Aspect Classification for Legal Depositions

Attorneys have interest in having a digital library with suitable services (e.g., summarizing, searching, and browsing) to help them work large legal deposition corpora. Their needs often involve understanding the semantics of such documents. In case tort litigation associated property casualty insurance claims, as relating an injury, it is important know not only about liability, but also events, accidents, physical conditions, treatments. We hypothesize that consists various aspects are discussed part deponent testimony. Accordingly, we developed ontology for accident injury cases. Using that, classifier can identify portions text each interest. Doing so was complicated by peculiarities this genre, e.g., transcripts generally consist data form question-answer (QA) pairs. our automated system starts pre-processing, then transforms QA pairs into canonical made up declarative sentences. Classifying sentences generated, according aspect, downstream tasks summarization, segmentation, question-answering, information retrieval. Our methods achieved classification F1 score 0.83. Having classified good accuracy will choosing be used candidate summary sentences, generate informative professionals or claim agents. methodology could extended depositions other kinds, aid like searching.