Table 2.

Examples of tokens eliminated and retained during the feature selection process on drug "warfarin"

r2	Examples of tokens
≈ 1 (perfect correlation)	only, when, however, well, another, same, results, other, observed, possible, different, since, even, could, though, occurring, therefore, high, although, also, both, so, result, appeared
≈ 0.90	restricted, controls, implicated, followed, diverse, stable, display, rate, plays, indicative, inhibit, typically, describe, excluded, terminal, excessive, largest, knowledge, employing, se
≈ 0.80	life, mature, loading, preincubation, problem, failure, binds, resolved, physiology, shock, signs, molecule, bind, elevations, chinese, usual, surface, aid, unit, accurate
≈ 0.70	intervention, stimulus, transition, closed, enable, bands, requiring, ester, nervous, sizes, electrophoresis, polymorphonuclear, aging, associations, accounts, practical, selective, choice, routine, attached
≈ 0.60	subset, undergoes, success, antagonist, artery, mr, depolarization, fields, suppression, precipitation, temperatures, records, mg2, adjustment, oxygen, picture, assembly, transcripts, encoded, organic
≈ 0.50	hydrogen, coated, glycol, antisense, coronary, adsorbed, histology, scan, formulation, foods, holding, resorption, gestational, filling, locus, memory, atrophy, ringer, prospectively, recruitment
≈ 0.40	diuretics, atrial, lysis, spinal, camp, bmax, vein, proteases, chelator, arachidonic, alzheimer, ascorbic, histamine, rhythm, ouabain, gas, preoperative, bladder, menopause, pertussis
≈ 0.33 (moderate correlation)	chromatographic, endothelin, relaxed, acceptable, stenosis, withdrawal, january, trypsin, oxidized, infiltration, forearm, et-1, enrolled, electrochemical, peroxidation, mothers, phosphodiesterase, cystic, compression, countries

r = Pearson's correlation coefficient

Examples of tokens retained by feature selection (in decreasing order of document frequencies)

Warfarin (8060), anticoagulation (2508), anticoagulant (1953), heparin (1699), thrombosis (1651), bleeding (1633), international (1324), venous (1238), aspirin (1231), fibrillation (1191), inr (1106), prothrombin (1035), thromboembolism (1017), anticoagulants (864), thromboembolic (860), coagulation (790), embolism (706), deep (698), prophylaxis (636), antithrombotic (606)

Examples of rare tokens eliminated by feature selection

vestige, bacteroides, ca-laurell, gd2, idaho, i475s, h2-blocker, depots, viic/viiam, left-hemispheric, p = .37, laboratory-developed, cardio, frames, thistle, thy1, homolog, videotapes, u-105665, five-years, cold-labeled, workups, fviiic

Examples of common tokens eliminated by feature selection