TAUS Massively Multilingual AI Conference Albuquerque

Practical applications, writing real prompts, including data preparation, evaluation methods, and benchmarking results.

An overview of the technical setup, financial, and performance considerations for various NLP tasks. When are LLMs suitable?

A novel approach to model fine-tuning.

Kicking off the workshop, we’ll dive into the mechanics of quality estimation (QE). What is quality estimation exactly and what function does it have in your workflow? What are the best use cases and benefits of QE? We’ll also take a look at the available models and scoring metrics. And discuss the reliability and limitations of QE models.

Exploring the practical applications of QE and its impact on transforming translation workflows. We'll examine various use cases where QE enhances traditional translation processes. Additionally, we'll consider how Automated Post Editing (APE) can be integrated with QE to further optimize translation quality and efficiency.

Time for some action. We start with a real-time demonstration of quality estimation with an MT workflow, using the TAUS Estimate API. During this demonstration we will go through several example scenarios to illustrate how you can utilize QE to improve your translation quality. We'll discuss how to set up your workflow, analyze and interpret the quality scores generated by the model, and test prompts for APE to further refine the process.

So where do LLMs fit in when it comes to quality estimation? In this session we’ll discuss the current and potential applications of LLMs in enhancing QE models, including the benefits and limitations. We’ll compare “traditional” QE models with LLM-enhanced models and look at any future improvements of QE models based on LLMs.

In the final session of the workshop we’ll provide practical guidance on how to get started with implementing QE into your workflow. Questions will be answered on:

• How to integrate QE into your existing workflows

• How to customize QE models to your specific needs

• What kind of data is required to train QE models

• How to ensure reliability of the QE models

Engage in open discussion, ask questions and exchange insights with presenters and fellow participants to further explore the nuances of QE and APE.

Hotel Albuquerque at Old Town