Existing studies often concentrate on processing the mutual information between the question and document. To tackle question-driven abstractive summarization, the answer summary should be highly related to the concerned question. Consequently, question-driven abstractive answer summarization is studied to generate the concise and salient short answer, which is also informative for answering the question. In contrast to extractive methods, abstractive methods produce summaries at the word level based on semantic comprehension. Though extractive summarization is more grammatical and coherent, the extractive sentences fail to have a logical connection. These traditional extractive summarization methods are mainly based on information retrieval methods to select sentences that heavily rely on feature engineering, and the results performance is restricted by pipelines. However, these methods are typically based on semantic relevance from query to context and neglect mutual information at the sentence level, which is helpful for the reasoning or inference process in question-driven summarization. Some related studies treat this QA data set as a summarization task and take the conclusion part of the abstract as the answer summary.Įarly works put emphasis on query-based summarization approaches in which the aim is to extract the sentences relevant to the given query. PubMedQA is a novel biomedical nonfactoid QA data set collected from PubMed articles in which the title is a question and can be answered by yes or no. Summaries for nonfactoid questions should be semantically consistent and identical with the context. The answer of factoid QA is a phrase or a sentence according to the question, but users prefer the detailed answer including more information to the accurate answer. It is different from a factoid question-answering (QA) system. In the biomedical domain, question-driven answer summarization can be particularly useful for people whether they have a biomedical background or not because the generated summary only covers the key information with respect to a specific question and filters out the explanation part. Automatic text summarization of natural language aims to summarize the source document to generate a concise and informative description for helping people efficiently and quickly capture the main idea.
0 Comments
Leave a Reply. |
Details
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |