Information Extraction: A Smart Calendar Application: Using NLP, Computational Linguistics, Machine Learning and Information Retrieval Techniques

The amount of information available on the web and other electronic formats is increasing at a rapid rate. Moreover, e-mails are now becoming the preferred mode of communication. This thesis investigates various Information Extraction techniques (Tokenization, POS Tagger, Chunker, NER, Co-reference Resolution) and develops a system that inferences calendar appointments from a user's e-mail account. More specifically, the system identifies the subject, date and time of an appointment and upon user confirmation enters it into a calendar service. It makes use of an intelligent user feedback mechanism that helps tailor the system towards individual users. A novel approach adopted towards constructing rules to identify entities in the absence of a domain relevant corpus, reinstates the importance of a rule-based approach towards building a Named Entity Recognizer. It allows the system to be easily extended and helps identify unseen patterns without much domain expertise. Finally, the thesis tries to provide a data format that could be used in future systems, paving the way for a world in which devices could truly communicate with each other.