I read and excellent article in the NYT technology section today and came across a term that hits home: sentiment mining. A long time ago we posted about “We feel fine” and since then, it seems that sentiment mining has gone from an interesting art project to a money-making technology.
In the article, the founders of Tweetfeel said that the best they could get at recognizing sentiment with automated systems was 70-80% effectiveness. After our brief, inexaustive trial of Tweetfeel we feel it was more around 50-75%. It is a safe bet that it will take a long time before automated systems will be effective enough to make a quantitative evaluation of sentiment.
Solutions such as Tweefeel and ScoutLabs are excellent for gauging the zeitgeist or the direction of the wind, and they are cost effective for that purpose. But business questions are often impossible to formulate simply – and emerging trends almost always start as eddies in the main wind. The mathematical sophistication to find these eddies in torrents of data must be coupled with a human analysis at some point to understand the particular linguistic and cultural differences that arise in each particular business context.
Sentiment mining is a great term, but a little optimistic when not coupled with some form of qualitative analysis. When processor power grows even cheaper and when the tools now used by folks such as our local Nstein move out of the enterprise software domain and become more available to consumers, sentiment mining might simply become part of a normal web search… at that time, and not before, could we say that the new (aka semantic) web has arrived.
