Pseudo-correlation feedback extended query method based on question-answering system

Abstract

The invention discloses a pseudo-correlation feedback extended query method based on a question-answering system. According to the pseudo-correlation feedback extended query method based on the question and answer system, mature semantic mining modules, such as attention mechanisms, in the question and answer system are used for reference, so that the model can truly understand the search intention of a user, and extended lexical items are selected according to interactive semantic information of query and documents. Compared with a traditional model, due to the fact that semantic interactioncharacteristics are added, the effect of selecting the extended lexical items is remarkably improved. In addition, a neural network based on a paired loss function is further added to understand statistical characteristics of lexical items, and word frequency, inverse document frequency and the like are used for correcting the possible semantic drift problem of the semantic model. Practice provesthat compared with a traditional pseudo-correlation feedback algorithm, the method has higher sorting accuracy and better robustness, and can be applied to various search scenes.

Description

technical field [0001] The invention relates to the field of information retrieval methods, in particular to the research on an extended query method based on a pseudo-correlation feedback algorithm in a search engine. Background technique [0002] Today, with the development of information technology, more and more people start to use search engines to search, browse and query relevant knowledge. Search engines use specific strategies to retrieve customized information from the Internet and return them to users based on user needs and some related algorithms. user. However, due to the diversification of the Internet ecology and the rapid growth of the amount of information, it is difficult for users to give the required queries accurately and efficiently, so users may only provide some short queries or a few query terms for search engines, which is It may cause the search engine to be unable to fully obtain the user's query intention, thus making it impossible to return th...

AI Technical Summary

Problems solved by technology

Its shortcoming is that only the context-related word embedding vector is used, and the semantic interaction information between the query and the document is completely ignored. At the same time, ignoring features such as word frequency may cause semantic drift of the extended query.

The problem with this method is that the influence of statistical features is calculated based on rules, and the generalization of the model is poor. It only uses the embedding vector of the term, and there is no effective means to mine the semantic connection between the query and the pseudo-related documents.

[0012] After analyzing relevant domestic and foreign patents and related research, the following conclusions can be drawn: the current pseudo-relevance feedback algorithm in the field of information retrieval cannot effectively utilize the semantic interaction information between user queries and pseudo-related documents, resulting in the relevant algorithm being unable to identify the user's Search intent, biased in query expansion

The use of statistical information such as word frequency cannot get rid of the fixed rule model, the generalization is poor, and the information of feature intersection cannot be fully utilized

Images