The SENSEI Overview of Newspaper Readers’ Comments

摘要

Automatic summarization of reader comments in on-line newsudis a challenging but clearly useful task. Work to date has produced extractiveudsummaries using well-known techniques from other areas of NLP.udBut do users really want these, and do they support users in realisticudtasks? We specify an alternative summary type for reader comments,udbased on the notions of issues and viewpoints, and demonstrate our userudinterface to present it. An evaluation to assess how well summarizationudsystems support users in time-limited tasks (identifying issues and characterizingudopinions) gives good results for this prototype.