Research Project Title:
Adapting PRIMERA for Query-Based Multi-Document Summarization
abstract:People and companies are inundated with written communication including email chains, texts, and financial documents. Natural language processing methods offers the prospect to significantly reduce reading time and expedite enterprise workflows with the implementation of models to condense knowledge and summarize text in formal and informal written commutation. Multi-document summarization (MDS) refers to the extraction of information written on the same topic. There are two main challenges to MDS: balancing between a concise and comprehensive summary, addressing conflicting views, and allowing multiple long text input. This project intends to investigate the use of an aspect-based entailment summarization approach to MDS that does not require any training. This novel method would allow synthesizing of summaries on relevant topics across multiple documents without any training.