A Platform for Event Extraction in Hindi

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 2241–2250 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 2241 A Platform for Event Extraction in Hindi Sovan Kumar Sahoo, Saumajit Saha, Asif Ekbal, Pushpak Bhattacharyya Department of Computer Science and Engineering Indian Institute of Technology Patna Patna, India {sovan.pcs17, saumajit.mtmc17, asif, pb}@iitp.ac.in Abstract Event Extraction is an important task in the widespread field of Natural Language Processing (NLP). Though this task is adequately addressed in English with sufficient resources, we are unaware of any benchmark setup in Indian languages. Hindi is one of the most widely spoken languages in the world. In this paper, we present an Event Extraction framework for Hindi language by creating an annotated resource for benchmarking, and then developing deep learning based models to set as the baselines. We crawl more than seventeen hundred disaster related Hindi news articles from the various news sources. We also develop deep learning based models for Event Trigger Detection and Classification, Argument Detection and Classification and Event-Argument Linking. Keywords: Event Extraction, Event Trigger, Arguments, Event-Argument Linking 1. Introduction With the massive growth of electronics technologies and a plethora of digital platforms, an enormous amount of digi- tal content is uploaded every day on the internet. Propelled by this technological advancement, the scope of journalism has also expanded. Nowadays all the leading news agencies have an online version of their newspapers which allow them to publish regular updates on any current event almost in- stantly. However, extracting relevant information manually about any event from this large amount of data is a near- impossible task. Event extraction, which is an essential task in Information Extraction (IE) deals with this challenge by developing tools and techniques to mine the required infor- mation from such unstructured data. Event is something that happens in a specific place at a time or during a par- ticular interval of time. Arguments are the mentions which are involved in any particular event like location, time, par- ticipants involved, and so on. Event trigger and argument trigger refer to any word or phrase which describe an event and argument, respectively. Event trigger detection, event trigger classification, argu- ment trigger detection, argument type classification and event-argument linking are the crucial sub-tasks of a typi- cal event extraction system. In event trigger detection task, word or group of words or phrase which indicates an oc- currence of an event are detected, whereas in event trigger classification, the task is to classify the detected event trig- gers into predefined event types. Similarly, in argument trigger detection, the word or group of words or phrase which indicate an argument of an event are identified, and the motive of argument type classification is to classify each argument trigger into predefined argument roles. The research community has attempted these problems with utmost attention. But trigger detection and classification are not sufficient for any event extraction system because there may be multiple event triggers of various types …

Content truncated to 500 words