Shallow parsing of Gujarati text

Abstract

Shallow parsing is the process of assigning tag to minimal, non recursive phrase of the sentence. It is useful for many applications like question answering system, information retrieval where there is no need of full parsing. Gujarati is one of the main languages of India and 26th most spoken native language in the world. There are more than 50 million speakers of Gujarati language worldwide. Natural language processing of Gujarati is in its infancy. Now days there are many data available in Gujarati on websites but due to lack of resources it is hard for users to retrieve it efficiently. So, shallow parsing of Gujarati can make task easier for another tasks like machine translation, information extraction and retrieval. In this thesis, we have worked on the automatic annotation of Shallow Parsing of Gujarati. 400 sentences have been manually tagged. Different Machine Learning techniques namely Hidden Markov Model and Conditional Random Field have been used. We achieved good accuracy and it is similar to Hindi chunker even though resources available for Gujarati are very less. The best performance is achieved using CRF with contextual information and Part-of-speech tags.