Automated Sentiment Analysis: the categorization of film reviews using Systemic Functional Linguistics (original) (raw)

Computers and emotions, up to this day, do not seem to have anything in common. Machines are assumed to be cold and insensitive and yet, considering that people spend a substantial part of their lives in front of computers, this is very unfortunate. Imagine discussing movies with your computer. The dream of articulate machines is so strong, nonetheless over half a century after the invention of the first data processing machines it remains just a dream. Therefore, having a conversation with a computer about the latest James Bond movie, as one would with a friend or a colleague, is excluded. A computer cannot watch a movie, as is colloquially intended, in the first place. However, take the case of a person who reads various articles and reviews about a certain film. Even though he has never watched the film he may still have gathered enough information to form an opinion about it. This information may be gathered from magazines, newspapers and books; however, most of it is readily available on the Web. Besides on the Web it is possible to find ancillary types of information, such as forums where users share their views with each other. Computers are excellent at manipulating data; in fact it is what they do best. With today’s technology such programs exist that can crawl the Web and collect data based on specific criteria. They are called Web crawlers and are commonplace. So far so good, a computer connected to the Internet has access to the necessary information and can use a Web crawler to gather it. But how can a machine convert that information into the knowledge required to form an opinion about a movie? The types of facts needed to interpret a given text have been thoroughly investigated. This report does not attempt to implement them in their full entirety – nobody has so far. Instead, on the principle of divide and conquer, a portion of these findings is tackled here that concentrates on the expressiveness of adjectives. Although taking into account such a small segment may sound rather limited the intent here is not of producing machines that can read a document and wholly comprehend it but rather of an application that can filter documents based on two opposing factors: positive and negative sentiments.

Sign up for access to the world's latest research.

checkGet notified about relevant papers

checkSave papers to use in your research

checkJoin the discussion with peers

checkTrack your impact