Event Extraction from Sport Forecasts

Electronic information exchanged online in different formats such as natural language documents, social media streams, news articles, media texts, blogs, sport fans websites etc, is rapidly increasing. Most of such prolific information is in an unstructured format, making it difficult for users to locate the most significant information they need. Here comes the vital role of Natural language techniques to extract the most important information from the unstructured fragmented noisy data with the aim of saving time and effort. This poses a challenging task for motivated researchers who have been introducing different approaches and algorithms to extract information, in particular its most prominent and common subtask, that is event extraction. This research is, to our knowledge, the first attempt, until this date, that uses a Convolutional Neural Network (CNN) with residual connections for event extraction from sports forecasts collected from football fans websites. It does not really stop at the task of extracting event mentions but also goes beyond that to extract named entities of competing teams as well as the winner and loser.