Findings of the Shared Task on Speech Recognition for Vulnerable Individuals in Tamil

Abstract

This paper illustrates the overview of the shared task on automatic speech recognition in the Tamil language. In the shared task, spontaneous Tamil speech data gathered from elderly and transgender people was given for recognition and evaluation. These utterances were collected from people when they communicated in the public locations such as hospitals, markets, vegetable shop, etc. The speech corpus includes utterances of male, female, and transgender and was split into training and testing data. The given task was evaluated using WER (Word Error Rate). The participants used the transformer-based model for automatic speech recognition. Different results using different pre-trained transformer models are discussed in this overview paper.

Exploring the Efficacy and Potential of Large Language Models for Depression: A Systematic Review

“…As noted, the systematic review included 34 studies published between February 2019 and March 2024 that investigate the application of LLMs in various aspects of depression research (3,7,24,24–54). These studies encompassed a wide variety of sample sizes, ranging from as few as 25 to over 632,000, and utilized data types ranging from clinical interview transcriptions and electronic health records to user-generated content on social media platforms ( Table 1 ).…”