Sourcerer 🔮 (until I find a better name) is a Machine-Learning powered language inferencing system at its core.
The system uses the
RoBERTa model fine-tuned on the MLNI corpus.
RoBERTa is a sentence transformer model that draws from
Google's BERT transformer architecture.
The problem that is solved under the hood is a flavor of
Natural Language Inferencing (NLI) to find the appropiate news source match.
Since the software is in development and testing stage, the rest of the system architecture is subject to change. Currently working to improve latency and parallelization.