Bank performance system question-answering robot corpus construction method and system

Abstract

The invention provides a bank performance system question and answer robot corpus construction method and system. The method comprises the steps of crawling a demand document of a first server through a web crawler model; preprocessing the demand document to generate a text unit; wherein the preprocessing comprises the following steps of: converting low-case characters and removing punctuations; extracting a keyword group of the text unit through a preset rule; and constructing a performance professional knowledge graph according to the key phrases. According to the method, the keywords in the document are extracted through NLP (Natural Language Processing), and manual intervention is not needed; the knowledge graph is constructed through knowledge reasoning, namely, bank performance assessment related content with extremely high professionality is generated, grammar and content are not different from natural languages, and the quality is guaranteed; performance specialty words are extracted and generated through a large number of demand document contents, and the part of demand documents are precipitated through long-term investigation and mining by demand personnel.

Description

technical field [0001] The invention relates to the technical field of artificial intelligence, in particular to a method and system for constructing a question-and-answer robot corpus for a bank performance system. Background technique [0002] With the development of artificial intelligence technology, intelligent question-answering robot systems are gradually emerging. Intelligent question-answering robots need a large amount of corpus to train in order to improve the accuracy of answers. The current mainstream corpus generation tools are mainly composed of corpus entities and corpus engines. The content of corpus entities is mainly vocabulary, that is, vocabulary such as subject-predicate-object. The functions of corpus engine include word frequency statistics, keyword indexing, etc. The engine generates the corpus we need by calculating the corpus entities. [0003] However, the corpus entities used by existing question-answering robots need to be manually collected, ...

AI Technical Summary

Problems solved by technology

[0003] However, the corpus entities used by existing question-answering robots need to be manually collected, and the amount of manual collection determines the quantity and quality of the final corpus to a large extent; the corpus collected manually cannot realize the association of synonyms and similar words, in other words To a certain extent, the corpus is just a combination of thesaurus; for highly specialized corpus, the collection and arrangement of the thesaurus requires professionals to spend a lot of time collecting and sorting, and the corpus engine may be different from natural language when segmenting these corpus entities. , so that the final corpus is far from expected, and the model and algorithm of the engine need to be readjusted, which increases a lot of workload

Images