How Do We Manage The Adoption Of Social Media?
Corporate Blogging Is About Marketing: Or A Self Serving Post About My Book

Subtext In Sentiment Analysis

If you want to mine the web for insights into products and brands using automated tools. One problem to overcome is the subjective nature of language. A consumer might review a product, and use words that indicate positive or negative sentiment towards the product, but really mean entirely the opposite.

Lillian Lee in her presentation to the AAAI in 2008 wrote the following:

"Can't we just look for words like "great" or "terrible"?

Yes, but ...

1) This laptop is a great deal.

2) A great deal of media attention surrounded the release of the new laptop.

3) This laptop is a great deal ... and I've got a nice bridge you might be interested in."

The first line is positive, the second could be positive or negative, and third is definitely negative even though in all three reviews use the phrase "great deal”. Reading the review you will understand the subtext of the sentiment, but a machine, that's much more difficult to achieve because the context of words in a sentence has to be understand. That’s where the science of sentiment analysis provides some answers to determine a writer’s sentiment in an automated way for thousands if not millions of documents online.

Comments