Automatic recognition of topics in free text answers
Led by Dr. Christoph Stanik, we develop our AI applications according to scientific standards and aim to publish the methods at conferences and in journals. But what is the story behind our AI applications? We would like to give you a short introduction to the feature, which is about the automatic recognition of frequently discussed topics.
Consider a questionnaire with free-text answers. While these can still be read and edited manually in small cohorts, surveys in large cohorts often have hundreds to thousands of responses. Occasionally, the questionnaires then have several items with free text fields. With such quantities, manual and representative processing of the free-text answers is extraordinarily laborious and time-consuming, if not impossible.
Solution and procedure
evasys already offers the possibility to categorize free text answers manually. In SDI, we are working on a solution that automatically groups all items you select with free-text answers into semantically similar categories. To optimally achieve the goal, we are experimenting with proven methods and the latest research. To ensure the relevance of the results, we develop the approaches on anonymized data of first pilot partners and engage in regular feedback sessions.
You can freely choose which items with free text answers should be analyzed. The AI can handle very large amounts of data and show you which topics are addressed and how often. By clicking on a topic, you can explore the specific comments. Through this AI application, you can save a lot of time and focus on topics that are of interest to you and get a first and condensed impression of the free text comments. Unlike fixed categories, this allows you to identify emerging topics and react to them if necessary.
Overall, this allows open-ended question types to be used much more extensively, since the automatic evaluation as well as clustering and analysis make their extensive processing possible in the first place. The increased use of free-text answers allows you to evaluate much more complex issues in future evaluation projects.