Complex valued word vector construction method based on positions and semantics

Abstract

The invention discloses a complex valued word vector construction method based on position and semantics, which comprises the following steps: searching a text classification corpus set, and dividingthe text classification corpus set into a training set, a verification set and a test set; preprocessing the text in the corpus set (removing stop words); constructing sentence representation by usingthe relative position information and the global semantic information; inputting the word vectors of the training corpus set into a complex-valued neural network, and training a semantic classification model; inputting the verification set text word vector into a complex-valued neural network model so as to calculate the prediction probability of each sample; testing the model obtained on the basis of the verification set on a test set; according to the method, the current situation that a text classification corpus set is relatively lacked is overcome, the feature information (position information) of the text can be extracted more fully, the position information and global semantic information of the text are fused, and the complex valued word vector is applied to the replicated neuralnetwork, so that neural network models have relatively strong discrimination capability.

Description

technical field [0001] The invention relates to the technical field of text classification, in particular to a method for constructing complex-valued vectors based on position and semantics. Background technique [0002] In the past few years, with the rapid development of science and technology, especially the rapid development of the Internet and social networks, all kinds of information are flooding the Internet, including comments and some of their own opinions published by users on social platforms. It has also become one of the main sources of information for users in their daily lives. People may obtain a large amount of information through the Internet, but how to manage these large amounts of information reasonably and effectively has become an issue that people are more and more concerned about. A very common management method for large amounts of information is classification, which shows that text classification contains huge social value. The present invention...

AI Technical Summary

Problems solved by technology

However, the current method only uses complex vectors, and does not make better use of complex-valued word vectors to mine the relative position information of words in sentences

Images