Using Semantic Analytics to Reduce Social Media Monitoring Blind Spots

My boss recently found an article on Social Media Explorer called Do Social Media Monitoring Services Leave Brands Blind?. It was largely about a comparison of the results surfaced by source-focused analytics product like Valuevine when compared to other keyword monitoring tools.  The comparison of the results was actually quite shocking but I’ll let you read the full article to get a better idea of the background.

My boss recently found an article on Social Media Explorer called Do Social Media Monitoring Services Leave Brands Blind?. It was largely about a comparison of the results surfaced by source-focused analytics product like Valuevine when compared to other keyword monitoring tools.  The comparison of the results was actually quite shocking but I’ll let you read the full article to get a better idea of the background.

Monitoring social media using keywords or Boolean expressions is easy to configure and to use but the results can be incomplete or inaccurate because this type of approach to social media analytics presumes you know all the terms that must be tracked in advance.  Keywords alone fail to disambiguate the meaning of “Crocs” the shoes or “crocs” the reptile. Unfortunately, as you add more and more expressions to exclude or include content the whole process begins to become very brittle.

But we’re not getting into any finger pointing because we actually use keywords and boolean expressions to identify and filter conversations. The difference is that we rely on latent semantic analysis (LSA) to refine our analytics by identifying and capturing conversations based on meaning.  LSA extracts specialized language features from large data sets, like social media conversations, and selects conversations based on their context and content. In other words, semantic technology is able to understand the difference between “Crocs” the shoes or “crocs” the reptile.

The next step is to apply natural language processing (NLP)  to extract specific language elements from a conversations. What exactly does that mean in layperson terms? It means that we apply another lens to the conversation to isolate consumer expressions around pricing, loyalty, quality or what we like to call dimensions. Dimensions can be customized to identify and extract specific elements of a conversations around consumer preferences, intentions or considerations. This approach allows you to surface consumer expressions or opinions in their own words, unsolicited and authentic.  And because the engine behind the analysis is semantic-based, author information is also included, so you end up with a very comprehensive view of the customer and their perspective on a given topic.

The system as a concept looks similar to this:

Click image to enlarge

CI’s approach addresses the inaccuracy and bluntness of keyword search and the speed and cost disadvantages of NLP techniques through the use of latent semantic analysis. It gives you the ability to identify and organize conversations that are most relevant to your analysis to help expand your view.