Most companies turn to listening tools if they want to discover evolving patterns in conversations. Such software platforms usually measure trends by counting frequencies of keywords. Unfortunately, keywords can be a very biased way to solve the information problem. An article might cover the reform of the Swiss pension scheme, without ever using any of the keywords directly related to the subject.
During the Swiss voting on #AV2020, September 2017, we monitored German-speaking media, and conversations on Twitter, to cluster the discussion and the news articles covering the voting. We used keyword frequency, a common method to discover trends or to build word maps. Unfortunately, the method is very biased, if your primary goal is to understand the meaning of large amounts of content pieces.
Articles or conversations (Tweets) can be about a certain subject without ever mentioning it in the text. In order to solve that problem and to have a valid base to understand and analyze huge amounts of content during certain events, we came up the idea to turn to “concepts” instead of keywords. A concept is on a higher level of abstraction and can be described by one or more words or phrases. A concept takes much more context into account, adding additional layers of information, helpful to evaluate, what people are discussing.
Today we’re delighted to tell you that our proposal for the “LNR Concept Mapper” has been selected for prototype funding by Google DNI.The LNR Concept Mapper will automatically identify and analyze, the concepts discussed in political discussions on Twitter and concepts as present in published articles online. The findings can be used to understand if there’s a gap between the topics discussed in publically visible conversations and the current media coverage, e.g. before and after elections. For the prototype, we will focus on conversations around political issues, not general market discussions. The domain knowledge we want to build is therefore focused on politics purely. It’s meant to be used in political journalism to get deeper insights into what’s really happening in a geographic market.