PopTip Launches Zipline, An Analytics Tool For Finding Common Phrases In Twitter Conversations

Social polling company PopTip has a new product for marketers, journalists, and others who want to know what people are really saying on Twitter about a given topic.

Analytics seems like a logical extension of PopTip’s existing services — it’s all about helping customers understand how people feel and what they’re talking about.

In the case of Zipline (that’s the new product), there are already a number of services for tracking brand mentions, sentiment, or related topics, but this product focuses on phrases. To cite an example from PopTip’s blog post, a company could identify “Grand Theft Auto V” as an important topic (in the parlance of Zipline, a “classifier”), and Zipline would show that people who are talking about GTA V are using phrases like “download gta v” and “the day girlfriends lose their friends”.

The lists of phrases may be useful in and of themselves, particularly paired with the “PopTip score” — founder Kelsey Falter told me that the score uses a bit of “secret sauce” to identify the phrases that aren’t just being said the most, but are also relevant because of, say, who’s actually doing the tweeting. One use case that was close to home for me was a journalist tracking reaction to the launch of a new product — instead of just searching for that product on Twitter and seeing a few results, or maybe using sentiment analysis to see that the results are mostly positive or mostly negative, they could see the actual things that many people are actually saying.

At the same time, the phrases can also be used to access other features. Customers can focus on a specific phrase and see all the actual tweets, or they can save a list of everyone who used the phrase. They can also set up alerts to get notified when specific words are used, or when there’s a big change in the conversation.

When Falter was giving me a demo of Zipline, I noticed that there were some very similar phrases in the lists (also known as ziplines) — for example, under Apple, there were a couple that were basically variations of “iPad Mini retina display”. Falter said the underlying natural language processing (NLP) technology does some combining of phrases, but there’s a tradeoff.

“Our focus is on speed, but we’re also going to continue to focus on quality in terms of sub categorization and essentially clustering around common phrases,” she said.

In the future, Falter said Zipline could dynamically create more ziplines that are relevant based on the conversations. So once you’ve entered your initial topics, you wouldn’t have to continue entering new info — you’d just keep an eye on the conversations and let Zipline do the rest.