Text analytics: How to get market insights from your text data

Are you interested in learning how to use text analytics to turn your unstructured texts into actionable insights?

Using Gavagai’s text analytics software you’ll be delivering insights to non-technical stakeholders in minutes. Companies like Kantar Sifo and Ipsos use our tool Explorer for their text analysis gain insights from qualitative data sets.

1. Code your text data into specific themes

This is the first step in trying to understand what your texts are about. Letting artificial intelligence do it for you will save you hours.

Explorer uses Gavagai’s own artificial intelligence to help you find the themes, and then sorts them according to prevalence. It recognizes multi-word expressions like “San Francisco” or “front desk”, synonyms (like “room” and “suite”) or other related terms (like “Fisherman’s Wharf” and “Union Square”) out of the box. It’s a fast tool and will code 250 000 verbatims in minutes.

Explorer will then analyze your data and find the most common themes. The tool also identifies the language of each response, and allows you to filter all texts that are not in your chosen language. Then it gathers all responses (or verbatims) that match and keeps track of them.

Explorer identifies common themes in your text and recognizes multi-word expressions
such as “Breaking Bad” and “Arrested Development”.

2. Create a model for your analysis

The second step is to dig further into your topics and start making sense of the text. What topics are popular, and how do people feel about them?

In Explorer, settings can be made to your data set to make the topics and sentiments as relevant as possible for your business. We call this “modeling”. You can set your own numeric codes or make the topics more specific or detailed. You can also merge the topics together, group them for hierarchies, ignore irrelevant themes, or give them arbitrary labels. You can even add your own topics manually, if you insist on displaying a low frequency topic in your model.

Explorer also detects tonality beyond positive and negative – we have 9 sentiments and you can customize them in your model to suit your business. Cold pizza, for example, is negative for a pizza delivery company – but positive for a frozen pizza brand.

When you are done modeling your data set you can save the model and reuse it with new data sets. You can even create a model in one language (like English) and then generate sub models for each language that you have collected responses for. This allows you to do native coding of 46 languages – even if you only know English.

3. Get instant insights from the dashboard

When you’re done modeling your data set it’s time to get your insights. The Explorer dashboard visualises findings and insights from your project and helps you understand how to use the results from your text analysis.

The dashboard shows the most influential topics and focuses on what drives satisfaction and dissatisfaction using correlation statistics. 

Click the image to see a live dashboard example.

The dashboard has a bar chart, an occurrence grade matrix and a comparison graph:

  • The Important Topics bar chart shows the frequency of mentions of a certain topic.
  • The Occurrence-Grade Matrix shows how the topics relate to grade:
    • On the X-axis (horizontally), a topic closer to the right means many mentions. A placement to the left means few mentions.
    • On the Y-axis (vertically), a topic with a high placement correlates with positivity/satisfaction. Low placement indicates that the topic correlates with negativity/dissatisfaction.
  • With Comparison Graphs, you can compare any topic in the analysis project against the other by sentiment or by occurence. The two types of comparison graphs are:
    • Timeline graph to compare different topics over a period of time.
    • Bar Graph to compare topics by any defining tag, for example geo location, gender or age group.
Creating a Grouped Comparison graph.
Creating a Times Series Comparison graph.

Export your dataset

If you want to export your dataset, you can download it as a .csv or .xlsx (Excel) file. The export will use rows for each response, and columns for each topic, keyword, summary, sentiment etc. in the model. 

Integrate your dataset

With our API solution, you can easily connect your dataset with your own favorite tools such as Tableau or Qlik Sense. 

Analyse your own dataset with Explorer

Do you want to see hands-on how Explorer works for market researchers? We will demonstrate the tool using your own data set to make sure you get relevant insights.

Don’t miss out – book your personalized, free demo today.

Or learn more about Explorer.