GenAI in Localization III - Prompt Workshop

Best Techniques of 2023


This workshop recaps everything that we’ve learned in course of 2023 about large language models (LLMs) and the application of generative AI in translation and localization. We will present detailed plans for specific process enhancements, such as automated language quality assurance, automated post-editing (APE), subtitle translations, terminology extraction, localization testing, and document processing. 


For localization professionals who are new to prompt writing, we offer guidance on applying GenAI, enabling you to design and structure your initial prompts in collaboration with seasoned professionals. Study by doing!


For participants with engineering backgrounds, we have prepared a series of challenges to hone skills in retrieval-augmented generation (RAG) and chain-of-thought prompts, techniques to minimize hallucinations and enhance reliability.

This edition of the workshop provides access to models in Amazon Bedrock, including Cohere, Claude, Jurassic, and OpenAI’s GPT-4 Vision model, which can be used for automating UI testing and PDF processing.


Workshop Tasks

  • Automatic Editing Improve machine translation, reduce human work
  • LQA Identify and label translation errors automatically
  • Source Quality Fix spelling errors, improve readability and reduce biases
  • MT Quality Estimation Determine whether automated translation quality is sufficient for publication
  • Human edits classifier Detect translation edits and classify them into necessary and unnecessary to guide editors
  • Subtitles Generate subtitles taking the context into consideration
  • Language Variants Convert UK to US English
  • Support Chatbot Design a chatbot that responds with facts from the user manual
  • Benchmark LLMs Devise a method to quickly compare model performance in a specific task

Workshop Model Access

GPT-4 Turbo

The latest OpenAI model combining better cost and performance

GPT-4 Vision

Experimental OpenAI model able to “see” images and describe them

Claude 2

Antropic’s model with 100k tokens of context


Rapidly growing competitor to OpenAI


Competitor to OpenAI optimized for business usage

Techniques Covered

  1. Chain-of-Thought Prompts – instruct the model to self-check, often breaking up the task into steps.
  2. Few-shot prompting – provide several examples of task execution to the model
  3. Retrieval-augmented generation (RAG) – feed the model examples dynamically from a factual database
  4. Batching – group together similar tasks to reduce prompt costs 
  5. Temperature – control the variability of model output
  6. EmotionPrompt – using politeness and emphasising  task importance to get better output.

Why you should join

Agenor Hofmann-Delbor
Agenor Hofmann-DelborHead of Localization at 3di Information Solutions
Read More
A great teamwork and a hands-on experience of what GTPs can help with in the translation industry. I can’t wait for the next prompting workshop.
Bridget Hylak
Bridget Hylak Senior Consultant and Founder at TongueTek Language Consulting
Read More
Workshop was enlightening, challenging yet not taxing, an interactive meeting of some pretty brilliant industry minds seeking to uplevel their skills in this brave, new chapter. If you have the chance, BE THERE.
Mark Jones
Mark JonesHead of Technology at Comtec Translations
Read More
The workshop was a great opportunity to learn ChatGPT.

Workshop plan

January XX, 16:00 pm – 18:00 pm 

Introductory Session

Outline of the workshop, task presentation, practical demonstration how to write prompts and how to iterate and improve them. A basic overview of retrieval and chain-of-thought prompts. Environment to test GPT-4 vision model capabilities.

Konstantin Dranch

Co-founder at Custom.MT

Katell Jentreau

Globalization Leader

18:00 pm – 19:00 pm 

Optional QA and Networking

Allocating an extra hour to questions and answers, as well as freeform online networking with localization peers.

December 9 – 13


January XX, 16:00 pm – 18:00 pm (CET)

Advanced Prompt Engineering

  • Chain-of-Thought Prompts
  • Focus on Retrieval-Augmented Generation (RAG)
  • Overview of model fine-tuning for localization
  • Completed prompt showcase

Dominik Wever


5:45 pm – 6:00 pm 

Wrap up & Takeaways

Save your spot today! 

Number of participants are limited.

Workshop Fee


you get: