used to determine the first similarity score for each question.

[0102] The server can filter out the 10 questions with the highest first similarity scores as designated questions, and determine that these 10 designated questions fall within three standard question groups. These three standard question groups are designated as group 1, group 2, and group 3. The server selects the maximum first similarity score within each designated standard question group as the first group score. At this point, the first group score for group 1 is 0.36, for group 2 it is 0.25, and for group 3 it is 0.14.

[0103] The server can use the encoded vector distance algorithm to calculate the second similarity between the standard question and the user question in each of the three specified standard question groups. Then, using the softmax function, it determines the second similarity score of the standard question within each specified standard question group, which is used as the second group score for that group. In this case, the second group score for group 1 is 0.4, for group 2 it is 0.3, and for group 3 it is 0.3.

[0104] The server can calculate the scores of the first and second groups for each specified standard question group using an average weighted average method, resulting in an integrated score of 0.38 for group 1, 0.275 for group 2, and 0.22 for group 3. In this case, the server can use the standard answer corresponding to group 1 as the answer to the user's question.

[0105] See Figure 7 This is a schematic diagram illustrating the determination of an integration score according to an embodiment of this application, as shown below. Figure 7 As shown, the FAQ corpus contains three standard question groups, with the standard questions in each group numbered 1, 2, and 3 respectively. Within the group containing standard question 1, the corresponding similar questions are numbered 1.1, 1.2, 1.3, and 1.4; within the group containing standard question 2, the corresponding similar questions are numbered 2.1, 2.2, 2.3, and 2.4; and within the group containing standard question 3, the corresponding similar questions are numbered 3.1, 3.2, 3.3, and 3.4.

[0106] The first similarity algorithm is used to determine the first similarity between the user's question and each question, and the first similarity score is obtained based on the first similarity. Figure 7 Scores 1, 1.1, ..., 3.4 represent the first similarity scores for each labeled question. The five highest first similarity scores are selected from these: score 1, score 1.3, score 1.4, score 2.2, and score 2.4. In this case, the designated standard question group can be determined as the group containing standard question 1 and standard question 2. Based on the first similarity scores within the designated standard question group, first group score 1 and first group score 2 are determined.

[0107] A second similarity algorithm is used to determine the second similarity between the user's question and standard question 1 and standard question 2, and a second similarity score is determined as the second group score 1 and the second group score 2. Based on the first group score 1 and the second group score 1, the integration score 1 of the group containing standard question 1 can be obtained; based on the first group score 2 and the second group score 2, the integration score 2 of the group containing standard question 2 can be obtained. After obtaining the integration score 1 and the integration score 2, the standard answer corresponding to the group with the higher integration score is taken as the answer to the user's question.

[0108] Figure 8 This is a block diagram of an answer retrieval device for a FAQ system according to an embodiment of the present invention, as shown below. Figure 8 As shown, the device may include:

[0109] The first determining module 810 is used to determine several standard question groups as specified standard question groups and a first group score for each specified standard question group based on the first similarity between the user's question and each question in the FAQ corpus; wherein, the questions in the FAQ corpus include standard questions and similar questions, and the standard question group includes one standard question and multiple similar questions;

[0110] The second determining module 820 is used to determine the second group score corresponding to each specified standard question group based on the second similarity between the standard question and the user question in each specified standard question group;

[0111] The synthesis module 830 is used to obtain a weighted sum of the first group score and the second group score for each specified standard question group to obtain the integrated score for each specified standard question group.

[0112] The third determining module 840 is used to determine the standard answer corresponding to the specified standard question group with the highest integrated score, which is the answer to the user question.

[0113] For details on the implementation process of the functions and roles of each module in the above-mentioned device, please refer to the implementation process of the corresponding steps in the answer retrieval method of the above-mentioned FAQ system, which will not be repeated here.

[0114] The apparatuses and methods disclosed in the several embodiments provided in this application can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatuses, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0115] In addition, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0116] If a function is implemented as a software module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

Claims

1. A method for retrieving answers to a FAQ system, characterized in that, include: Based on the first similarity between the user's question and each question in the FAQ corpus, several standard question groups are determined as designated standard question groups, and a first group score is determined for each designated standard question group; wherein, the questions in the FAQ corpus include standard questions and similar questions, and the standard question group includes one standard question and multiple similar questions; Based on the second similarity between the standard question and the user question in each specified standard question group, the second group score corresponding to each specified standard question group is determined; The scores of the first group and the second group for each specified standard question group are weighted and summed to obtain the integrated score for each specified standard question group; The standard answer corresponding to the specified standard question group with the highest integration score is determined as the answer to the user question; The determination of several standard question groups as designated standard question groups based on the first similarity between the user's question and each question in the FAQ corpus, and the first group score for each designated standard question group, includes: Based on the first similarity between the user's question and each question, a first similarity score is determined for each question relative to the user's question; Based on the first similarity score of all questions, determine several specified standard question groups; Based on the first similarity score corresponding to each question within each specified standard question group, determine the first group score corresponding to each specified standard question group; The step of determining the second group score corresponding to each specified standard question group based on the second similarity between the standard question and the user question in each specified standard question group includes: Based on the second similarity between the standard question and the user question in each specified standard question group, a second similarity score is determined for each standard question relative to the user question; For each specified standard question group, the second similarity score of the standard questions within the specified standard question group is used as the second group score of the specified standard question group.

2. The method according to claim 1, characterized in that, Based on the first similarity score of all questions, several specified standard question groups are determined, including: Select a specified number of questions with the highest first similarity score as the designated questions; Determine several standard question groups to which the specified question belongs, and use them as the specified standard question groups.

3. The method according to claim 1, characterized in that, The step of determining the first group score corresponding to each specified standard question group based on the first similarity score corresponding to each question within each specified standard question group includes: For each specified standard question group, the maximum, average, or median value of multiple first similarity scores in the specified standard question group is determined as the first group score of the specified standard question group.

4. The method according to claim 1, characterized in that, Before determining a plurality of standard question groups as designated standard question groups, and a first group score for each designated standard question group, the method further includes: For each question in the FAQ corpus, a first similarity is determined between the question and the user's question according to a first similarity algorithm.

5. The method according to claim 1, characterized in that, Before determining the second group score corresponding to each specified criterion group, the method further includes: For each standard question in a group of specified standard questions, a second similarity between the standard question and the user question is determined according to a second similarity algorithm.

6. A device for retrieving answers to a FAQ system, characterized in that, include: The first determining module is used to determine several standard question groups as specified standard question groups and a first group score for each specified standard question group based on the first similarity between the user's question and each question in the FAQ corpus; wherein, the questions in the FAQ corpus include standard questions and similar questions, and the standard question group includes one standard question and multiple similar questions; The determination of several standard question groups as designated standard question groups based on the first similarity between the user's question and each question in the FAQ corpus, and the first group score for each designated standard question group, includes: Based on the first similarity between the user's question and each question, a first similarity score is determined for each question relative to the user's question; Based on the first similarity score of all questions, determine several specified standard question groups; Based on the first similarity score corresponding to each question within each specified standard question group, determine the first group score corresponding to each specified standard question group; The second determining module is used to determine the second group score corresponding to each specified standard question group based on the second similarity between the standard question and the user question in each specified standard question group; Specifically, based on the second similarity between the standard question and the user question in each specified standard question group, a second group score is determined for each specified standard question group, including: Based on the second similarity between the standard question and the user question in each specified standard question group, a second similarity score is determined for each standard question relative to the user question; For each specified standard question group, the second similarity score of the standard questions within the specified standard question group is used as the second grouping score of the specified standard question group; The synthesis module is used to obtain a weighted sum of the first group score and the second group score for each specified standard question group, so as to obtain the integrated score for each specified standard question group. The third determining module is used to determine the standard answer corresponding to the specified standard question group with the highest integrated score, which is the answer to the user question.

7. An electronic device, characterized in that, The electronic device includes: processor; Memory used to store processor-executable instructions; The processor is configured to execute the answer retrieval method of the FAQ system according to any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The storage medium stores a computer program that can be executed by a processor to perform the answer retrieval method of the FAQ system according to any one of claims 1-5.

Images