Seth Grimes, a business analytics strategist at Alta Plana Corporation presented on the power of sentiment analysis on the web. Like a genie, sentiment has magical powers, but we need to figure out how to release it. He opened with three assertions:
1. Human communications, online and offline, are inherently subjective
2. Online facts have business value
3. Opinion often masquerades as fact
Grimes demonstrated examples of sentiment, including web toys, booking sites, and search engines. Sentiment analysis is more than just counting, it’s “the task of identifying positive and negative opinions, emotions, and evaluations” according to a 2005 academic report. Decision support that uses sentiment requires automated tools. The automated tools turn attitudes into data. Grimes shared some examples of tools for sentiment analysis. In his examples, he reminded the audience that what is positive in sentiment for one company, it is often negative for a competitor.
Sentiment analysis isn’t easy to do right. Document-level analysis is often too blunt an instrument. Grimes argues that we need to go beyond keyword-based analysis, because keywords lack semantics. Neither human-only nor machine-only analysis is sufficient because human lack the power of machines and the ability to scale, and machine lack the sensitivity and flexibility of humans. Essentially, machines lack the common sense to appreciate some context.
Grimes believes that breaking sentiment into positive, negative, and neutral is insufficient, especially in a business case. Without offering more gradations of emotion, businesses cannot read intent signals from the data.