Question and answer method and system based on BERT and knowledge base

Abstract

The invention discloses a questioning and answering method and system based on BERT and a knowledge base, is applied to the field of information retrieval. A named entity recognition model based on BERT-CRF and a language model and a text similarity dichotomy model based on BERT and the language model in order to overcome at the defects of an existing knowledge base questioning and answering system. the two models are trained, the two trained models are adopted to process the linguistic data of the to-be-answered questions, correct answers to the questions can be obtained, and answers are automatically rewritten.

Description

technical field [0001] The invention belongs to the field of information retrieval, in particular to a question-and-answer search technology. Background technique [0002] Traditional question and answer search, based on keyword retrieval, does not take into account the semantic information of the question text. In the knowledge base question answering system, the questioner inputs specific question texts, analyzes and processes the question texts online, and then retrieves and outputs the most matching answer texts to obtain fast and accurate answers to questions. [0003] Knowledge base question answering systems and methods are mainly divided into three categories: [0004] 1) Method based on information retrieval [0005] By extracting question entities and attribute relations from the question text, and then searching in the knowledge base. [0006] 2) Method based on semantic analysis [0007] By retrieving the logical expression of the question text in the knowled...

AI Technical Summary

Problems solved by technology

[0011] 1. Based on semantic analysis methods, there is a barrier between logical expressions and natural language semantics;

[0012] 2. The method based on information retrieval cannot analyze the semantic information in the question text, especially the context information cannot be fully utilized to get the disambiguation of the entity;

[0013] 3. Existing models such as CNN, RNN, Bi-LSTM, etc., the model training effect, accuracy, F1 value, etc. are not as good as BERT, Transformer and other cutting-edge models, and lack the correlation between words or words in the question text sex analysis

Images