Patent Claims Summarization using LSTM based on Bahdanua Attention

Patents are the legal right granted for a disclosed invention. To figure out meritorious patents from the large collection of patent documents, manual reading, and the summarization approach is critically slow and inefficient. A novel patent claim summarization method is proposed to solve this issue. In this work, a new model named BbLSTM is proposed to perform automatic patent summarization. The summarization system identifies the semantic relationships among patents and corresponding summaries. Summaries created using BbLSTM outperforms the results from the other evaluated summarization techniques.