KEC_DAlab @ EventXtract-IL-FIRE2017: Event Extraction using Support Vector Machines

Nowadays, Social media has become a major part to transfer the message that must be shared with the people ideas and express the information globally in our day-to-day life. Through social media can able to connect the people together, they are vulnerable to crimes like Identity thefts, false information, and identity masking etc. Identifying the event from the social media messages and news headlines are the important area of research in the current era. This paper illustrates work done on Event Extraction for Indian language shared task which is conducted in Forum for Information Retrieval Evaluation (FIRE) 2017. For this Event extraction task, organizers release the dataset with three languages Tamil, Hindi, and Malayalam. Each language dataset consists of two files Original Tweet file and Annotation files. We only participated in the Tamil event extraction task. In this task, we converted the original tweet file into Bio-format to apply the machine learning directly. Then analyzing each chunk of the word is an event is said to [B] beginner and the other events will be given as Intermediate and the others are assigned as O tag. Each word or chunk should be trained whether it is the event or not an event with the help of rich features and SVM classifier. Here we also find out the CrossValidation accuracy using Natural language techniques.