Research on Chinese Text Summarization Generation Based on LSTM Models

This study develops an LSTM-based Seq2Seq model for Chinese text summarization. The encoder combines LSTM and CNN layers to capture contextual and local semantic features, while the decoder uses attention mechanisms for dynamic content focus during generation. Trained on the LCSTS dataset (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{2. 4 M}$</tex> Weibo texts) with Jieba segmentation and word embedding preprocessing, the model generates coherent summaries. Evaluation shows promising results with ROUGE metrics and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{2 5 \%}-\mathbf{5 0 \%}$</tex> accuracy on samples. Future work will explore Transformer architectures and pretrained embeddings for enhanced performance.