Project !Translate

  • !Translate

risotto di zucca

Project !Trans (no-translate) is developed at UniBo's Department of Interpreting and Translation, within the framework of Alma Idea 2022, an initiative of Alma Mater Studiorum-Università di Bologna in agreement with the Italian National Recovery and Resilience Plan (PNRR).

!Translate aims at developing technology to spot texts for which even modern machine translation  technologies run short at due to terms/domains heavily rooted in a region/culture, causing valid translations to hardly exist in other languages (aka realia).

!Translate aims at developing technology to explain concepts across languages in order to increase comprehensibility under such an unfavourable scenario.

Objective

The main objective of !Translate is developing (semi-)automatic pipelines to explain complex or untranslatable terms, specific to the intangible Italian gastronomy heritage, to enhance the comprehension and evaluate the usability of texts in (incidental) learning contexts for Italian as L2.

Specific objectives

The project draws upon two specific objectives:

  1. Design and develop supervised models to assess the level of (cross-language) comprehensibility of a text, as well as the feasibility of its automatic translation.
  2. Design and develop supervised models for the retrieval of cross-language definitions to produce explanations of complex/specific/non-translatable terms from a culture/language.

Main results

We have carried out a number of experiments with deep learning models to assess the suitability of machine translating diverse types of texts. Our models aim at predicting whether a translation request should be sent to an automatic model (neural machine translation) or its difficulty leads to send it to either a professional or a non-professional human translator. The results were presented at the 24th Annual Conference of the European Association for Machine Translation (EAMT 2023) in Tampere, Finland.

We also worked on models for the identification of realia words and expressionsterms and for the automatic extraction of their definitions from encyclopaedic texts with deep learning technology, be it in Italian or Englis. This research has been presented at Recent Advances in Natural Language Processing (RANLP 2023) in Varna, Bulgaria.

Publications

  1. Federico Garcea, Margherita Martinelli, Maja Milicević Petrović, and Alberto Barrón-Cedeño. 2023. !Translate: When You Cannot Cook Up a Translation, Explain. In Proceedings of the 14th International Conference on Recent Advances in Natural Language Processing (RANLP 2023), pages 392–398, Varna, Bulgaria. INCOMA Ltd., Shoumen, Bulgaria.
  2. Francesco Fernicola, Silvia Bernardini, Federico Garcea, Adriano Ferraresi, and Alberto Barrón-Cedeño. 2023. Return to the Source: Assessing Machine Translation Suitability. In Proceedings of the 24th Annual Conference of the European Association for Machine Translation (EAMT2023), pages 79–89, Tampere, Finland. EAMT