Skip to main content

Table 5 Effects of each feature subset on the final classification performance for Knowledge Type

From: Identification of research hypotheses and new knowledge from scientific literature

Feature Subset

Only This Feature

All Except This Feature

 

P

R

F1

P

R

F1

Constituency

—

—

—

0.815

0.727

0.763

Dependency

—

—

—

0.823

0.728

0.765

Parse Tree

0.428

0.281

0.340

0.823

0.730

0.776

Participant

0.383

0.252

0.243

0.831

0.740

0.776

Sentence

0.474

0.442

0.453

0.785

0.705

0.738

Lexical

0.592

0.449

0.478

0.794

0.722

0.754

Structural

0.558

0.495

0.517

0.791

0.665

0.709

All

0.823

0.725

0.764

0.823

0.725

0.764

  1. Results are only shown in cases where it was possible to produce a reliable model. The final row denotes the performance of the classifier when using all feature subsets
  2. Values in bold represent the best performing feature subset for each column