Table 3.
Dimension | Abstract, title, and keyword | Abstract alone | Title and keyword | Title alone | Keyword alone |
---|---|---|---|---|---|
Behavioral domain | 0.534 | 0.537 | 0.501 | 0.440 | 0.448 |
Cognitive paradigm class | 0.464 | 0.464 | 0.471 | 0.420 | 0.394 |
Instruction type | 0.534 | 0.538 | 0.498 | 0.488 | 0.456 |
Response modality | 0.745 | 0.744 | 0.731 | 0.710 | 0.694 |
Response type | 0.706 | 0.720 | 0.699 | 0.660 | 0.662 |
Stimulus modality | 0.815 | 0.814 | 0.794 | 0.770 | 0.805 |
Stimulus type | 0.496 | 0.478 | 0.470 | 0.410 | 0.430 |
Table presents F1-micro values (see text) for naive Bayes under the binary relevance transformation across the five corpora that vary the feature space: words from (1) abstracts, titles, and MeSH keywords; (2) words from abstract text alone; (3) words from both titles and MeSH keywords; (4) title words alone; (5) MeSH keywords alone. Highest F1-micro highlighted in boldface; this does not indicate statistical significance. See text for details.