"Human-in-the-Loop Machine Learning"
The goal of the proposed research is the study of the ways machine and human intelligence can be combined to optimize solutions to problems that are handled using Machine Learning techniques, such as Active Learning, in the context of Situational Awareness. The objectives of the study include the:
- Review of current literature and proposals on the use of Deep Learning and Natural Language Processing of social media information related to crisis management.
- Survey on existing crisis datasets retrieved from social media platforms to be processed by Deep Learning-based Natural Language Processing (NLP) algorithms.
- Application of Transfer Learning techniques to finetune Transformer-based language models on crisis identification and Named Entity Recognition.
- Proposal (and evaluation) of a practical human-in-the-loop solution that can enhance the automated identification of emergency-related social media content and enable the active participation of users during the Machine Learning training and evaluation process.
