2010-03-02

Predicting Sentiment in Advance

There are now a number of tools that monitor social networks looking at:

  • Sentiment analysis – General sentiments related organizations and their brands
  • Topic trend analysis – the relative frequency that topics are mentioned over time

Therefore, if I make blog post or tweet, my topic and sentiments will be captured by automated systems, analyzed and reported. There are some pretty sophisticated tools being used by Marketers, and while some are free, others are quite expensive. However, as a recent blog post noted: "A technology glitch demonstrates how fragile marketing measurement technology really is." That said, let's assume they'll get better or this Technorati glitch was atypical.

I can also manually get some trend information using Google trends for example looking at 'ECM' over time: http://trends.google.com/trends?q=%22ecm%22&ctab=0&geo=all&date=all&sort=0. However, the information on ECM is too sparse, and there is too much 'contamination' with other definitions of ECM, such as Engine Control Management.

But I have to admit I don't use these tools. So I need a tool like all great advances that caters to laziness – or increased efficiency as I might prefer to characterize it! I need help. I need push rather than pull technology. It occurs to me that I wouldn't mind knowing how my proposed post relates to other posts already made. I'd get a report something like:

"Your post on 'ECM' would be the 47th post on this topic so far this year. This topic is declining in frequency."

"Your 'negative' post on 'content system metadata' would align with 19% negative, 25% neutral and 63% positive posts on this topic."

Besides putting my proposed post in context, I wouldn't mind getting a sample of the most relevant posts so that I could potentially revise my post, add links, references, rebuttals, etc.

At some level this would be a form of assisted authoring. It wouldn't have to be limited to blog posts. I'd like to do the same for content authored in an enterprise context. The reality that many reports, white papers and similar work products of knowledge workers duplicate things already available, but people generally don't look. It's easier to start typing, imagining your work to be original, than to look first if it's already been done or if there is something close than you can build on.

Syndicated at http://conversations.opentext.com/

1 comment: