Microtalks


Track: Inside Track | IN9 |
Wednesday, October 30, 2024, 3:45pm – 4:30pm
Held in: Steinbeck 1 B&C
Presenters:
Zidian Guo - Middlebury Institute of International Studies at Monterey (MIIS) 
Sheriff Issaka - African Languages Lab 
Alon Lavie - Phrase 
Iti Sahai - Klaviyo 
George Zhao - VideoLocalize.com
Host: Daniel Goldschmidt

In this session, we will offer several quick and dynamic presentations covering topics and ideas too interesting to ignore!

The Impact of QA Processes on Video Dubbing Projects and Leveraging AI Dubbing for Improvement – George Zhao – VideoLocalize
In video localization projects, the traditional quality assurance (QA) process can create challenges, particularly for dubbing projects. This case study explores a project that involved localizing four videos with three characters into 22 languages. For a dubbing project, if the translation needs to be changed after the voice-over recording is completed, re-recording will be necessary.   The client recognized the expense and potential delays caused by re-recording and aimed to avoid it. Despite the client signed-off translations before the recording start, the traditional QA process still necessitated re-recording, leading to increased costs and project delays. This presentation will examine how this issue arose, the limitations of the traditional QA process, and how an AI dubbing solution can improve the workflow.
Takeaways: Understand the complexities of voice-over recording and dubbing; Recognize the limitations of traditional QA processes in video localization projects; Learn about a practical use case for AI dubbing technology.

Building a Comprehensive Localization Pipeline for 100+ Languages: The base implementation framework – Sheriff Issaka – African Languages Lab
This talk explores innovative approaches to localizing complex, cross-platform applications across 100+ languages. We’ll explore a comprehensive localization strategy, integrating cutting-edge machine translation, i18next for internationalization, and dynamic API data extraction from translation memories. Learn how to balance automation with human expertise to ensure cultural relevance and linguistic accuracy. Discover methods for overcoming challenges in managing diverse language expertise and developing a cross-cultural team. Gain insights into success metrics and the future of AI in bridging global linguistic divides. This talk offers a concise yet powerful glimpse into next-generation localization practices for multilingual, global-focused, AI-enhanced platforms.
Takeaways: Attendees will gain practical insights into implementing a scalable, AI-driven localization pipeline for complex platforms supporting 100+ languages. They will learn strategies for balancing machine translation with human expertise to ensure cultural relevance and linguistic accuracy across diverse markets. Participants will understand key performance indicators and best practices for measuring the success of large-scale localization efforts in AI-enhanced community platforms.

The Science of Evaluation: Automated quality measurement and scoring throughout the Localization Journey – Alon Lavie – Phrase
Enterprise grade localization needs a reliable way to measure translation quality in order to avoid inaccurate or culturally inappropriate translations. What combination of automations and scoring engines are required to score high volumes of content? And more importantly, do these systems actually work? In this microtalk, hear from an industry expert on what they utilize to assess translation quality and the learning curve to set up a system that works. Hear about what is available today, what should be developed in the future, and why that matters for speed, accuracy, and connectivity.
Takeaways: Understand the nuances of translation quality scoring; Hear about the industry standards for enterprise grade localization; Consider what you can implement at your organization to level up quality scoring.

From Idea to Global Impact: Unwrapping Role of International Product Managers – Iti Sahai – Klaviyo
This micro-talk offers a concise exploration of the pivotal role played by international product managers in driving global success. This micro talk delves into the multifaceted responsibilities of product managers, from conceptualizing innovative ideas to executing strategic plans that resonate with diverse markets worldwide. Through real-world examples and insights, participants gain a deeper understanding of the unique challenges and opportunities faced by international product managers as they navigate cultural nuances, market dynamics, and technological advancements to bring products to global prominence. Join us for an illuminating talk on the transformative impact of international product management in today’s interconnected world.
Takeaways: Global Vision, local execution: International product managers must possess a deep understanding of diverse cultures, markets, and consumer preferences to effectively navigate global landscapes and drive product success; Strategic Adaptability: Flexibility and adaptability are paramount as product managers must continuously refine strategies, iterate on product features, and respond to evolving market dynamics to maintain competitive advantage in international markets; Cross-Functional Collaboration: Effective collaboration with cross-functional teams, including marketing, engineering, and localization, is essential for successful product launches and sustained growth in international markets.

Understanding Confusing Prompts for ChatGPT in Localization Tasks – Zidian Guo – Middlebury Institute of International Studies at Monterey (MIIS)
When assigning ChatGPT a localization task, a confusing prompt can hinder accurate results. Localization adapts content linguistically and culturally for specific audiences. For ChatGPT, a confusion may occur due to a punctuation or a verb which is used correctly for humans but misleading for LLMs. For instance, saying “translate and edit” may mislead ChatGPT to translate and summarize. The presentation provides real examples and conclusions to help localizers avoid these mistakes in prompts and fine-tuning, ensuring the model understands the desired outcome and cultural nuances.
Takeaways: Identify vague language misdirection. Recognizing how unspecific terms or phrases can cause ChatGPT to produce generalized or off-mark translations; Learn how to provide mixed Instructions without confusing ChatGPT, using translations and QA tasks as examples