Terminology-Grounded Translation

Bridging the Gap Between Wikipedians and Scientists with Terminology-Aware Translation: A Case Study in Turkish

The pipeline of our proposed project.

This project addresses the gap between the growing volume of English-to-Turkish Wikipedia translations and the limited number of contributors, particularly in technical domains. By leveraging the collaborative efforts of academics in building terminology dictionaries, we propose a pipeline system to improve translation quality. Our focus is on fostering collaboration between academic and Wikipedia communities, creating datasets, and developing NLP models for terminology identification, retrieval, and terminology-aware translation. The goal is to encourage sustained contributions and enhance the overall quality of Turkish Wikipedia articles.

Project Goals

The project will focus on the following tasks:

  • High-Quality Parallel Corpora for Terminology-Aware Translation: We aim to generate 3,000 parallel English-Turkish sentences that include:
    1. English text annotated with technical terms,
    2. Links to the correct terminology entries in the database,
    3. Edited translations using the correct Turkish terminology.

  • Term Identification: Develop models to identify technical terms in a multilingual setup.

  • Term Linking: Build models to link identified terms to a terminology database. If the term is not present in the database, a notification system will alert domain experts to add it.

  • Terminology-Aware Translation: Develop post-editing and translation systems that adhere to the terminology database to ensure consistent and accurate translations.

Project Team

  • Asst. Prof. Gözde Gül Şahin (Principal Investigator)
  • Ali Gebeşçe (Master’s Student)
  • Ege Uğur Amasya (Intern)

Former Team Members

  • Mina Durhasan (Intern)

Funding

This project is funded by the Wikimedia Research Fund. The official project page can be found here.