Sophisticated Text Mining Can Provide Context for Sentiment Analysis

One of the most powerful features of semantic technology is its ability to contextually understand content; to derive meaning from language. Isolating specific terms around positive or negative reaction in relation to a specific context can reveal the sentiment behind a consumer expression of intent or consideration.

We have an interesting dashboard of data related to consumer TV behavior around:

Viewing Intent – captures text related to what consumers intend to watch or not watch. For example, “I missed the show last night and have the TV set to remind me to watch it at 8pm tonight.”

Watching – relates information specific to shows that are actively being watched.

Favorite – captures specific and strong indicators of positive reaction, capturing variations of phrases such as “my new fave”, “the best…”, “one of my favorites”, etc.

Affinity – identifies specific positive phrases such as “I love…”, “…is fantastic”, etc…

Each of these consumer intentions or expressions represents a dimension, a sort of status of interest in the following shows:

Pawn Stars
Antiques Roadshow
Storage Wars
American Pickers
Hardcore Pawn
Auction Kings

Click image to enlarge

Once we have filtered conversations to include only on-topic text related to Watching, we can then apply a sentiment filter.

Click image to enlarge

In the image above, two important views of data related to consumer behavior are being monitored. Not only are you able to identify a specific audience of active viewers watching a specific show but you can also begin to surface sentiment related to that behavior. Monitoring these metrics over time can provide insights on any outreach efforts, campaigns or general perception of a show.

If you’re interested, you can also view the video we created around the use of dimensions and sentiment.