Context Sensitive Verb Similarity Dataset for Legal Information Extraction

Existing literature demonstrates that verbs are pivotal in legal information extraction tasks due to their semantic and argumentative properties. However, granting computers the ability interpret meaning of a verb its properties relation given context can be considered as challenging task, mainly polysemic domain specific behaviours verbs. Therefore, developing mechanisms identify behaviors evaluate how artificial models detect with significant importance. In this regard, comprehensive dataset used an evaluation resource, well training data set, major requirement. paper, we introduce LeCoVe, which is similarity intended towards facilitating process identifying similar meanings context. Using dataset, evaluated both generic embedding models, were developed using state-of-the-art word representation language modelling techniques. As part experiments carried out announced Sense2Vec BERT trained corpus opinion texts order capture behaviours. addition demonstrate neural network model, was by combining semantic, syntactic, contextual features obtained from outputs perform comparatively well, even low resource scenario.