peshkova - Fotolia
Falcon.io sells a social media marketing platform that covers basic needs for creating, planning and measuring content.
Falcon.io helps keeps its users abreast of how brands are perceived on its own channels, as well as how they come across on third-party websites, blogs and social media profiles, according to Justina Bubele, Director of Product Management at Falcon.io.
The platform not only measures how many mentions companies get online, but also uses sentiment analysis to measure brand perception, to see just what those mentions are actually saying. To measure sentiment, Falcon.io has turned to Lexalytics, a vendor of a sentiment and intent analysis platform.
Sentiment analysis is "a very important tool when dealing with any kind of PR crisis," Bubele said.
By combing through what's being said online during a crisis, like the PR crisis around the delayed Iowa caucus results right now, brands can monitor how well public relations efforts are working. They can measure if efforts are intensifying the crisis or making it worse and can then adjust strategies accordingly.
Using Lexalytics, Falcon.io can automatically perform sentiment analysis. Lexalytics, founded in 2003 and headquartered in Amherst, Mass., supports 22 different languages with dozens of machine learning models per language, said Paul Barba, chief scientist at Lexalytics.
With NLP, machine learning and text analytics, Lexalytics can scan text documents and extract meaning. It can be used for sentiment analysis, to automatically summarize large pieces of texts and to determine intent in pieces of writing.
Falcon.io, founded in 2010 and based in Copenhagen, Denmark, has used the platform for the last five years to help power its own social media marketing platform so it can automatically scan social media and web posts to determine sentiment for their clients' brands. It works well, but not perfectly, said Bubele.
The challenges with automatic sentiment analysis in marketing, Bubele said, don't necessarily lie with Lexalytics, but rather with the technology in general.
Justina Bubele Director of Product Management, Falcon.io
"Sentiment is definitely not an exact science," Bubele said. It's difficult to teach a machine sarcasm or slang, for example, especially over text.
News articles are particularly tricky to automatically perform sentiment analysis on, Bubele said, as they can contain numerous negative and positive statements about a company in one article. Even for a human, it can become difficult to assign a positive or a negative sentiment to one news story, she said.
"It's a bit of a hit or miss, but the general trend is quite clear," she said. Individual social media posts, news articles and blog posts can be difficult to automatically pull sentiment from, but, given enough posts or articles, a picture becomes clearer.
Over the last few years, the Lexalytics platform has gotten better at picking out sentiment, and it has added support for more languages, which has been positive for Falcon.io, which supports around 20 languages, said Bubele. Since partnering with Lexalytics, the company hasn't considered using another tool.