Engineering

From 0 to 180K+ Dialogue Clips: The Architecture Evolution of MandarinClips

A transparent breakdown of how MandarinClips evolved from raw video clipping to pragmatic intent graphs, the reality of running at a loss under heavy token costs, and why V4 is currently on hold.

What is MandarinClips? An Authentic Cinema Corpus for Spoken Chinese

An authentic audiovisual corpus indexing 180,000+ real film dialogue snippets by pragmatic functions, grammatical patterns, and cultural scenarios.