معرفی
David M. Howcroft is an Advanced Research Fellow in Natural Language Generation at the School of Natural and Computing Sciences, University of Aberdeen. He previously held research fellowships at Edinburgh Napier University, Heriot-Watt University, and Saarland University, contributing to major NLP projects including ASICA, NLG for Low-Resource Domains, and Madrigal. His work bridges computational linguistics, psycholinguistics, and statistical modeling.
His research focuses on natural language generation, particularly in low-resource settings. He develops machine learning methods for data-to-text generation, creates novel corpora (e.g., for Scottish Gaelic), and improves human evaluation practices via crowdsourcing and rigorous statistical analysis. A key interest is the application of Bayesian nonparametrics and ordinal mixed-effects models to better understand and evaluate generated text. He also explores readability, referring expressions, and AI planning for rule-based NLG systems.
His recent publications reveal a strong trend toward methodological rigor and inclusivity in NLP. He advocates for better evaluation standards, transparency in metric usage, and participatory design in NLP research. His work spans corpus development, evaluation methodology, low-resource language support, and human-centered NLP. He has led efforts to create datasets for under-resourced languages and to standardize best practices in human assessment.
He has received small grant funding for projects such as Scottish Gaelic Generation for Exhibits and has contributed to software tools for data collection and evaluation. He mentors and collaborates widely, though no formal advisees are listed. He has developed backend systems and Android apps for healthcare applications, notably in melanoma patient support via the ASICA project.
He is actively involved in research labs and teams including:
- ASICA Project Team (University of Aberdeen)
- NLG for Low-Resource Domains (Edinburgh Napier University)
- Madrigal Project (Heriot-Watt University)
- SFB 1102 Project A4 (Saarland University)
- Language Science and Technology (LSV, Saarland University)
His scientific contributions are widely disseminated through top-tier venues such as ACL, EMNLP, and INLG. He maintains an active online presence with tutorials and technical blog posts on tools like OpenCCG and Treex.
حوزههای پژوهشی
