Evaluation of the ontology was done informally using corpus coverage of terms.
DARE gives serious treatment to a massive corpus of prominent terms, only a few of which are evocative names for heavy rainstorms.
Table 3 The overview of the corrected CHEMDNER corpus in terms of the number of PubMed abstracts (#Articles), the number of CEMs (#CEMs), and the number of CEMs for each of the CEM classes in C = {SYSTEMATIC, IDENTIFIER, FORMULA, TRIVIAL, ABBREVIATION, FAMILY, MULTIPLE, NO CLASS} × means the resulting figure is unknown.
(3) We need to apply the method to a large corpus of hand-tagged terms, which we have already in the case of the Latin and Arabic texts in the Filemaker databases.
Now we are going to compare the student corpus and the expert corpus in terms of the percentage of verbal processes.
Figures 6 and 7 represent the distribution of the outbreak news reports in our corpus in terms of the publication date and affected country, respectively.
SketchEngine (Kilgarriff et al. 2014) was used to analyse the corpus in terms of concordances, collocations and word sketches and to extract examples from the corpus.
When applied to text corpora, farthest-first was outperformed by d-Confidence on the imbalanced corpus and presented similar performance on the balanced corpus, in terms of finding unknown classes, but with lower accuracy.
Lastly we hand-curated the remaining set of approximately 700 terms to remove non-scientifically relevant words yielding a final corpus of 490 evidence-derived terms.
Standard preprocessing [26] includes tokenization, stemming, unigram and bigram construction, removing terms which do not appear at least twice in the corpus, and construction of term frequency (TF) feature vectors.3 Additionally, Twitter-specific preprocessing [8, 13, 24] transforms usernames, hashtags, and collapses repetitive letters.
