Table 2.
Tuning of the search engine regarding the traditional technical survey task.
| Strategy | Tuning | P0 | MAP |
|---|---|---|---|
| Impact of the description field | Description | 15.87% | 1.49% |
| No description | 19.51% | 1.48% | |
| Impact of the metadata field | No metadata | 30.92% | 1.74% |
| Metadata on TAC | 30.30% | 1.60% | |
| Metadata on TACD | 19.51% | 1.48% | |
| Impact of the weighting model | PL2 | 19.51% | 1.48% |
| BM25 | 20.05% | 1.59% | |
| Impact of the co-citation network | Without re-ranking | 20.05% | 1.59% |
| With re-ranking | 21.24% | 1.60% | |
| Impact of the use of IPC codes | No IPC codes | 21.24% | 1.60% |
| IPC codes | 23.28% | 1.62% | |
Results in bold are the most effective and are therefore selected to perform further experiments. P0 indicates the top-precision and MAP represents the mean average precision.