Deep Hash Model for Similarity Text Retrieval

Text hashing transforms a text into a binary hash code, making similar texts have similar hash codes. Text hashing can reduces storage and improves retrieval efficiency of similar texts retrieval, but integrating semantic characteristics into hash code of text is difficult and is a hot topic. A deep hashing network model based on SE-Inception + Bi-LSTM + Attention mechanism is proposed in this paper. The Bert pre-training language model is used to preprocess text content to obtain the input vector, and the hash layer is added to generate the binary hash code of the text. The proposed model deeply combines semantic information to generate hash code and adds SE module with attention mechanism to improve the performance of neural network. The experimental results showed that the proposed model is superior to other model in extraction accuracy and classification accuracy.