SignTrace offers a natural‑language front‑end to a Chinese sign‑language dictionary, tackling the long‑standing reverse‑lookup problem where learners recall a sign's movement but lack its meaning. The architecture comprises five components: LLM‑driven dictionary enrichment, action extraction, dictionary‑style rewriting, seven‑channel retrieval, and candidate reranking. The dictionary contains 6,699 entries, enabling multi‑view matching of motion descriptions.
On a benchmark of 500 movement‑description queries derived from the dictionary, the system achieves 94.0% Hit@1, 97.4% Hit@9, and a mean reciprocal rank of 0.9540. Reranking lifts Hit@1 from 71.8% to 94.0%, demonstrating the impact of enriched entry descriptions.
With six concurrent queries, the median processing time is 13.37 seconds. SignTrace has been deployed for user trials and received positive informal feedback. By linking everyday motion descriptions to documented signs and meanings, it provides a practical tool for identifying unfamiliar signs.
Review