Three of our researchers here at LaSTUS lab - Piotr Przybyła, Santiago Egea, and Nico Duran - recently participated in the AuTexTification (Automated Text Identification) challenge. This challenge tasks teams of researchers to develop models which can accurately distinguish automatically-generated from human written text across five domains ranging from legal documents to social media posts. Participants also have to work across two languages, English and Spanish, as well as covering a range of writing registers and degrees of narrative structure.

We are pleased to announce that our entry from UPF-TALN came top of the field out of 76 entries to win the English and Spanish first shared task! They also scooped up fourth place in the Spanish second shared task. They will present their work at the IberLEF@SEPLN workshop in Jaén later this year (26th September), so be sure to look out for the paper publication of their work.

Elsewhere, Euan McGill - in collaboration with Prof. Horacio Saggion, Santiago Egea, and Luis Chiruzzo from the Universidad de la República in Uruguay - is presenting their work at two computational linguistic workshops in the Nordic countries. At RESOURCEFUL@NoDaLiDa, which this year was hosted in the Faroe Islands, they presented a paper which proposes a part-of-speech tagging methodology for Spanish Sign Language and uses these tags to improve translation results between Spanish text and Spanish Sign Language glosses and vice versa.

The Nordic House, Tórshavn - venue of RESOURCEFUL 2023

In addition, later this month, Euan McGill will present his work on creating a novel corpus for British Sign Language based on data from the UK Parliament's live interpreted Sign Language feed. He will present it at AT4SSL at this year's Annual Conference of The European Association for Machine Translation in Tampere, Finland on the 15th June.