Amharic Phrase Chunking with Conditional Random Fields
Alemebante Mulu, Vishal Goyal · Journal of scientific research · 2020
This paper particularly presents a Conditional Random Field (CRF) based phrase chunking system for Amharic language.Chunking is the series of actions which has divided or segmented the sentence into phrase by the arrangements of correlated word groups.Therefore, chunking is recognized by the identity of chunk labels and the boundary that describes chunks.In this research work, our goal is to develop chunking for Amharic language by using different tagging schemes for identifying the chunk boundaries and incorporate the tangs in form of contextual information.Therefore we have identified the problem, common stetting, solution and improvement of chunking as well.In addition to this, we have also constituted the special word, Part-of-Speech (POS) tagging information, morphological analysis as an input to increase the performance of the system.Totally, we are using 400,000 tagged words and have evaluated the result by the combination boundary identification and labeling.Since we are using large amount of tagged data for Part of Speech tagging, it although performs good results with chunking.Finally, the average accuracy of the system is reached to 94.2%.