means of Natural Language Processing Viorel Morari Chair of Web - PowerPoint PPT Presentation

Omission: Hypozeugma Hypozeugma - placing last, in a construction containing several words or phrases of equal value, the word or words on which all of them depend. (Silva Rhetoricae) A rooster, a prince and a lion walk into a bar… Stanford Dependencies A rooster, a prince and a lion walk into a bar… governor-dependent relation 52

Omission: Hypozeugma Hypozeugma - placing last, in a construction containing several words or phrases of equal value, the word or words on which all of them depend. (Silva Rhetoricae) A rooster, a prince and a lion walk into a bar… Stanford Dependencies A rooster, a prince and a lion walk into a bar… UIMA Ruta rooster, a prince and a lion walk 53

Omission: Hypozeugma Hypozeugma - placing last, in a construction containing several words or phrases of equal value, the word or words on which all of them depend. (Silva Rhetoricae) A rooster, a prince and a lion walk into a bar… Stanford Dependencies A rooster, a prince and a lion walk into a bar… UIMA Ruta rooster, a prince and a lion walk 54

Omission: Hypozeugma Hypozeugma - placing last, in a construction containing several words or phrases of equal value, the word or words on which all of them depend. (Silva Rhetoricae) A rooster, a prince and a lion walk into a bar… Stanford Dependencies A rooster, a prince and a lion walk into a bar… UIMA Ruta rooster, a prince and a lion walk 55

Omission: Hypozeugma Hypozeugma - placing last, in a construction containing several words or phrases of equal value, the word or words on which all of them depend. (Silva Rhetoricae) A rooster, a prince and a lion walk into a bar… Stanford Dependencies A rooster, a prince and a lion walk into a bar… UIMA Ruta rooster, a prince and a lion walk 56

Omission: Hypozeugma Hypozeugma - placing last, in a construction containing several words or phrases of equal value, the word or words on which all of them depend. (Silva Rhetoricae) A rooster, a prince and a lion walk into a bar… Stanford Dependencies A rooster, a prince and a lion walk into a bar… UIMA Ruta rooster, a prince and a lion walk 57

Omission: Hypozeugma Hypozeugma - placing last, in a construction containing several words or phrases of equal value, the word or words on which all of them depend. (Silva Rhetoricae) A rooster, a prince and a lion walk into a bar… Stanford Dependencies A rooster, a prince and a lion walk into a bar… UIMA Ruta rooster, a prince and a lion walk 58

Omission: Hypozeugma Hypozeugma - placing last, in a construction containing several words or phrases of equal value, the word or words on which all of them depend. (Silva Rhetoricae) A rooster, a prince and a lion walk into a bar… Stanford Dependencies A rooster, a prince and a lion walk into a bar… UIMA Ruta rooster, a prince and a lion walk 59

Repetition: Epanalepsis Epanalepsis - repeats the beginning word of a sentence at the end. Our eyes saw it, but we could not believe our eyes. 60

Repetition: Epanalepsis Epanalepsis - repeats the beginning word of a sentence at the end. Our eyes saw it, but we could not believe our eyes. 61

Repetition: Epanalepsis Epanalepsis - repeats the beginning word of a sentence at the end. Our eyes saw it, but we could not believe our eyes. Our eyes saw believe our eyes. 62

Repetition: Epanalepsis Epanalepsis - repeats the beginning word of a sentence at the end. Our eyes saw it, but we could not believe our eyes. Our eyes saw believe our eyes. 63

Repetition: Epanalepsis Epanalepsis - repeats the beginning word of a sentence at the end. Our eyes saw it, but we could not believe our eyes. Our eyes saw believe our eyes. 64

Repetition: Epanalepsis Epanalepsis - repeats the beginning word of a sentence at the end. Our eyes saw it, but we could not believe our eyes. Our eyes saw believe our eyes. 65

Repetition: Epanalepsis Epanalepsis - repeats the beginning word of a sentence at the end. Our eyes saw it, but we could not believe our eyes. Our eyes saw believe our eyes. 66

Repetition: Epanalepsis Epanalepsis - repeats the beginning word of a sentence at the end. Our eyes saw it, but we could not believe our eyes. Our eyes saw believe our eyes. 67

Repetition: Epanalepsis Epanalepsis - repeats the beginning word of a sentence at the end. Our eyes saw it, but we could not believe our eyes. Our eyes saw believe our eyes. 68

Custom: If-conditional 2 If-conditional 2 - expresses consequences that are totally unrealistic or will not likely happen in the future. If I were president, I would cut taxes. 69

Custom: If-conditional 2 If-conditional 2 - expresses consequences that are totally unrealistic or will not likely happen in the future. If I were president, I would cut taxes. Stanford Dependencies If I were president, I would cut taxes. 70

Custom: If-conditional 2 If-conditional 2 - expresses consequences that are totally unrealistic or will not likely happen in the future. If I were president, I would cut taxes. Stanford Dependencies If I were president, I would cut taxes. governor-dependent relations 71

Custom: If-conditional 2 If-conditional 2 - expresses consequences that are totally unrealistic or will not likely happen in the future. If I were president, I would cut taxes. Stanford Dependencies If I were president, I would cut taxes. 72

Custom: If-conditional 2 If-conditional 2 - expresses consequences that are totally unrealistic or will not likely happen in the future. If I were president, I would cut taxes. Stanford Dependencies If I were president, I would cut taxes. P-clause Q-clause 73

Custom: If-conditional 2 If-conditional 2 - expresses consequences that are totally unrealistic or will not likely happen in the future. If I were president, I would cut taxes. Stanford Dependencies If I were president, I would cut taxes. UIMA Ruta If I were president I would cut 74

Custom: If-conditional 2 If-conditional 2 - expresses consequences that are totally unrealistic or will not likely happen in the future. If I were president, I would cut taxes. Stanford Dependencies If I were president, I would cut taxes. UIMA Ruta If I were president I would cut 75

Custom: If-conditional 2 If-conditional 2 - expresses consequences that are totally unrealistic or will not likely happen in the future. If I were president, I would cut taxes. Stanford Dependencies If I were president, I would cut taxes. UIMA Ruta If I were president I would cut 76

Custom: If-conditional 2 If-conditional 2 - expresses consequences that are totally unrealistic or will not likely happen in the future. If I were president, I would cut taxes. Stanford Dependencies If I were president, I would cut taxes. UIMA Ruta If I were president I would cut 77

Custom: If-conditional 2 If-conditional 2 - expresses consequences that are totally unrealistic or will not likely happen in the future. If I were president, I would cut taxes. Stanford Dependencies If I were president, I would cut taxes. UIMA Ruta If I were president I would cut 78

Custom: If-conditional 2 If-conditional 2 - expresses consequences that are totally unrealistic or will not likely happen in the future. If I were president, I would cut taxes. Stanford Dependencies If I were president, I would cut taxes. UIMA Ruta If I were president I would cut 79

Evaluation dataset Literature Political Expert speeches websites Rhetorical Commercials Bible Devices 80

Evaluation dataset Literature Political Expert speeches websites Rhetorical Commercials Bible Devices Evaluation measures 𝑢𝑞 𝑢𝑞 𝐺1 𝑡𝑑𝑝𝑠𝑓 = 2 ∙ 𝑞𝑠𝑓𝑑𝑗𝑡𝑗𝑝𝑜 ∙ 𝑠𝑓𝑑𝑏𝑚𝑚 𝑄𝑠𝑓𝑑𝑗𝑡𝑗𝑝𝑜 = 𝑆𝑓𝑑𝑏𝑚𝑚 = 𝑢𝑞 + 𝑔𝑞 𝑢𝑞 + 𝑔𝑜 𝑞𝑠𝑓𝑑𝑗𝑡𝑗𝑝𝑜 + 𝑠𝑓𝑑𝑏𝑚𝑚 81

Evaluation Results Balance schemes Omission schemes 1 1 1 0.84 0.9 0.9 0.8 0.8 0.68 0.69 0.69 0.7 0.7 Enumeration Hypozeugma 0.6 0.6 Epizeugma 0.5 0.5 0.4 Isocolon 0.4 0.4 Pysma Asynd. 0.3 0.3 0.2 0.2 0.1 0.1 0 0 82

Evaluation Results Balance schemes Omission schemes 1 1 0.95 1 0.84 0.9 0.9 0.8 0.8 0.68 0.69 0.69 0.7 0.7 Enumeration Hypozeugma 0.6 0.6 Epizeugma 0.5 0.5 0.4 Isocolon 0.4 0.4 Pysma Asynd. 0.25 0.3 0.3 0.2 0.2 0.1 0.1 0 0 Precision Recall 83

Evaluation Results Balance schemes Omission schemes 1 1 0.95 1 0.84 0.9 0.9 0.8 0.8 0.68 0.69 0.69 0.7 0.7 Enumeration Hypozeugma 0.6 0.6 Epizeugma 0.5 0.5 0.4 Isocolon 0.4 0.4 Pysma Asynd. 0.25 0.3 0.3 0.2 0.2 0.1 0.1 0 0 Precision Recall Enumeration Asyndeton Old farmer had a pig, a dog, a cow and a horse. 84

Evaluation Results Balance schemes Omission schemes 1 1 1 0.84 0.9 0.9 0.8 0.8 0.68 0.69 0.69 0.7 0.7 Enumeration Hypozeugma 0.6 0.6 Epizeugma 0.5 0.5 0.4 Isocolon 0.4 0.4 Pysma Asynd. 0.3 0.3 0.2 0.2 0.1 0.1 0 0 Custom schemes Repetition schemes 1 1 0.91 1 0.85 0.87 0.9 0.9 0.73 0.78 0.78 0.74 0.78 0.8 0.74 0.74 0.74 0.73 0.8 0.72 0.67 0.7 0.56 0.61 0.7 Whether-cond 0.59 If-counterFact 0.56 Polysyndeton Passive Voice 0.6 0.6 Unless-cond Epanalepsis Anadiplosis Comp. Adv. Super. Adv. Comp. Adj. Super. Adj. 0.5 0.5 Mesarchia 0.4 Epiphoza Epizeuxis If-cond 1 If-cond 3 If-cond 0 If-cond 2 Diacope 0.4 0.4 Mesod. 0.3 0.3 0.2 0.2 0.1 0.1 85 0 0

Evaluation Results Balance schemes Omission schemes 1 1 1 0.84 0.9 0.9 0.8 0.8 0.68 0.69 0.69 0.7 0.7 Enumeration Hypozeugma 0.6 0.6 Epizeugma 0.5 0.5 0.4 Isocolon 0.4 0.4 Pysma Asynd. 0.3 0.3 0.2 0.2 0.1 0.1 0 0 Custom schemes Repetition schemes 1 1 0.91 1 0.85 0.87 0.9 0.9 0.73 0.78 0.78 0.74 0.78 0.8 0.74 0.74 0.74 0.73 0.8 0.72 0.67 0.7 0.56 0.61 0.7 Whether-cond 0.59 If-counterFact 0.56 Polysyndeton Passive Voice 0.6 0.6 Unless-cond Epanalepsis Anadiplosis Comp. Adv. Super. Adv. Comp. Adj. Super. Adj. 0.5 0.5 Mesarchia 0.4 Epiphoza Epizeuxis If-cond 1 If-cond 3 If-cond 0 If-cond 2 Diacope 0.4 0.4 Mesod. 0.3 0.3 0.2 0.2 0.1 0.1 86 0 0

Evaluation Results F1-Score Balance schemes Omission schemes 1 1 1 0.84 0.9 0.9 0.8 0.8 0.68 0.69 0.69 0.7 0.7 Enumeration Hypozeugma 0.6 0.6 Epizeugma 0.5 0.5 0.4 Isocolon 0.4 0.4 Pysma Asynd. 0.3 0.3 0.2 0.2 0.1 0.1 0 0 Custom schemes Repetition schemes 1 1 0.91 1 0.85 0.87 0.9 0.9 0.73 0.78 0.78 0.74 0.78 0.8 0.74 0.74 0.74 0.73 0.8 0.72 0.67 0.7 0.56 0.61 0.7 Whether-cond 0.59 If-counterFact 0.56 Polysyndeton Passive Voice 0.6 0.6 Unless-cond Epanalepsis Anadiplosis Comp. Adv. Super. Adv. Comp. Adj. Super. Adj. 0.5 0.5 Mesarchia 0.4 Epiphoza Epizeuxis If-cond 1 If-cond 3 If-cond 0 If-cond 2 Diacope 0.4 0.4 Mesod. 0.3 0.3 0.2 0.2 0.1 0.1 87 0 0

Pipeline output 88

Pipeline output 89

Analysis of 2 Rhetorical Devices 90

Pipeline output 91

Pipeline Group distribution Frequency Combinations output + + + + + + 92

Pipeline Precision Device Recall output Confidence interval n * precision (rd) n/recall (rd) Al-Khatib et al. [2017] 93

Pipeline Significance Test A B Effect-size Test output A B 94

Pipeline Type Topic output 95

Pipeline 96

Analysis 3 Experiments Data Preparation 97

Experiments: datasets The New York Times US Presidential Debates 2016 Ben Wiseman [2016] 98

Data dimensionality Language Mode Communication Author Audience English Written Monological Identity U.S. Type Genre Topic Medium Descriptive Editorial Education Newspaper Argumentative Review Science Presidential Debates Biography Art Debate Politics 99

NYT Experiment: data subsampling NYT Dataset Random Article-length based Matching 1000 articles 600 articles 343 articles 100