Lianet Sepulveda Torres, ER3a

Location: Pangeanic, Spain

Project Title: Improving hybrid translation tool. Full-text search engine approach

Project Description: 

I started work at Pangeanic and the Expert Project in April 2016 as an Experienced Researcher (ER). Ever since I have been focused my research mainly on the integration of Natural Language Processing (NLP) techniques to improve the Translation Memory (TM) matching and retrieval of translations segments on hybrid Machine Translation (MT) scenarios.

I have combined language-independent and language-specific methods in order to optimize the reuse of the existing translations segments. More specifically, I have included linguistic knowledge such as segmentation rules, grammatical similarities and named-entity recognition between the segments.

I have worked jointly with my colleague at Pangeanic Alexander Raginsky, that has developed a fast and scalable TM system based on full-text search engine, whose goal is to overcome limitation of current solutions for TM storage and query. We have included the NLP techniques in the full-text search engine system to improve the hybrid MT workflow.

We have evaluated the fuzzy-match performance testing traditional MT Evaluation metrics. We also have analyzed the benefits of TM system in hybrid computer-aided translation scenarios. 

Research Interests: machine translation, evaluation, information retrieval and full-text search engine.