Sign In

Communications of the ACM

ACM TechNews

Mining the Web for Feelings, Not Facts

View as: Print Mobile App Share:

Credit: iStockphoto

Opinions expressed on the Internet can determine the eventual success or failure of a product, and tools stemming from the emergent field of sentiment analysis could not only help businesses bolster their revenues, but ultimately facilitate a transformation of the experience of searching for information online.

For example, Jondage provides a service aimed at online publishers that enables them to incorporate opinion data gathered from more than 450,000 sources, including mainstream news sources, blogs, and Twitter. The service uses a sophisticated algorithm that assesses sentiments about specific topics while also identifying the opinion holders with the most influence. Jondage's early investors include the U.S. National Science Foundation, and it is presently developing an algorithm that could tap opinion data to anticipate future occurrences, such as projecting the effect of newspaper editorials on a company's stock price.

Meanwhile, the Financial Times' experimental Newssift program tracks sentiments about business topics in the news, combined with a specialized search engine that lets users organize their queries according to topic, organization, place, person, and theme. The simplest sentiment analysis programs scan keywords to interpret a statement as positive or negative based on binary analysis, but important subtleties of human language—irony, sarcasm, slang, and so forth—are overlooked.

"We are dealing with sentiment that can be expressed in subtle ways," says Yahoo researcher Bo Pang. She has created software that analyzes several different filters, including polarity, intensity, and subjectivity, to determine a statement's actual intent.

From The New York Times
View Full Article


Abstracts Copyright © 2009 Information Inc., Bethesda, Maryland, USA


No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account