Reproduced Anthropic’s Contextual Retrieval (contextual embeddings + contextual BM25), then designed systematic comparative experiments across structured and government-document datasets. This work is the basis of my first empirical research paper.
Experiments
| Phase | Dataset | Content |
|---|---|---|
| 1 | Canteen menu (structured lists) | Baseline vs CR vs Jieba+KG |
| 2 | Flood-prevention plans (gov docs) | Baseline vs CR vs Deep KG |
| 3 | Same, OneKE-13B + OpenKG | Knowledge-graph schema optimization |
Core finding: CR is a double-edged sword across data types; knowledge graphs still hit real bottlenecks under current conditions.