Text Representation Based on WT-GloVe Word Vector Weighting Model

The text representation issue in text classification is thoroughly examined in this paper, with a focus on handling massive volumes of text data. The paper suggests the word vector weighting model based on WT-GloVe in response to the issues with excessive sparsity and subpar classification performance that classic neural network text representation models frequently confront. To find important information in the text, the model first extracts features from the text using a feature weighting approach based on word spacing and interclass contribution. This work then presents a filtering technique that tries to enhance the training quality and relevance of word vectors in order to address the issue of irrelevant words that may be introduced by the GloVe model in word vector training. This work effectively builds a weighted word vector model that can precisely capture the value of feature words and rich semantic information by combining the feature weighting technique with irrelevant word screening. According to experimental results, the model performs better in numerous text classification assignments when compared to existing models, demonstrating its efficacy in enhancing text representation performance. The paper enriches the field of text representation by offering a fresh approach and aiding in the advancement of Natural Language Processing technology.