Word vector matrix compression method and device, and method and device for obtaining word vectors

Abstract

The present application discloses a word vector matrix compression method and device, and a method and device for obtaining word vectors, which relate to the technical field of data processing and help to save storage space of client devices. The word vector matrix compression method includes: generating a word vector matrix for representing the vocabulary to be processed based on the word vector model, and using the generated word vector matrix as the word vector matrix to be compressed; one row or column of the word vector matrix to be compressed is a word Vector, a word vector in the word vector matrix to be compressed is used to represent a word in the vocabulary to be processed; according to the semantic information of the vocabulary to be processed, the word vectors included in the word vector matrix to be compressed are classified to obtain at least two categories; Compressing at least one of the at least two categories, and constructing a compressed word vector matrix according to the compressed word vector of the at least one category, and storing the compressed word vector matrix.

Description

technical field [0001] The present application relates to the technical field of data processing, and in particular to a word vector matrix compression method and device, and a method and device for obtaining word vectors. Background technique [0002] With the development of deep learning technology, the performance of natural language processing (NLP) tasks has been greatly improved. Among them, NLP tasks can include: word segmentation, part-of-speech tagging, named-entity recognition (NER ), sentence classification, dialogue system, etc. Based on deep learning technology, the core of NLP tasks is to represent vocabulary through language models. The language model not only contains the meaning of the words themselves, but also reflects the relationship between different words, such as synonyms, antonyms, and contextual relationships. [0003] The word vector matrix is ​​a specific representation of the language model. A word vector matrix is ​​a matrix composed of one or...

AI Technical Summary

Problems solved by technology

[0004] Because the language has the characteristics of large vocabulary and complex syntax and grammar, a large number of features are required to describe it. Therefore, the generated word vector matrix is ​​large in size and occupies a relatively large storage space. Usually, it cannot be directly applied to the client device.

Images