Named entity recognition (NER) – taking stock at ELRC’s Technology Workshop
Named entities play a key role in the performance of machine translation (MT) systems. However, sometimes, they can be challenging and complicate the machine translation process. While some named entities have to remain identical in the target language after the translation process has been run, others need to be localized in the target language. At the same time, some named entities may be inflected in the source language, but not in the target language. And in yet other cases, named entities must be anonymised to avoid GDPR-related issues.
On 5 July 2020, CrossLang organised the online workshop “Named Entity Recognition and Machine Translation”, aiming to shed some light on the importance of Named Entity Recognition (NER) and how to approach it in the context of MT. During the online event, speakers from academia, industry and the public sector gave valuable insights into the creation of multilingual NER tools and resources, including those for morphologically complex languages. In addition, the integration of such tools and resources in MT systems was illustrated and participants engaged in critical discussions. The workshop was attended by representatives from the European Commission, including from DGT and DG CNECT, and by experts from the research community. Exclusive impressions were shared during the workshop via the ELRC accounts on Twitter, LinkedIn and Facebook.
Following the presentations, an interactive panel session took place, allowing the participants and speakers to share their experiences and ideas. The talks and discussions clearly demonstrated that despite the advanced approaches and solutions, some issues related to NER are still open. In their closing statement, the workshop organisers declared: “We did not only hear answers but found new questions and challenges”. As such, the ELRC community looks forward to taking further steps towards solving these questions.