News from the 45th Translating and the Computer Conference - Asling TC45

Last week, multilingual generative AI was the focus at Asling Translating-and-the-Computer.

The 45th AsLing conference brought together over 200 participants from EU institutions, academia, language technology, service providers and corporate businesses at the European Convention Centre in Luxembourg. Lively discussion centred around how genAI will deliver gains and productivity, changing the landscape for people, their roles and tools.

Helena Moniz highlighted the major ethical challenges of genAI in her opening keynote. Vilelmini Sosoni talked about changing roles with the rise of language flow architects, language solution experts, enablers of digital transformation, prompt engineers, AI deployment engineers, data analysis, for example. The need to raise AI-literacy, mitigate risks and retain the humantouch was made. Linguistic roles may focus on domain/terminology expertise, quality estimation as well as post-editing, etc.

Hybrid solutions and use cases integrating genAI and Large Language Models (LLM) with Computer-Assisted Translation (CAT tools) were on show:

• Enhancing Neural Machine Translation
• Improving linguistic error detection, accuracy post-editing
• Reducing time-consuming formatting such as tag placement
• Facilitating brevity customization (for text length restrictions)
• Automating terminology extraction, definition and context generation
• Assisting meaningful quality predictions and confidence scores

Early game-changing AI-enhanced applications set to improve translation shown by STAR included (1) Terminology extraction, domain recognition, translation, definition proposal as well as (2) Quality confidence scoring, blended with the COMET metric and post-editing distance calculations.

CAT tools were predicted to evolve into agile, flexible, modular and API-driven hybrid architectures with AI capabilities to orchestrate language resources, processes and teams. Translation management will make full use of prompt engineering in custom workflows to gain maximum targeted value from AI.

GenAI almost eclipsed advances in Neural Machine Translation. The notable exception came during a tour of the European Parliament and keynote address by Valter Mavrič. Mr Mavrič reported on initial tests blending Translation Memories, custom CAT tools and Machine Translation (MT) with their ‘human-in-the-centre’ design across 24 languages.

Impressive multimedia localisation projects were shown including live speech to text with machine translation and custom subtitling environments.

Contact STAR today to benefit from latest AI-led advances. STAR – a leader in multilingual information technologies and services –  is a long-term sponsor of AsLing sharing decades of experience shaping language technologies and services to help today’s language professional.