Automating text analytics for a research software platform

Background

Our client, a new market research software platform, wanted to make it quicker and easier to analyse the vast amounts of unstructured text, video and audio data the platform captures from research participants’ social media interactions.

The research hosted on the platform covers a variety of different topics, so the solution would need to be able to adapt.

Approach

We recommended using a combination of techniques, powered by pre-trained deep learning language models, to build the back-end text analytics.

​Our approach involved intelligently extracting multi-word keyphrases and single-word keywords using parts-of-speech tagging.

We also automatically analysed responses to extract the most representative comments using network analysis.

Outcome

The analytics successfully extracted important keyphrases from the responses, making it easy to see what research participants were talking about in their responses to each question.

Additionally, the analytics made it easy to get the context around those keyphrases by automatically extracting representative responses.

Overall, the analytics made it far quicker to extract the important points from respondents’ answers – reducing time spent on analysis.