Method, device and electronic device for determining word representation vectors

By introducing glyph unit information and multi-level word representation methods, the problem of out-of-scope characters in pictographic languages ​​is solved by the charCNN model, which improves the accuracy of word representation vectors and the prediction accuracy of language models.

CN112926334BActive Publication Date: 2026-04-28BEIJING SAMSUNG TELECOM R&D CENT +1
2 Cites 0 Cited by

Patent Information

Application Number
CN202010609924.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2019-12-06
Filing Date
2020-06-29
Publication Date
2026-04-28
Estimated Expiration
2040-06-29

AI Technical Summary

Technical Problem

Existing neural network language models based on charCNN cannot effectively handle words composed of characters beyond the syllabus when processing pictographic languages, resulting in inaccurate word representation vectors and affecting the performance of the language model.

Method used

By introducing glyph unit information and adopting a multi-level word representation method, this method extracts and fuses features at the glyph unit, character, and word levels. Using an attention-based feature extraction method, it extracts semantically relevant information from the characters that make up a word, weakens irrelevant information, and improves the accuracy of word representation vectors.

Benefits of technology

Effectively obtaining word representation vectors for pictographic languages ​​improves the performance of language models, especially for characters or words beyond the syllabus, enhancing the accuracy of word representation vectors and the overall prediction accuracy of the language model.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

Embodiments of the present application provide a method and device for determining word representation vectors, electronic equipment and computer readable storage medium, and belong to the field of natural language processing. The method comprises: obtaining a set of word units of a text; obtaining a context vector of the text based on the set of word units; and predicting a next word of the text based on the context vector. The method for determining word representation vectors can effectively obtain a corresponding set of word units even for pictographic characters or languages evolved from pictographic characters that are prone to out-of-vocabulary words, thereby improving the accuracy of determining word representation vectors.
Need to check novelty before this filing date? Find Prior Art