The blossoming of a significant number of social networking sites, blogs, and microblogs has given a podium for general masses to voice their opinion regarding social topics, economic issues, political matters, market trends etc. However, this sudden eruption of review data had opened floodgates to unmanageable records as it is almost impossible for any individual or organization to manually extract any useful information from it. Opinion mining or sentimental analysis is a natural language processing which can obtain the opinion or feeling of people about any particular product or subject. The main focus of this paper is to find a method to perform sentiment analysis of Twitter which is one of the most prevalent microblogging sites. The lexical method proposed in this paper classifies the tweets as positive, negative or neutral depending on the polarity of the words in it. Also, the role of negation words has been investigated.