Systems Architecture and Algorithm for Co-reference Resolution in Texts

Co-reference resolution is a process in which all noun phrases referring to a single entity will be categorized in a distinct class. This contributes a lot to the accurate understanding of textual semantic as well as disambiguation; hence, it has notable usage NLP applications including machine translation, automatic text summarization, and automatic question answering as well as information extraction. This research area was first emerged through linguistic theories and rule-based methods. Then, since these linguistic methods were inefficient in situations where understanding the semantics was necessary for choosing the reference, machine learning based methods were proposed and in a number of these methods, linguistic theories were added to the data in the form of features and were mostly supervised learning. In the current study, different system architectures, methods, and algorithms of co-reference resolution are reviewed and mostly those studies are considered which have significant fundamental or structural distinction.