Table 3 Effectiveness of bm25 and the Lucene ranking algorithm combining evidence from spans with evidence from abstracts and articles.
From: Is searching full text more effective than searching abstracts?
MAP | ||
|---|---|---|
| Â | Ivory (bm25) | Ivory (Lucene) |
Span (max) | 0.240 | 0.206 |
Span (max) + Abstract | 0.257 (+7%)° | 0.216 (+5%)° |
Span (max) + Article | 0.257 (+7%)° | 0.262 (+27%)** |
P20 | ||
| Â | Ivory (bm25) | Ivory (Lucene) |
Span (max) | 0.357 | 0.332 |
Span (max) + Abstract | 0.382 (+7%)° | 0.349 (+5%)° |
Span (max) + Article | 0.343 (-4%)° | 0.404 (+22%)** |
IP@R50 | ||
| Â | Ivory (bm25) | Ivory (Lucene) |
Span (max) | 0.212 | 0.189 |
Span (max) + Abstract | 0.215 (+1%)° | 0.190 (+1%)° |
Span (max) + Article | 0.257 (+21%)° | 0.244 (+29%)** |