Lucia Specia
Lucia Specia researches data-driven natural language processing, with interests in multilingual and multimodal models. With Gustavo Paetzold and Carolina Scarton, she developed QuEst++, an open-source tool for predicting translation quality at word, sentence and document levels. Her research also includes image captioning and text adaptation. She received a professorial appointment at Dublin City University in 2020, worked within ADAPT, and had affiliations with Imperial and Sheffield. Her work connects language generation with the practical problem of identifying outputs that need revision.
- Now at
- Imperial College London
- Works
- 362
- Citations
- 9,356
- h-index
- 48
Publication numbers from OpenAlex, as of 14 October 2026.
Timeline
-
2015
Estimate translation quality
Specia, Gustavo Paetzold and Carolina Scarton presented QuEst++ at ACL-IJCNLP. The open-source tool predicts the quality of translated words, sentences and documents, providing a basis for finding text that may need human revision.
-
2015
Connect predictions at different levels
The system supports a pipeline in which word-level predictions can become features for sentence-level models. Its architecture combines feature extraction with machine learning, making different quality-estimation tasks available through a shared framework.
-
2020
An international NLP appointment
DCU announced her professorial appointment on 17 April 2020, and she worked within its ADAPT center and had affiliations with Imperial and Sheffield. Her research includes multilingual and multimodal models at the intersection of language and vision.
-
As of October 2026
Use context across language tasks
Her research includes machine translation, image captioning, quality estimation and text adaptation as applications. An ERC project focuses on multimodal context for translation.
Sources and credits
Photo: www.dcu.ie; creator not stated. License: Not stated by source. Source: dcu.ie. Cropped to a square and resized. Ask us to remove or replace this photo.