June 16, 2014, 10:01 AM — In the NLP (natural language processing) business for a while, Attensity sees an opportunity to get new customers with Q, a visualization tool it says can help non-technical users like marketers find insights in oceans of social media data.
"Any type of end user can pick up and start listening," said Martin Onofrio, Attensity's chief revenue officer. "You spin it up and start listening. It's that high-level observation tool."
Attensity's underlying platform provides real-time access to Facebook, more than 150 million blogs and forums, and the Twitter firehose.
Of course, within those data sets and particularly Twitter, "there's a massive amount of garbage" that Attensity's technology can clean out through spam filters, heuristics, taxonomies, domains and other techniques, said Sean Timmins, director of enterprise solutions.
However, with Q "the goal is not to target individual people," Timmins said. "We're not an ad engine."
Q users can simply put in topic phrases and the system will start harvesting conversations that seem to be related, and presenting the data in a visual format.
If a Q user spots something interesting -- say, a sudden burst of social media messages that are critical of a company's product -- that data set can be moved into another Attensity tool, Analyze, for deeper, root-cause analysis. Attensity has created a workflow that makes the data migration process simple, according to Timmins.
There are at least a dozen components to Attensity's NLP engine that examine various parts and types of speech, such as entities, voices, topics and intent, Timmins said. The system can also be tweaked to properly handle entities with multiple meanings, such as by separating the McDonald's restaurant chain from a local farm called McDonald's, he added.
Q will be available as an on-demand service later this month.
There are many vendors in the NLP market, all of which are scrambling to develop the best algorithms, filters and presentation tools to make sense of the Web's ever-larger pool of social media data.
Recently, NLP vendors have made progress in doing the same things as before, but for additional languages, said analyst Curt Monash of Monash Research. They've also gotten better at parsing "shorter messages with more questionable spelling and grammar," such as Twitter posts, he added.
However, NLP vendors haven't made much progress in deepening their engines' understanding of language, according to Monash.
What Q does should also be taken into context, he said.
"Analyzing the Twitter firehose is a bit of a fad," Monash said. "The largest consumer companies certainly gain value from it; other cases are less clear."