ImpactWatch

I have spent this summer working for the Bivings Group as an intern. The main focus of my internship has been working on ImpactWatch. One feature that strikes me as particularly interesting is the use of human editors to rate the sentiment of articles.

Many media measurement companies now use automated systems to rate the sentiment of news articles. While this system may be able to read and rate the articles faster than a human, we feel that that the complete automation of the system causes the information to lose value. The Bivings Group will continue to monitor the automated sentiment analysis technology and will switch to it if and when it is warranted.

We have found that while great strides have been made in the field of sentiment analysis, things like sarcasm and colloquial speech can often throw the system off. Another weakness of automated sentiment analysis programs is that they tend to mishandle trends. If a company has made a mistake, but is being praised for their efforts to correct that mistake, automated systems will often rate that as negative, when it should be positive. Sentiment analysis software also seems to run into problems when it comes to rating the sentiment of different entities in an article.

It is for all these reasons that the Bivings Group has chosen to stay with human editors for the sentiment rating in ImpactWatch. We are confident at this point that the extra time it may take to rank the articles is worth it because we know that we are giving our clients the most accurate data possible.

This entry was posted on Thursday, July 17th, 2008 at 4:19 pm and is filed under Blogs, ImpactWatch, ImpactWatch Features, Measurement. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.