A Text Hash Retrieval Method Based on Deep Learning

Abstract

The invention discloses a text hash retrieval method based on deep learning, which is characterized by first using a bidirectional LSTM model to extract the semantic code corresponding to each original vocabulary data in the word embedding matrix, and then connecting the text volume in parallel after the bidirectional LSTM model Integrate the neural network and add the attention mechanism, and then use the sign function to convert the output value of the second fully connected layer into the corresponding hash code, use the hash code to reconstruct the category label, and finally, find the hash code in the text library The vector data closest to the Hamming distance coded by the retrieval text hash code completes the hash retrieval process of the retrieval text data. The advantage is that the hash model has a high learning ability for short texts, and the added attention mechanism can further improve the features. The classification layer uses hash coding to reconstruct category labels, so that the hash model can use label information more finely while learning binary coding, so the retrieval accuracy is higher.

Description

Technical field [0001] The present invention relates to a text hash retrieval method, in particular a depth learning text hash retrieval method. Background technique [0002] As the data size and dimension increases, the price of semantic retrieval increases, text hash as an important way to achieve efficient semantic retrieval; however, most text hash algorithms are directly using machine learning. The mechanism will map the explicit feature or keyword feature of this article. Inventive content [0003] The technical problem to be solved in the present invention is to provide a method of retrieving precision-based text hashing retrieval methods. [0004] The present invention solves the technical solution used in the above technical issues is: a method based on deep learning text hash retrieval method, including the following steps: [0005] 1 Gets to be retrieved by S original vocabulary data, and preprocessing the original vocabulary data for cleaning and particle pretreatmen...

AI Technical Summary

Problems solved by technology

[0002] As the scale and dimension of data increase, the cost of semantic retrieval increases sharply. As an important way to achieve efficient semantic retrieval, text hashing has received extensive attention; however, most text hashing algorithms directly use machine learning The mechanism maps the explicit features or keyword features in this paper to binary codes, which cannot effectively guarantee the semantic similarity between texts, resulting in low retrieval efficiency of codes