All projects
Academic project
Transformer-Based Stack Overflow Question Classification
NLP Academic Project
PythonBERTNLPTokenisationTransformersText classification
Problem
Technical support platforms receive large volumes of unstructured questions. Automatically categorising those questions can support routing, search and knowledge management.
Technical approach
- Stack Overflow question text
- Text cleaning and preprocessing
- Transformer tokenisation
- BERT-based classification
- Class-level evaluation
- Practical routing interpretation
Business or societal relevance
The workflow demonstrates how unstructured enquiries can be routed to relevant teams, topics or knowledge resources.
Limitations
- Domain-specific labels
- Class imbalance
- Text length and ambiguity
- Academic rather than production deployment
Visuals and artefacts
Visual placeholder
BERT classification workflow
Asset →
/images/projects/bert-question-classification.pngTechnologies
PythonBERTNLPTokenisationTransformersText classification