Word vector-based high-efficiency semantic expansion retrieval method and device and storage medium

Abstract

The invention relates to a word vector-based high-efficiency semantic expansion retrieval method and device and a storage medium. The method comprises the following steps: S1, carrying out corpus cleaning to obtain corpus in a uniform format; s2, performing word segmentation on the corpus to obtain vocabularies; s3, converting all vocabularies into word vectors by utilizing the trained word vector model; s4, establishing an index for the word vector by adopting a binary tree method based on the cosine distance; s5, receiving a retrieval keyword, converting the retrieval keyword into a word vector, and obtaining a synonym by using the established index; s6, carrying out word segmentation on the keywords; step S7, performing research and report retrieval on the original word, the synonym and the segmented word to obtain an original word retrieval result, a synonym retrieval result and a segmented word retrieval result; and S8, taking the research report which appears in the original word retrieval result, the synonym retrieval result and the word segmentation retrieval result at the same time as a highest priority result to be presented. Compared with the prior art, the method has the advantages of improving response speed and accuracy and the like.

Description

technical field [0001] The invention relates to the field of text retrieval, in particular to a high-efficiency semantic extended retrieval method, device and storage medium based on word vectors. Background technique [0002] Financial companies such as banks and securities firms will conduct research on specific industries, markets and policies and form research reports. Massive reports will consume a lot of time for practitioners to select suitable research reports, and strict searches for specific vocabulary may filter out reports with similar information. [0003] In this regard, some existing technologies provide fuzzy query or retrieval methods to supplement the retrieval results. Usually, in the existing technologies, the search vocabulary is supplemented, and the synonym derivation algorithm is applied to the search scene of the research report, thereby Expanding proprietary words not only ensures that the relevant content of the original search terms can be displa...

AI Technical Summary

Problems solved by technology

However, due to maintenance costs and other reasons, first of all, the number of words included in this type of dictionary is limited, and generally only the query of high-frequency words can be guaranteed, and proper nouns in specific fields may not be supplemented. Secondly, its update frequency will be very high. Low, difficult to cover new words and hot words in time

Second, its response speed will be very slow

[0005] In this regard, some existing technologies have proposed a method of pre-establishing indexes, which can improve the response speed, but at present, very few of them are used in professional fields. The problem is that many of the vocabulary in the professional field are niche words. Many synonyms and participle have very different meanings. In addition, during the retrieval process, if the input retrieval formula is a sentence, too many invalid words will interfere with the retrieval results.

Images