Ruixing Liang bio photo

Email

LinkedIn

Github

Google Scholar

ResearchGate

Zhihu

(†: equal contribution)

Published Papers


Contributing Published paper

Ongoing Projects

Conversational Audio Speech of Peer Engagement Recordings (CASPER) introduces a novel dataset designed to advance AI research in naturalistic dialogue systems. CASPER provides a rich resource of spontaneous and casual conversations, specifically tailored to enhance the training of AI models in understanding semantic context, maintaining topical coherence, and executing effective turn-taking. Collected through a cross-platform web application and open-sourced in collaboration with universities, the dataset encompasses a diverse range of speakers and scenarios. The pilot phase of CASPER has 100 hours of transcribed and human validated English conversational audio amongst more than 100 users which in next phase aims to amass 100k hours, with plans to extend to other languages and global university partners. To facilitate the research on this corpus, we provide results of several benchmark models. Comparative results show that for this dataset, our current models are not able to provide significant improvement by introducing background knowledge/topic. Therefore, the proposed dataset should be a good benchmark for further research to evaluate the validity and naturalness of multi-turn conversation systems. This dataset addresses critical gaps in existing resources, which predominantly focus on text or formal speech and often lack the flexibility to adapt to the diverse linguistic landscape. By facilitating scalable data collection for casual, expressive, and conversational speech across various languages, CASPER stands to significantly enhance the development of AI systems capable of engaging in more human-like interactions.

-More projects are work in progress stay tuned in next few months for more details



Early Projects

Update: 29th Sep 2024