Algemeen: http://pad.constantvzw.org/p/certainty Logbook: http://pad.constantvzw.org/p/certainty_logbook Questions: http://pad.constantvzw.org/p/certainty_questions Modality paper notes: http://pad.constantvzw.org/public_pad/certainty_notes_Modality-and-Negation modality.py close reading --------------------------------------------------------- start sentence: This automated decision-making is designed to take the human out of the equation. n = 0.0 (float) m = 0 (integer) --------------------------------------------------------- epistemic_weasiling dict sentence: This automated decision-making is popular belief and of course designed to take the human out of the equation. prints: phrase: popular belief k: -0.75 n: -0.75 m: 2 phrase: of course k: 1.0 n: 0.25 m: 4 >>> -0.75 + 1.0 / 2 = 0.25 >>> 2 + 2 = 4 --------------------------------------------------------- MODIFIER dict sentence = 'This strongly hypothesized decision-making is popular belief and of course designed to take the human out of the equation.' sentence = 'This strongly hypothesize decision-making is popular belief and of course designed to take the human out of the equation.' <<< hypothesized --> hypothesize <<< there is no lemmatization on word level in Word(), so hypothesized is not detected in the epistemic prints: *said i+word+type+weight: 2 Word(u'said/VBD') VB 2 *MODIFIER DETECTED* word: Word(u'said/VBD') weight: 3 (normally a verb has a +2 weight, but +1 from the MODIFIER, as 'hypothesized' is followed by 'strongly', a MODIFIER) k: 0.25 n: 28.0 m: 231 --------------------------------------------------------- epistemic_type dict sentence = 'This strongly hypothesize decision-making is popular belief and of course designed to take the human out of the equation.' i+word+type+weight: 6 Word(u'belief/NN') NN 1 k: -0.25 n: 27.75 m: 232 >>> belief gets another k here as a NN >>> 'popular belief' is already rated in the epictemic_weaseling dict >>> 'belief' is rated twice >>> all the epistemic_weasling phrases that appear in the other dict's, are rated twice --------------------------------------------------------- reverse score for negated terms sentence = 'This strongly said decision-making is popular belief and of course designed to wish and not wish the human out of the equation.' code: if i > 0 and s(S[i-1]) in ("not", "n't", "never", "without"): *k = -k * 0.5 prints: i+word+type+weight: 12 Word(u'wish/VB') VB 2 former n : 0.75 weight : 2 k : -0.25 weight * k: -0.5 new n : 0.25 m : 10 i+word+type+weight: 14 Word(u'not/RB') RB 2 i+word+type+weight: 15 Word(u'wish/VB') VB 2 negated k : 0.125 former n : 0.25 weight : 2 k : 0.125 weight * k: 0.25 new n : 0.5 m : 12 --------------------------------------------------------- all words that appear in one of the epistemic dicts how is m constructed? >>> an addition of all weights >>> on what basis are the different weights choosen? >>> MD = 4, VB = 2, NN = 1, epistemic_weaseling = 2 *epictemic_weaseling dicts >>> m += 2 phrase: popular belief k: -0.75 n: -0.75 m: 2 phrase: of course k: 1.0 n: 0.25 m: 4 *epistemic_type dicts >>> m += weight Word(u'said/VBD') VB 2 *MODIFIER DETECTED* weight: 3 --- former n : 0.25 weight : 3 k : 0.25 weight * k: 0.75 new n : 1.0 m : 7 Word(u'belief/NN') NN 1 former n : 1.0 weight : 1 k : -0.25 weight * k: -0.25 new n : 0.75 m : 8 Word(u'wish/VB') VB 2 former n : 0.75 weight : 2 k : -0.25 weight * k: -0.5 new n : 0.25 m : 10 Word(u'not/RB') RB 2 Word(u'wish/VB') VB 2 negated k : 0.125 former n : 0.25 weight : 2 k : 0.125 weight * k: 0.25 new n : 0.5 m : 12 >>> m is a sort of weight? a sum of all the weights? ------------------------------------------------------- numbers are heavy (or, and then the numbers are coming) sentence = 'This strongly said decision-making is popular belief and of course designed to wish and not wish the human out of the two equations.' >>> what is the reference of this assumption? Word(u'two/CD') WP 1 CD former n : 0.5 new n : 1.25 m : 13 ------------------------------------------------------- modality rate calculation code: if m == 0: print 1.0 # No modal verbs/adverbs used, so statement must be true. print 'modality rate:', max(-1.0, min(n / (m or 1), +1.0)) prints: modality rate: 0.0961538461538 n = 1.25 m = 13 min((1.25 / 13), +1.0) min(0.0961, 1) >>> algorithm chooses the lowest of the two max(-1, 0.0961) >>> algorithm chooses the highest level ------------------------------------------------------- the words that influence the modality rate: (strongly) said popular belief belief of course two ------------------------------------------------------- modality.py results (black box results) sentence : This strongly said decision-making is popular belief and of course designed to wish and not wish the human out of the two equations. parsed : This/DT/O/O/this strongly/RB/B-VP/O/strongly said/VBD/I-VP/O/say decision-making/NN/B-NP/O/decision-making is/VBZ/B-VP/O/be popular/JJ/B-NP/O/popular belief/NN/I-NP/O/belief and/CC/O/O/and of/IN/B-PP/B-PNP/of course/NN/B-NP/I-PNP/course designed/VBN/B-VP/I-PNP/design to/TO/I-VP/O/to wish/VB/I-VP/O/wish and/CC/O/O/and not/RB/B-VP/O/not wish/VB/I-VP/O/wish the/DT/O/O/the human/JJ/B-ADJP/O/human out/IN/B-PP/B-PNP/out of/IN/I-PP/I-PNP/of the/DT/B-NP/I-PNP/the two/CD/I-NP/I-PNP/two equations/NNS/I-NP/I-PNP/equation ././O/O/. Sentence(s) : Sentence('This/DT/O/O/this strongly/RB/B-VP/O/strongly said/VBD/I-VP/O/say decision-making/NN/B-NP/O/decision-making is/VBZ/B-VP/O/be popular/JJ/B-NP/O/popular belief/NN/I-NP/O/belief and/CC/O/O/and of/IN/B-PP/B-PNP/of course/NN/B-NP/I-PNP/course designed/VBN/B-VP/I-PNP/design to/TO/I-VP/O/to wish/VB/I-VP/O/wish and/CC/O/O/and not/RB/B-VP/O/not wish/VB/I-VP/O/wish the/DT/O/O/the human/JJ/B-ADJP/O/human out/IN/B-PP/B-PNP/out of/IN/I-PP/I-PNP/of the/DT/B-NP/I-PNP/the two/CD/I-NP/I-PNP/two equations/NNS/I-NP/I-PNP/equation ././O/O/.') modality rate: 0.0961538461538 >>> RIGHT! word : This modality : 1.0 ------------- word : strongly modality : 1.0 ------------- word : said modality : 0.25 ------------- word : decision-making modality : 1.0 ------------- word : is modality : 1.0 ------------- word : popular modality : 1.0 ------------- word : belief modality : -0.25 ------------- word : and modality : 1.0 ------------- word : of modality : 1.0 ------------- word : course modality : 1.0 ------------- word : designed modality : 1.0 ------------- word : to modality : 1.0 ------------- word : wish modality : 1.0 ------------- word : and modality : 1.0 ------------- word : not modality : 1.0 ------------- word : wish modality : 1.0 ------------- word : the modality : 1.0 ------------- word : human modality : 1.0 ------------- word : out modality : 1.0 ------------- word : of modality : 1.0 ------------- word : the modality : 1.0 ------------- word : two modality : 0.75 ------------- word : equations. modality : 1.0 -------------