October 15, 2020
As a part of ‘Co-creating Service Innovations in Europe’ - a project funded by the European Commission’s Horizon 2020 programme - we were interested in advancing the active shaping of service priorities by end users and their informal support networks. We aimed to engage citizens, especially groups often called ‘hard to reach’, in the collaborative design of public services. We were particularly looking for new ways of finding and analyzing data on how to create and deliver social services. In this context, one of the groups of people that we were most interested to hear from is the “marginalized youth” group. They often use the social service but it has proven difficult to attract them in workshops and they rarely fill-in surveys or answer feedback requests.
We identified a relevant discussion group at ylilauta.org, an anonymous discussion group that attracts marginalized people to partake in the discussion group ’Hikikomero’. Among the 76 000 messages posted in 2018-2019, there were more than 3 000 messages that referred to social service organizations. Such a large number of qualitative data is of course a very positive finding as it gives rich content for analysis. On the other hand, analyzing 3 000 messages is a real burden: text analysis requires coding and re-coding the data, in order to make sense of it all.
Therefore, we planned on using different machine learning and text analytics tools, such as Luuppi, which has an integrated search robot, machine teaching, message handling and visualization tools. Luuppi could have provided us with interesting results. However, we were on a tight schedule for the project and the tools required us to, by ourselves, teach the algorithm how to analyze the texts - which is pretty time consuming.
“DrAI saved me a couple of months and gave better results”
At this phase, we decided to utilize DrAI’s algorithm & full-service text analytics solution. DrAI is a Finnish startup that had just been launched a couple of months before and we were one of their first customers. The sleek part of DrAI is that it required no effort from our part in terms of teaching the machine. Also, DrAI did all the text, topic and sentiment analysis “heavy lifting” as well as testing the statistical meaningfulness of the findings. We were able and invited to assist in the process by utilizing our prior knowledge on interesting keywords. But, other than that, DrAI just delivered us a complete report of their analyses of the 3 000 messages. I can honestly say that their text analytics full-service saved me at least a couple of months of work. Moreover, I can say it gave us more conclusive results than what we could have achieved with other tools or unassisted text, topics and sentiment analyses. We were able to get all the statistical analysis on topics and sentiments that emerged from the data. This enabled us to go even deeper in the analysis, deep-diving into the negative messages and gain better understanding on how social services could be developed and meet the needs of the marginalized youth. In the end, we were able to submit our manuscript the very next month.
Results show that social media provide a window into the every-day life of socially withdrawn youths in a way that can be used for enriching the knowledge-base of service co-creation processes. With new text analytics services, getting insights from publically available data is more efficient, fast and easy than ever.
Turku University of Applied Sciences