TASK 1: USE THE WEB Two groups jumped ahead and tried thesauruses, which worked out well. The next batch used variations of MI. The MI batch came next, then some novel approaches. Quite a range of ideas! -------------------------------------------------- GROUP SCORE APPROACH XIII 80.000000 thesaurus XVI 77.000000 analyzed search results of google XII 73.500000 thesaurus VII 73.500000 MI with negative evidence and threshold XV 71.500000 penalizes next to each other XIV 70.500000 regression on collection of vars V 70.000000 MI threshold + filters VIII 69.000000 Check if (A NEAR B)/(A AND B), threshold IV 68.500000 use "near" or "and" depending on frequency II 67.000000 MI plus filter, vary with absolute count VIII 66.500000 (A NEAR B)/(A AND B), threshold, secondary POS filter III 63.500000 MI with threshold XVII 63.000000 ? I 62.500000 modified MI and threshold XI 62.000000 MI (without the log) with a threshold IX 61.500000 MI with threshold X 59.500000 MI with threshold X 58.500000 MI with threshhold XVI 55.000000 count "ors"/"ands" and keep top 90 VI 54.000000 % of nearby words in common and threshold XVI 43.500000 count "ands" and keep top 87 TASK 2: USE A THESAURUS Results went up a lot from last time. There was some variation (using overview or not, stemming or not), but the basic approaches seemed pretty similar to each other. Nonetheless, the results differed a lot, even among approaches that sounded the same to me, so the groups must have been doing *something* differently! -------------------------------------------------- GROUP SCORE APPROACH II 86.0 WN: syns, look for common words in the list of "senses" XVII 85.0 thesaurus with MI XIII 80.0 online thesaurus (from last time) IV 78.0 WN: syns, look for common words in syns. Chop ed&s VII 78.0 WN: Intersections between synsets VI 77.5 WN: compare homonym, synonym, antonym lists V 77.0 WN: search for A, its forms and syns in B's list, swap XII 77.0 WN+web thesaurus XV 77.0 WN: intersections of syn sets XVI 76.5 WN: analyzed results IX 72.5 WN: trimming suffixes from words, etc XIV 72.5 WN: syns plus removal of non-matching parts of speech XII 71.5 WN: solo (no combination, as above) I 71.0 WN: hypernyms, syns, similars, etc. for POS, stem III 70.0 WN: looked for A (stem s&ed) in B's description, swap VIII 67.5 WN: intersected 'relevant' words (substrings) X 64.0 WN: syns and overview XI 61.0 WN: syns (using online commands)