tf-idf of Babel: Using text mining to optimize automated processes in multilingual contexts

Minor EU reaches migrants via Facebook but drowns in irrelevant notifications. We build text-mining filters that surface the posts that matter.

Exploratory AnalysisModeling Process

Status: Finished
Project Period: February 2019 – November 2019
Outputs: Blogpost
Partner: Minor Kontor

Minor – Project Office for Education and Research is supported by an interdisciplinary and intercultural team. Diverse, committed, and scientifically grounded, Minor works with and for socially marginalized groups. Minor’s members are active nationwide and transnationally in research, program and project consulting, continuing education, networking, and model project development. (adapted from https://minor-kontor.de/ueber-minor/)
Homepage

The project "Migrationsberatung 4.0" by Minor EU focused on supporting newly arrived migrants in their labor market integration. The organization proactively engaged with the target groups on social media, particularly in relevant Facebook groups based on migrants' countries of origin. To streamline their work, the Minor team set up email notifications for new posts but faced the challenge of manually filtering out the bulk of notifications for irrelevant content. The CorrelAid team used R and Python to assist Minor by creating an interactive tool for post labeling, improving email Outlook filters based on labeled data, providing descriptive statistics for reporting purposes, and conducting experiments like topic modeling and sentiment analysis. The project resulted in time savings, better identification of relevant posts, and an improved report for funders. The project emphasized the importance of clear project goals, investment of time, and quality data for successful data projects.