Table 3. Performances metrics of the classification to predict the Main Classifier (left) and the Main Classifier + Modulars (right) classes (relevant/non-relevant), in terms of precision, recall and F-score.

	Main classifier relevance			Main classifier + Modulators relevance
	Classifier: logistic regression, pre-processing: stop-words removal, unigrams			Classifier: random forest, pre-processing: lemmatised text, no stop-words removal, unigrams and bigrams
	Precision	Recall	F-score	Precision	Recall	F-score
Relevant	0.57	0.44	0.50	0.62	0.71	0.67
Not relevant	0.73	0.82	0.77	0.59	0.49	0.53
Weighted average	0.68	0.69	0.68	0.61	0.61	0.61