Open information extraction from texts. Part III. Question answering system

In this paper, we propose a question answering system prototype that works on top of an au-tomatically generated knowledge base. For the knowledge base construction, methods of open infor-mation extraction are used, as well as unsupervised learning methods. In particular, various deep clus-tering methods are investigated and applied. Using open information extraction methods, triplets of the form (object1; predicate; object2) are extracted, which are then clustered into semantic relations. The clustered triplets are collected into a graph database, which is a source of information to generate an answer. This study demonstrates the applicability of unsupervised relation extraction methods.