Event detection on literature by utilizing word embedding

초록

The events in literature refer to what the characters are going through, and the story revolves around them. In other words, figuring out events is understanding literature, which is an important concept in assessing the value of it. Event detection is the field of Information Retrieval and previous studies have been mainly based on Automatic Content Extraction (ACE) corpus. However, it costs a lot to make large datasets such as ACE and time consuming because all components like signals for events (entity, event mention, and event argument) are annotated. In addition, it is difficult to apply this large dataset to the special domain of literature. Therefore, we approach event detection in literature as using word embedding. Firstly, we make a set of keywords corresponding to each event, and then we apply Ranking by using word embedding. By utilizing this, this paper will provide a way of finding the sentences relating to the given queries, i.e. the sentences are supposed to be the event. Our method suggests a novel approach in event detection without creating large-scale datasets. © Springer Nature Switzerland AG 2020.

키워드