You are here

Summarizing sentiment in user generated content - Rob Blaauboer voor COMMIT/

The internet is big; really mind-bogglingly big … you just wouldn’t believe how big it is. This paraphrase of science fiction writer Douglas Adams’ quote on another big ‘thing’ (he was talking about space) is the perfect way to start an article on Sentiment summarization.

Because it is true, the internet is growing rapidly with new sites, pages, tweets and likes being added every second of every day. Users add their voices and opinions about almost everything imaginable: people, films, brands, companies and so on.
In these contributions there is often an opinion: their point of view on a specific matter. For companies this is a valuable source of information to gauge how people (both clients and non-clients) view their company, product or brand.

Researchers from Erasmus University Rotterdam and Teezir are looking into how they can summarize the sentiment of conversational documents in order to make them easily accessible.

Their empirical study (a part of the total research performed in the COMMIT program) focusses on identifying sentences that are most relevant to the opinion expressed in the article. Would a long sentence be an indication of sentiment and a good candidate for a summery? Or perhaps a sentence towards the end of the article, since a conclusion is naturally found at the end

Ook dit is een COMMIT/project