Conference Theme

Generative AI

Generative AI agents (such as ChatGPT, Google's Gemini, CoPilot, Siri, etc.) are essentially question-answering systems that rely on knowledge text they are trained on or learning dynamically. These systems have limitations in the languages they are trained on, but they can extend to other languages through LLM-based translation systems. However, translation accuracy is a challenge, as localization may fail when adapting across different language structures and cultural contexts.

As Generative AI systems integrate into production lines, the need for robust evaluation metrics remains critical. This conference aims to explore this aspect in depth.

Machine Translation Systems

With the emergence of LLMs, developing generic machine translation support for any language has become easier with minimal data. However, accuracy remains a significant challenge. Many industries now deploy machine-translated content with disclaimers stating that only English is authoritative, which undermines linguistic diversity.

This conference will focus on evaluating third-party machine translation systems for various language pairs.

Automatic Speech Recognition (ASR) Systems

ASR systems transcribe speech into text following the linguistic and script standards of a language. While the process is straightforward for simple speech, complexities arise as content gets more intricate. This conference welcomes discussions, papers, posters, and demonstrations on evaluating ASR systems.

Text to Speech Systems

New TTS models have improved Indian language synthesis. However, evaluating TTS remains more subjective than objective. To achieve more natural and culturally accurate speech synthesis, we may require community-centric evaluation methods.

Optical Character Recognition in Indian Languages

OCR technology has been a crucial application in computer vision, closely tied to verbal languages. India is home to more than 13 widely used scripts, along with several evolving ones. This conference will explore the evaluation methodologies for OCR systems in the Indian linguistic landscape.