Advisor: Dr. Hyoil Han
Click here to view the project homepage
We are conducting research to find more effective ways to generate concise extractive summaries from sets of documents focused on a common topic. An extractive summary is one that is generated by extracting key sentences from source documents. An example of a document set with a common topic would be a set of documents from digital news sources which are focused on a particular current event. A multi-document summary can be used for several purposes. For example, suppose a user searches through a large document collection using an information retrieval tool and obtains a set of related documents in the results of a particular search. A summary of these documents could be generated, and it is possible that this summary could contain all of the information that is required by the user for his or her purpose. If the summary itself does not contain all of the information needed by the user, it will give the user a good sense of whether the required information is contained in the set of documents for which the summary is written. We are developing two proposed approaches to the MDS problem. In the first approach, we will examine sentence features directly obtainable as statistics of the text data. In second, we will utilize a semantic parsing of source document sentences.