Thomas Compton

WORK HISTORY

EDUCATION

ABOUT THOMAS COMPTON

Interdisciplinary NLP Researcher | Bridging social science, humanities, and computer science to build rigorous, interpretable, and socially aware natural language processing systems.https://comptonportfolio.netlify.appI apply NLP to historical texts, policy documents, and academic corpora, with a focus on topic modelling, OCR pipelines, and RAG-based analysis. My work emphasizes evaluation, reproducibility, and open science—because understanding how well models work is as important as getting them to work at all.Technical Focus:Topic Model Evaluation (BERTopic, LDA, Gini, PUV)OCR + LLM Accuracy Benchmarking (Tesseract, Gemini, Qwen, Deepseek)FAISS, Sentence Transformers, RAG, and LLM IntegrationPython, PyTorch, Hugging Face, spaCy, Gradio, AWSResearch & Education:PhD Researcher at the University of York (ESRC 1+3), exploring historical continuities in community unionism using NLP. Proven experience in undergraduate teaching, public engagement, and interdisciplinary collaboration. Open Source & Communication:I believe in transparent, reusable research. All projects are open-source and designed for accessibility—whether through Google Colab notebooks, Hugging Face demos, or educational repositories. Selected Work:OCR Accuracy Evaluation : Benchmarking 7 OCR engines + LLMs (WER/CER)BERTopic Evaluation Framework : Metrics for stability, coverage, and inequalitygram-2-topic : Interpretable hybrid topic modellingSamuel Smiles DA : Teaching historical text analysis with embeddingsolistic Evaluations of Topic Models (arXiv:25•••••64)Passionate about teaching, responsible AI, and democratizing access to NLP tools—especially in education, policy, and the digital humanities. Let’s connect if you’re working in applied NLP, digital history, policy analytics, or research engineering.