CN-113821624-B - Answer retrieval method and device of FAQ system, electronic equipment and storage medium

CN113821624BCN 113821624 BCN113821624 BCN 113821624BCN-113821624-B

Abstract

The application provides an answer retrieval method and device of a FAQ system, electronic equipment and a computer readable storage medium, wherein the method comprises the steps of 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 first similarity between a user question and each question in a FAQ corpus; the method comprises the steps of determining a second grouping score corresponding to each specified standard question group based on a second similarity between the standard questions and the user questions in each specified standard question group, obtaining weighted summation of the first grouping score and the second grouping 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 the specified standard question group with the highest integrated score as an answer of the user questions. According to the scheme, the integrated score determined by the first grouping score and the second grouping score can be used for considering the breadth characteristic and the precision characteristic of the search, so that the answer of the user question can be obtained more accurately.

Inventors

LI LONGWEI
JIAN RENXIAN
MA YONGNING

Assignees

竹间智能科技（上海）有限公司

Dates

Publication Date: 20260508
Application Date: 20211008

Claims (8)

1. An answer retrieval method of a FAQ system, comprising: Determining a plurality of standard questions groups serving as specified standard questions groups and a first group score of each specified standard questions group based on first similarity between the user questions and each question in the FAQ corpus, wherein the questions in the FAQ corpus comprise standard questions and similar questions, and the standard questions groups comprise one standard question and a plurality of similar questions; determining a second packet score corresponding to each specified challenge packet based on a second similarity between the challenge and the challenge in each specified challenge packet; The first grouping score and the second grouping of each specified question grouping are weighted and summed to obtain an integrated score of each specified question grouping; determining standard answers corresponding to the specified standard question groups with the highest integrated scores as answers of the user questions; Wherein determining a number of question groupings as specified question groupings based on a first similarity between the user question and each question in the FAQ corpus and a first grouping score for each specified question grouping comprises: Determining a first similarity score for each question relative to the user question based on a first similarity between the user question and each question; determining a plurality of specified standard question groups according to the first similarity scores of all the questions; determining a first group score corresponding to each specified question group based on the first similarity score corresponding to each question in each specified question group; The determining a second packet score corresponding to each specified standard question packet based on a second similarity between the standard questions and the user questions in each specified standard question packet comprises: Determining a second similarity score for each question relative to the user question based on a second similarity between the question and the user question in each specified question group; For each specified inter-criterion group, a second similarity score for inter-criterion within the specified inter-criterion group is taken as a second group score for the specified inter-criterion group.
2. The method of claim 1, wherein determining a number of specified groups of questions based on the first similarity scores for all questions comprises: Screening out the specified number of problems with the highest first similarity score as specified problems; and determining a plurality of standard question groups where the specified questions are located as the specified standard question groups.
3. The method of claim 1, wherein determining the first group score for each specified question group based on the first similarity score for each question within each specified question group comprises: For each specified inter-criterion group, a maximum or average or median of a plurality of first similarity scores in the specified inter-criterion group is determined as a first group score for the specified inter-criterion group.
4. The method of claim 1, wherein prior to said determining a number of challenge packets as specified challenge packets, and a first packet score for each specified challenge packet, the method further comprises: and determining a first similarity between each question and each user according to a first similarity algorithm aiming at each question in the FAQ corpus.
5. The method of claim 1, wherein prior to said determining the second packet score for each specified inter-criterion packet, the method further comprises: For each question of a number of specified groupings of questions, a second similarity between the question and the user is determined according to a second similarity algorithm.
6. An answer retrieval device of a FAQ system, comprising: The first determining module is used for 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 first similarity between the user questions and each question in the FAQ corpus, wherein the questions in the FAQ corpus comprise standard questions and similarity questions, and the standard question groups comprise one standard question and a plurality of similarity questions; Wherein determining a number of question groupings as specified question groupings based on a first similarity between the user question and each question in the FAQ corpus and a first grouping score for each specified question grouping comprises: Determining a first similarity score for each question relative to the user question based on a first similarity between the user question and each question; determining a plurality of specified standard question groups according to the first similarity scores of all the questions; determining a first group score corresponding to each specified question group based on the first similarity score corresponding to each question in each specified question group; A second determining module, configured to determine a second packet score corresponding to each specified criterion group based on a second similarity between the criterion and the user in each specified criterion group; wherein determining a second packet score for each specified question group based on a second similarity between a question of the standard and the question of the user in each specified question group comprises: Determining a second similarity score for each question relative to the user question based on a second similarity between the question and the user question in each specified question group; For each specified challenge group, taking a second similarity score of the challenge within the specified challenge group as a second group score of the specified challenge group; the synthesis module is used for carrying out weighted summation on the first grouping score and the second grouping score of each specified standard question grouping to obtain an integrated score of each specified standard question grouping; And the third determining module is used for determining the standard answers corresponding to the specified standard question group with the highest integrated score as the answers of the user questions.
7. An electronic device, the electronic device comprising: A processor; A memory for storing processor-executable instructions; wherein the processor is configured to perform the answer retrieval method of the FAQ system of any one of claims 1-5.
8. A computer-readable storage medium storing a computer program executable by a processor to perform the answer retrieval method of the FAQ system of any one of claims 1-5.

Description

Answer retrieval method and device of FAQ system, electronic equipment and storage medium Technical Field The present application relates to the field of natural language processing technologies, and in particular, to an answer retrieval method and apparatus for a FAQ system, an electronic device, and a computer readable storage medium. Background The FAQ (Frequently Asked Questions, common problem solution) system may reply to the determined answer to the problem submitted by the user. The FAQ question-answering system is built, and the FAQ corpus is firstly arranged. The FAQ corpus can be composed of three parts, namely a standard question, a standard answer and a similar question. The standard questions are questions of formal and standard language comparison, a plurality of standard questions can be included in the FAQ corpus, the distinction between different standard questions is high, and the standard questions cover a plurality of different questions possibly asked by a user. The standard answers are answers to the standard questions, and if the complicated conditions such as dimension and the like are not considered, the standard questions are in one-to-one correspondence with the standard answers. For example, the standard question is that "whether there is a valid period after the opening of the science popularization board. "similar questions are extensions to standard questions, one standard question can correspond to a plurality of similar questions, and similar questions express the same meaning as corresponding standard questions, but the language of the similar questions is often not standard, and various question methods of the standard questions are covered as much as possible. For example, the standard question is "whether there is a valid period after the opening of the traumatology plate. Referring to fig. 1, an organization chart of a FAQ corpus according to an embodiment of the present application is shown in fig. 1, where each standard question maps a standard answer, and the standard questions correspond to a plurality of similarity questions. The FAQ search question-answering system is a form of the FAQ system. The FAQ search question-and-answer system, after receiving the user questions, may match the user questions with the standard questions and similar questions in the FAQ corpus. When the user is most similar to the user question, the standard answer mapped by the standard question is used as a system answer, and when the user is most similar to the user question, the standard answer mapped by the standard question is used as a system answer. However, the language of the question is normal, the degree of distinction is high, and the search question has a precision characteristic. When the question-asking method of the user is more common, the error probability of finding the standard question is lower. The number of the similar questions is large, the covered question is wide, and the search similar questions have breadth characteristics. When the question-asking method of the user is rare, by finding a similar question to the question-asking solution of the user, a correct reply can be given. The response mode of the related scheme does not effectively distinguish between different characteristics of the standard questions and the similar questions. Disclosure of Invention An object of an embodiment of the present application is to provide a method and an apparatus for retrieving answers of a FAQ system, an electronic device, and a computer readable storage medium, which are used for determining an optimal answer to a question under the condition of considering different characteristics of a standard question and a similar question. In one aspect, the present application provides an answer retrieval method for a FAQ system, including: Determining a plurality of standard questions groups serving as specified standard questions groups and a first group score of each specified standard questions group based on first similarity between the user questions and each question in the FAQ corpus, wherein the questions in the FAQ corpus comprise standard questions and similar questions, and the standard questions groups comprise one standard question and a plurality of similar questions; determining a second packet score corresponding to each specified challenge packet based on a second similarity between the challenge and the challenge in each specified challenge packet; The first grouping score and the second grouping of each specified question grouping are weighted and summed to obtain an integrated score of each specified question grouping; and determining the standard answers corresponding to the specified standard question group with the highest integrated score as the answers of the user questions. In one embodiment, the determining a number of question groupings that are specified question groupings based on a first similarity between the user question and each question in the