A Framework for Semantic Information Extraction from Social Media Data

Abstract

Information available in digital format has exceeded the capability of human brain to process, thus the processing mechanism has become an important responsibility of machines. Understating and extracting valuable information from such text have been a challenge over the past decade. Many valuable research works are done in this area and as a result there are many efficient existing approaches. Most of them are domainbased approaches and employ only shallow syntactic features in the extraction process. Also, these available approaches have limited capabilities of analysing the content semantically. This work introduces a novel approach which identifies the basic relations of the entities in a digital text in terms of triplets, in the given domain of music news retrieved from social media and extracts information to a structured format. This approach has been implemented as an independent system which receives content in unstructured format as the input and gives a structured file as the output. Proposed approach consistss of three main components, namely Data Retrieval, Pre-processor, Triplet Extractor and gives a special attention to analyse content in a digital text semantically. This would be a unique solution when finding considerable amount of information from free text as it consists of a methodology of semantically analysing text content and provide requested information by the user. Question and Answering module has been implemented to perform evaluation of the main system which is included as the fourth component of the system. Through the Evaluation performed using this Q&A module, 67% of accuracy was shown, thus this research work presents a powerful generic information extraction mechanism to grab knowledge from digital text available in various resources.