Abstractive Summary of Public Opinion News Based on Element Graph Attention

The summary of case–public opinion refers to the generation case-related sentences from public information related judicial cases. Case–public news cases (intentional homicide, rape, etc.) that cause large opinion. in these usually contains case element such as suspect, victim, time, place, process, and sentencing case. In multi-document opinion, due problem cross redundancy between different documents under same case, order generate a concise smooth summary, this paper proposes an abstractive model based on attention diagram. Firstly, multiple are split into paragraphs, then paragraphs elements coded transformer method construct heterogeneous graph containing paragraph nodes nodes. Finally, decoding two-layer mechanism is applied node node, so can effectively solve generation.