Sentiment: Semantics brings context to life
Last week, we participated in the panel discussion “Beyond Sentiment: Mining Social and other Media for More than Positive and Negatives” at the SemTechBiz conference in London. Our group Tom Reamy (KAPS Group), Marie Wallace (IBM), along with my colleague Marcello Pellacani from Expert System and our partner, Fabio Lazzarini from Cribis D&B, had a lively session around our individual perspectives and trends, techniques and ideas for the future.
As with all discussions of technology, the topics of “where we’re going” and “what’s next” were top of mind. We’re constantly looking forward, trying to anticipate future applications, functionality and drivers for our technology, not to mention staying on top of an industry that is forever working to create products and applications that will change our lives in ways we haven’t yet imagined.
As Marie mentioned in her thoughtful recap of the event, traditional content analytics have “hit a wall” in social media analysis. Semantic technologies are increasingly being recognized for their ability to provide a new level of understanding of context for analyzing social media. And as Tweets and posts on Facebook and Google+ continue to influence consumers, and therefore marketers, advertisers, product development, there is a real need to understand intent and opinion, beyond Like and sarcasm.
At Expert System, we continue to make investments in our technology to better understand sentiment in context and to better understand what’s behind an expressed opinion. These are some of the most important scenarios where semantics are making a difference:
- Improving processing of anaphora, which are used commonly in social media communications, such as posts on blogs, Facebook, Google+ or Twitter. The interpretation of anaphora is important for a correct analysis of sentiment among several posts, or even in the same post where information is not always explicity expressed. Correct processing of anaphora allows you to almost double the collection of opinions and sentiment expressed online (i.e., “I love the new iPhone” and, a few posts later “But I don’t love it any more” where the person is still talking about the iPhone).
- This also applies to identifying influencers. Analyzing style and tone of writing, you can also understand if someone writing under two different usernames is actually the same person. The same opinion expressed by different people has a different weight than the same opinion expressed repeatedly by the same person.
- Refining the analysis of sentences in which there is a comparison between entities, companies or products (“I think the Fiat 500 is much cooler than the Beetle.”).
- Improving analysis of difficult to interpret information such as that written in ungrammatical slang or acronyms, which is common in Tweets, Facebook posts and other communications between friends or limited by 140 characters.
- Distinguishing whether an expression is emotional or intentional (“It would be amazing if that dress was also in red” or “If there was a grey iPhone I’d buy one right now.”).
From the perspective of our customers, they want to be able to use sentiment to drive better decision making, and this is where semantic technology excels, and where we will continue to grow.
Author: Luca Scagliarini