Event classification from text based commentaries for sports analytics

Abstract

Despite being relatively a new field, which is still evolving, sports analytics have already proven to provide a competitive advantage to sports teams. Especially with large number of commercial leagues taking place across the globe for every sport, the insights generated through analytics has helped franchise owners in bidding auctions for selecting players, where every decision the owners make, will cost them billions of money. However, most of the research work done in sports analytics such as score predictions, player profile analysis has been involved with mainly number crunching, with score boards being the predominant information source of such analytical work. Over time research had been focused on exploring other sources of information such as video and audio analysis, social media text analysis, micro blogging analysis etc. But there still exists large amount of untapped data with potential of being used for sports analytics. One such data source is online sports commentaries where the web sites give live updates on games, in the form of a temporal series such as minute by minute commentary for soccer or ball by ball commentary for Cricket. These publicly available commentaries give not only the scores, but coverage of all events happening during the match, including the ones which do not go into the scoreboards such as "dropping a catch" in a cricket match. If this information can be captured through event extraction and stored as structured data, that could be useful for analytical purposes. In this research an end to end system is proposed to produce structured data from online sports commentaries, while extracting information from the commentary text during the process.