| { | |
| "paper_id": "N15-1023", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T14:34:51.697302Z" | |
| }, | |
| "title": "Inducing Lexical Style Properties for Paraphrase and Genre Differentiation", | |
| "authors": [ | |
| { | |
| "first": "Ellie", | |
| "middle": [], | |
| "last": "Pavlick", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "University of Pennsylvania", | |
| "location": {} | |
| }, | |
| "email": "epavlick@seas.upenn.edu" | |
| }, | |
| { | |
| "first": "Ani", | |
| "middle": [], | |
| "last": "Nenkova", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "University of Pennsylvania", | |
| "location": {} | |
| }, | |
| "email": "nenkova@seas.upenn.edu" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "We present an intuitive and effective method for inducing style scores on words and phrases. We exploit signal in a phrase's rate of occurrence across stylistically contrasting corpora, making our method simple to implement and efficient to scale. We show strong results both intrinsically, by correlation with human judgements, and extrinsically, in applications to genre analysis and paraphrasing.", | |
| "pdf_parse": { | |
| "paper_id": "N15-1023", | |
| "_pdf_hash": "", | |
| "abstract": [ | |
| { | |
| "text": "We present an intuitive and effective method for inducing style scores on words and phrases. We exploit signal in a phrase's rate of occurrence across stylistically contrasting corpora, making our method simple to implement and efficient to scale. We show strong results both intrinsically, by correlation with human judgements, and extrinsically, in applications to genre analysis and paraphrasing.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Abstract", | |
| "sec_num": null | |
| } | |
| ], | |
| "body_text": [ | |
| { | |
| "text": "True language understanding requires comprehending not just what is said, but how it is said, yet only recently have computational approaches been applied to the subtleties of tone and style. As the expectations on language technologies grow to include tailored search, context-aware inference, and analysis of author belief, an understanding of style becomes crucial.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "Lexical features have proven indispensable for the good performance of most applications dealing with language. Particularly, more generalized characterizations of the lexicon (Brown et al., 1992; Wilson et al., 2005; Feng et al., 2013; Ji and Lin, 2009; Resnik, 1995) have become key in overcoming issues with lexical sparseness and in providing practical semantic information for natural language processing systems (Miller et al., 2004; Rutherford and Xue, 2014; Velikovich et al., 2010; Dodge et al., 2012) . Most work on stylistic variation, however, has focused on larger units of text (Louis and Nenkova, 2013; Danescu-Niculescu-Mizil et al., 2012; Greene and Resnik, 2009; Xu et al., 2012) and studies of style at the lexical level have been scant. The few recent efforts (Brooke et al., 2010; Brooke and Hirst, 2013b; Formal/Casual Complex/Simple jesus/my gosh great/a lot 18 years/eighteen cinema/a movie respiratory/breathing a large/a big yes/yeah music/the band decade/ten years much/many things 1970s/the seventies exposure/the show foremost/first of all relative/his family megan/you there matters/the things somewhere/some place april/apr this film/that movie journal/diary full/a whole bunch the world/everybody otherwise/another thing burial/funeral father/my dad rail/the train recreation/hobby physicians/a doctor Table 1 : Paraphrases with large style differences. Our method learns these distinctions automatically. Brooke and Hirst, 2013a) have been motivated by the need to categorize genre in multiple continuous dimensions and focused on applying standard methods for lexical characterization via graph propagation or crowdsourcing.", | |
| "cite_spans": [ | |
| { | |
| "start": 176, | |
| "end": 196, | |
| "text": "(Brown et al., 1992;", | |
| "ref_id": "BIBREF3" | |
| }, | |
| { | |
| "start": 197, | |
| "end": 217, | |
| "text": "Wilson et al., 2005;", | |
| "ref_id": "BIBREF23" | |
| }, | |
| { | |
| "start": 218, | |
| "end": 236, | |
| "text": "Feng et al., 2013;", | |
| "ref_id": "BIBREF9" | |
| }, | |
| { | |
| "start": 237, | |
| "end": 254, | |
| "text": "Ji and Lin, 2009;", | |
| "ref_id": "BIBREF14" | |
| }, | |
| { | |
| "start": 255, | |
| "end": 268, | |
| "text": "Resnik, 1995)", | |
| "ref_id": "BIBREF20" | |
| }, | |
| { | |
| "start": 418, | |
| "end": 439, | |
| "text": "(Miller et al., 2004;", | |
| "ref_id": "BIBREF17" | |
| }, | |
| { | |
| "start": 440, | |
| "end": 465, | |
| "text": "Rutherford and Xue, 2014;", | |
| "ref_id": "BIBREF21" | |
| }, | |
| { | |
| "start": 466, | |
| "end": 490, | |
| "text": "Velikovich et al., 2010;", | |
| "ref_id": "BIBREF22" | |
| }, | |
| { | |
| "start": 491, | |
| "end": 510, | |
| "text": "Dodge et al., 2012)", | |
| "ref_id": "BIBREF7" | |
| }, | |
| { | |
| "start": 592, | |
| "end": 617, | |
| "text": "(Louis and Nenkova, 2013;", | |
| "ref_id": "BIBREF16" | |
| }, | |
| { | |
| "start": 618, | |
| "end": 655, | |
| "text": "Danescu-Niculescu-Mizil et al., 2012;", | |
| "ref_id": "BIBREF6" | |
| }, | |
| { | |
| "start": 656, | |
| "end": 680, | |
| "text": "Greene and Resnik, 2009;", | |
| "ref_id": "BIBREF12" | |
| }, | |
| { | |
| "start": 681, | |
| "end": 697, | |
| "text": "Xu et al., 2012)", | |
| "ref_id": "BIBREF24" | |
| }, | |
| { | |
| "start": 780, | |
| "end": 801, | |
| "text": "(Brooke et al., 2010;", | |
| "ref_id": "BIBREF2" | |
| }, | |
| { | |
| "start": 802, | |
| "end": 826, | |
| "text": "Brooke and Hirst, 2013b;", | |
| "ref_id": "BIBREF1" | |
| }, | |
| { | |
| "start": 827, | |
| "end": 827, | |
| "text": "", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 1439, | |
| "end": 1463, | |
| "text": "Brooke and Hirst, 2013a)", | |
| "ref_id": "BIBREF0" | |
| } | |
| ], | |
| "ref_spans": [ | |
| { | |
| "start": 1335, | |
| "end": 1342, | |
| "text": "Table 1", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "We propose a simple and flexible method for placing phrases along a style spectrum. We focus on two dimensions: formality and complexity. We evaluate the resulting scores in terms of their correlation with human judgements as well as their utility in two tasks. First, we use the induced dimensions to identify stylistic shifts in paraphrase, allowing us to differentiate stylistic properties in the Paraphrase Database (PPDB) with high accuracy. Second, we test how well the induced scores capture differences between genres, and explore the extent to which these differences are due to topic versus lexical choice between stylistically different expressions for the same content. We show that style alone does differentiate between genres, and that the combined indicators of style and topic are highly effective in describing genre in a way consistent with human judgements.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction", | |
| "sec_num": "1" | |
| }, | |
| { | |
| "text": "We focus on two style dimensions: formality and complexity. We define formal language as the way one talks to a superior, whereas casual language is used with friends. We define simple language to be that used to talk to children or non-native English speakers, whereas more complex language is used by academics or domain experts.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Method", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "We use the Europarl corpus of parliamentary proceedings as an example of formal text and the Switchboard corpus of informal telephone conversations as casual text. We use articles from Wikipedia and simplified Wikipedia (Coster and Kauchak, 2011) as examples of complex and simple language respectively. For each style dimension, we subsample sentences from the larger corpus so that the two ends of the spectrum are roughly balanced. We end up with roughly 300K sentences each for formal/casual text and about 500K sentences each for simple/complex text. 1 Given examples of language at each end of a style dimension, we score a phrase by the log ratio of the probability of observing the word in the reference corpus (REF) to observing it in the combined corpora (ALL). For formality the reference corpus is Europarl and the combined data is Europarl and Switchboard together. For complexity, the reference corpus is normal Wikipedia and the combined data is normal and simplified Wikipedia together. Specifically, we map a phrase w onto a style dimension via:", | |
| "cite_spans": [ | |
| { | |
| "start": 220, | |
| "end": 246, | |
| "text": "(Coster and Kauchak, 2011)", | |
| "ref_id": "BIBREF5" | |
| }, | |
| { | |
| "start": 556, | |
| "end": 557, | |
| "text": "1", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 719, | |
| "end": 724, | |
| "text": "(REF)", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Method", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "FORMALITY(w) = log P (w | REF ) P (w | ALL) .", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Method", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "We assign formality scores to phrases up to three words in length that occur at least three times total in ALL, regardless of whether they occur in both corpora. Phrases which do not occur at all in REF are treated as though they occurred once.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Method", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "Phrase-level human judgements For each of our style dimensions, we take a random sample of 1,000 phrases from our corpora. We show each phrase to 7 workers on Amazon Mechanical Turk (MTurk) and ask the worker to indicate using a sliding bar (corresponding to a 0 to 100 scale) where they feel each word falls on the given style spectrum (e.g. casual to formal). Workers were given a high-level description of each style (like those given at the beginning of Section 2) and examples to guide their annotation. We estimate inter-annotator agreement by computing each rater's correlation with the average of the others. The inter-annotator correlation was reasonably strong on average (\u03c1 = 0.65). However, not all phrases had equally strong levels of human agreement. Table 2 shows some examples of phrases which fell \"obviously\" on one end of a style spectrum (i.e. the variance between humans' ratings was low) and some other examples which were less clear.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 765, | |
| "end": 772, | |
| "text": "Table 2", | |
| "ref_id": "TABREF1" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Method", | |
| "sec_num": "2" | |
| }, | |
| { | |
| "text": "We compute the correlation of our method's score with the average human rating for each phrase. The results are summarized in Table 4 . The log-ratio score correlates with the human score significantly above chance, even matching inter-human levels of correlation on the formality dimension.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 126, | |
| "end": 133, | |
| "text": "Table 4", | |
| "ref_id": "TABREF4" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Quality of automatic scores", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "We evaluate the acquired style mappings in two tasks: finding paraphrase pairs with differences in style and characterizing genre variation. ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Applications", | |
| "sec_num": "4" | |
| }, | |
| { | |
| "text": "Paraphrases are usually defined as \"meaning equivalent\" words or phrases. However, many paraphrases, even while capturing the same meaning overall, display subtle differences which effect their substitutability (Gardiner and Dras, 2007) . For example, paraphrasing \"I believe that we have...\" as \"I think we got...\" preserves the meaning but causes a clear change in style, from a more formal register to a casual one. It has been proposed that paraphrases are rarely if ever perfectly equivalent, but instead represent near synonyms (Edmonds and Hirst, 2002) , which contain subtle differences in meaning and connotation.", | |
| "cite_spans": [ | |
| { | |
| "start": 211, | |
| "end": 236, | |
| "text": "(Gardiner and Dras, 2007)", | |
| "ref_id": "BIBREF11" | |
| }, | |
| { | |
| "start": 534, | |
| "end": 559, | |
| "text": "(Edmonds and Hirst, 2002)", | |
| "ref_id": "BIBREF8" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Differentiating style in paraphrases", | |
| "sec_num": "4.1" | |
| }, | |
| { | |
| "text": "We test whether our method can tease apart stylistic variation given a set of \"equivalent\" phrases. We use phrase pairs from the Paraphrase Database (PPDB) (Ganitkevitch et al., 2013) . Using the scoring method described in Section 2, we identify paraphrase pairs which display stylistic variation along a particular dimension. We can find pairs w1, w2 in PPDB for which FORMALITY(w1) \u2212 FORMALITY(w2) is large; Table 1 gives some examples of pairs identified using this method. We can also view paraphrases along a continuum; Table 3 shows groups of paraphrases ordered from most formal to most casual and Figure 1 shows how paraphrases of the phrase money rank along the formality and complexity dimensions. For example, we capture the fact that money is more formal but simpler than the idiomatic expression a fortune. Pairwise human judgements To evaluate the goodness of our style-adapted paraphrases, we take a random sample of 3,000 paraphrase pairs from PPDB and solicit MTurk judgements. We show workers each paraphrase pair and ask them to choose which of the words is more casual, or to indicate \"no difference.\" We also carry out the analogous task for the complexity distinction. We take the majority of 7 judgements as the true label for each pair.", | |
| "cite_spans": [ | |
| { | |
| "start": 156, | |
| "end": 183, | |
| "text": "(Ganitkevitch et al., 2013)", | |
| "ref_id": "BIBREF10" | |
| } | |
| ], | |
| "ref_spans": [ | |
| { | |
| "start": 411, | |
| "end": 418, | |
| "text": "Table 1", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 526, | |
| "end": 533, | |
| "text": "Table 3", | |
| "ref_id": "TABREF3" | |
| }, | |
| { | |
| "start": 606, | |
| "end": 614, | |
| "text": "Figure 1", | |
| "ref_id": "FIGREF0" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Differentiating style in paraphrases", | |
| "sec_num": "4.1" | |
| }, | |
| { | |
| "text": "In only 9% of the 3,000 paraphrase pairs, turkers decided there was no stylistic difference in the pair, indicating that indeed formality and complexity differences are truly characteristic of paraphrases. In further analysis we ignore the pairs for which the consensus was no difference but note that in further work we need to automate the identification of stylistically equivalent paraphrases.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Differentiating style in paraphrases", | |
| "sec_num": "4.1" | |
| }, | |
| { | |
| "text": "Automatically differentiating paraphrases Using the human judgements, we compute the accuracy of our method for choosing which word in a pair is more formal (complex). We use the magnitude of the difference in formality (complexity) score as a measure of our method's confidence in its prediction. E.g. the smaller the gap in FORMALITY, the less confident our method is that there is a true style difference. Table 5 shows pairwise accuracy as a function of confidence: it is well above the 50% random baseline, reaching 90% for the highconfidence predictions in the complexity dimension. ", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 409, | |
| "end": 416, | |
| "text": "Table 5", | |
| "ref_id": "TABREF6" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Differentiating style in paraphrases", | |
| "sec_num": "4.1" | |
| }, | |
| { | |
| "text": "Now we explore if the dimensions we learned at the sub-sentential level can be used to capture stylistic variation at the sentence and genre level.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Genre characterization", | |
| "sec_num": "4.2" | |
| }, | |
| { | |
| "text": "Sentence-level human judgements We gather human ratings of formality and complexity for 900 sentences from the MASC corpus (Ide et al., 2010) : 20 sentences from each of 18 genres. 2 Recently data from this corpus has been used to study genre difference in terms of pronoun, named entity, punctuation and part of speech usage (Passonneau et al., 2014) . We use the data to test a specific hypothesis that automatically induced scores for lexical style are predictive of perceptions of sentence-and genre-level style. We average 7 independent human scores to get sentence-level style scores. To get genre-level style scores, we use the the average of the 20 sentencelevel scores for the sentences belonging to that genre.", | |
| "cite_spans": [ | |
| { | |
| "start": 123, | |
| "end": 141, | |
| "text": "(Ide et al., 2010)", | |
| "ref_id": "BIBREF13" | |
| }, | |
| { | |
| "start": 326, | |
| "end": 351, | |
| "text": "(Passonneau et al., 2014)", | |
| "ref_id": "BIBREF18" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Genre characterization", | |
| "sec_num": "4.2" | |
| }, | |
| { | |
| "text": "In human perception, the formality and complexity dimensions are highly correlated (Spearman \u03c1 = 0.7). However, we see many interesting examples of sentences which break this trend (Table 6 ). Overall, inter-annotator correlations are reasonably strong (\u03c1 \u2248 0.5), but as in the phrase-level 2 Court transcripts, debate transcripts, face-to-face conversations, blogs, essays, fiction, jokes, letters, technical writing, newspaper, twitter, email, ficlets (short fan fiction), government documents, journal entries, movie scripts, non-fiction, and travel guides. We omit the \"telephone\" genre, since it is too similar to the Switchboard corpus and may inflate results.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 181, | |
| "end": 190, | |
| "text": "(Table 6", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Genre characterization", | |
| "sec_num": "4.2" | |
| }, | |
| { | |
| "text": "annotations, we see some sentences for which the judgement seems unanimous among annotators and some sentences for which there is very little consensus (Table 7) . We discuss this variation further in Section 5.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 152, | |
| "end": 161, | |
| "text": "(Table 7)", | |
| "ref_id": "TABREF8" | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Genre characterization", | |
| "sec_num": "4.2" | |
| }, | |
| { | |
| "text": "Formal/Simple has dr. miller left the courtroom? Formal/Simple i want to thank you for listening tonight. Casual/Complex right. cuz if we have a fixed number of neurons-? Casual/Complex i was actually thinking we could use the warping factors that we compute for the mfcc's Table 6 : Some examples of sentences for which the generally high correlation between formality and complexity does not hold.", | |
| "cite_spans": [], | |
| "ref_spans": [ | |
| { | |
| "start": 274, | |
| "end": 281, | |
| "text": "Table 6", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Genre characterization", | |
| "sec_num": "4.2" | |
| }, | |
| { | |
| "text": "Automatically characterizing genre The extent to which genre is defined by topic versus style is an open question. We therefore look at two methods for genre-level style characterization, which we apply at the sentence-level as well as at the genre-level. First, we take the average formality (complexity) score of all words in the text, which we refer to as the \"all words\" method. Using the style score alone in this way will likely to conflate aspects of topic with aspects of style. For example, the word birthday receives a very low formality score whereas the phrase united nations receives a very high formality score, reflecting the tendency of certain topics to be discussed more formally than others. We therefore use a second method, which we refer to as \"PP only\", in which we look only at the words in the text which belong to one of our paraphrase sets (as in Figure 3 ), allowing us to control for topic and focus only on stylistic word choice. In \"PP only\", we consider a word to be formal if it appears on the formal side of the set (i.e. there are more phrases to its left than to its right). We then score the overall formality of the text as the proportion of times a formal phrase was chosen when a more casual paraphrase could have been chosen instead. The intuition is captured in Figure 2 : when an author is writing about a given topic, she encounters words for which there exist a range of paraphrases. Her lexical choice in these cases signals the style independent of the topic. Table 8 shows how well our two scoring methods correlate with the human judgements of sentences' styles. The \"all words\" method performs very well, correlating with humans nearly as well as humans correlate with each other. Interestingly, when using paraphrases only we maintain significant correlations. This ability to differentiate stylistic variation without relying on cues from topic words could be especially important for tasks such as bias detection (Recasens et al., 2013) and readability (Callan, 2004; Kanungo and Orr, 2009 Table 8 : Spearman \u03c1 of automatic rankings with human rankings. Genres are the concatenation of sentences from that genre. In \"all words,\" a text's score is the average log-ratio style score of its words. In \"PP only,\" a text's score is the proportion of times a formal term was chosen when more casual paraphrases existed, effectively capturing style independent of topic.", | |
| "cite_spans": [ | |
| { | |
| "start": 1966, | |
| "end": 1989, | |
| "text": "(Recasens et al., 2013)", | |
| "ref_id": "BIBREF19" | |
| }, | |
| { | |
| "start": 2006, | |
| "end": 2020, | |
| "text": "(Callan, 2004;", | |
| "ref_id": "BIBREF4" | |
| }, | |
| { | |
| "start": 2021, | |
| "end": 2042, | |
| "text": "Kanungo and Orr, 2009", | |
| "ref_id": "BIBREF15" | |
| } | |
| ], | |
| "ref_spans": [ | |
| { | |
| "start": 874, | |
| "end": 882, | |
| "text": "Figure 3", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 1304, | |
| "end": 1312, | |
| "text": "Figure 2", | |
| "ref_id": "FIGREF1" | |
| }, | |
| { | |
| "start": 1507, | |
| "end": 1514, | |
| "text": "Table 8", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 2043, | |
| "end": 2050, | |
| "text": "Table 8", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Genre characterization", | |
| "sec_num": "4.2" | |
| }, | |
| { | |
| "text": "Characterization of style at the lexical level is an important first step in complex natural language tasks, capturing style information in a way that is portable across topics and applications. An interesting open question is the extent to which style is defined at the lexical level versus at the sentential level: how strongly are human perceptions of style influenced by topic and context as opposed to by lexical choice? One interesting phenomenon we observe is that inter-annotator correlations are lower at the sentence level (\u03c1 \u2248 0.5) than at the wordand phrase-level (\u03c1 \u2248 0.65). Tables 7 offers some insight: for many of the sentences for which human agreement is low, there seems to be some mismatch between the topic and the typical style of that topic (e.g. talking formally about family life, or talking in relatively complex terms about Facebook). When humans are making judgements at the lexical level, such contextual mismatches don't arise, which might lead to higher overall agreements. Interesting future work will need to explore how well humans are able to separate style from topic at the sentenceand document-level, and how the lexical choice of the author/speaker affects this distinction.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Discussion", | |
| "sec_num": "5" | |
| }, | |
| { | |
| "text": "We present a simple and scalable method for learning fine-grained stylistic variation of phrases. We demonstrate good preliminary results on two relevant applications: identifying stylistic differences in paraphrase, and characterizing variations between genres. Our method offers a simple and flexible way of acquiring stylistic annotations at web-scale, making it a promising approach for incorporating nuanced linguistic information into increasingly complex language applications. 3", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion", | |
| "sec_num": "6" | |
| }, | |
| { | |
| "text": "Number of words: casual (2MM), formal (7MM), simple (9MM), complex (12MM).3 EvaluationWe first assess the intrinsic quality of the scores returned by our method by comparing against subjective human judgements of stylistic properties.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "All human and log-ratio scores discussed are available at http://www.seas.upenn.edu/\u223cnlp/resources/ style-scores.tar.gz", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "Hybrid models for lexical acquisition of correlated styles", | |
| "authors": [ | |
| { | |
| "first": "Julian", | |
| "middle": [], | |
| "last": "Brooke", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Graeme", | |
| "middle": [], | |
| "last": "Hirst", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2013, | |
| "venue": "Proceedings of the Sixth International Joint Conference on Natural Language Processing", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "82--90", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Julian Brooke and Graeme Hirst. 2013a. Hybrid mod- els for lexical acquisition of correlated styles. In Pro- ceedings of the Sixth International Joint Conference on Natural Language Processing, pages 82-90.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "A multidimensional bayesian approach to lexical style", | |
| "authors": [ | |
| { | |
| "first": "Julian", | |
| "middle": [], | |
| "last": "Brooke", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Graeme", | |
| "middle": [], | |
| "last": "Hirst", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2013, | |
| "venue": "Proceedings of the 2013 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "673--679", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Julian Brooke and Graeme Hirst. 2013b. A multi- dimensional bayesian approach to lexical style. In Proceedings of the 2013 Conference of the North American Chapter of the Association for Computa- tional Linguistics: Human Language Technologies, pages 673-679.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "Automatic acquisition of lexical formality", | |
| "authors": [ | |
| { | |
| "first": "Julian", | |
| "middle": [], | |
| "last": "Brooke", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Tong", | |
| "middle": [], | |
| "last": "Wang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Graeme", | |
| "middle": [], | |
| "last": "Hirst", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2010, | |
| "venue": "Proceedings of the 23rd International Conference on Computational Linguistics: Posters", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "90--98", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Julian Brooke, Tong Wang, and Graeme Hirst. 2010. Au- tomatic acquisition of lexical formality. In Proceed- ings of the 23rd International Conference on Compu- tational Linguistics: Posters, pages 90-98.", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "Classbased n-gram models of natural language", | |
| "authors": [ | |
| { | |
| "first": "F", | |
| "middle": [], | |
| "last": "Peter", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Peter", | |
| "middle": [ | |
| "V" | |
| ], | |
| "last": "Brown", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Robert", | |
| "middle": [ | |
| "L" | |
| ], | |
| "last": "Desouza", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Vincent", | |
| "middle": [ | |
| "J" | |
| ], | |
| "last": "Mercer", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Jenifer", | |
| "middle": [ | |
| "C" | |
| ], | |
| "last": "Della Pietra", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Lai", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1992, | |
| "venue": "Comput. Linguist", | |
| "volume": "18", | |
| "issue": "4", | |
| "pages": "467--479", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Peter F. Brown, Peter V. deSouza, Robert L. Mercer, Vin- cent J. Della Pietra, and Jenifer C. Lai. 1992. Class- based n-gram models of natural language. Comput. Linguist., 18(4):467-479, December.", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "A language modeling approach to predicting reading difficulty", | |
| "authors": [ | |
| { | |
| "first": "Kevyn", | |
| "middle": [], | |
| "last": "Collins-Thompson", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Jamie", | |
| "middle": [], | |
| "last": "Callan", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2004, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kevyn Collins-Thompson Jamie Callan. 2004. A lan- guage modeling approach to predicting reading diffi- culty.", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "Simple english wikipedia: a new text simplification task", | |
| "authors": [ | |
| { | |
| "first": "William", | |
| "middle": [], | |
| "last": "Coster", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "David", | |
| "middle": [], | |
| "last": "Kauchak", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2011, | |
| "venue": "Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies: short papers", | |
| "volume": "2", | |
| "issue": "", | |
| "pages": "665--669", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "William Coster and David Kauchak. 2011. Simple en- glish wikipedia: a new text simplification task. In Pro- ceedings of the 49th Annual Meeting of the Associa- tion for Computational Linguistics: Human Language Technologies: short papers-Volume 2, pages 665-669. Association for Computational Linguistics.", | |
| "links": null | |
| }, | |
| "BIBREF6": { | |
| "ref_id": "b6", | |
| "title": "You had me at hello: How phrasing affects memorability", | |
| "authors": [ | |
| { | |
| "first": "Cristian", | |
| "middle": [], | |
| "last": "Danescu-Niculescu-Mizil", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Justin", | |
| "middle": [], | |
| "last": "Cheng", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Jon", | |
| "middle": [], | |
| "last": "Kleinberg", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Lillian", | |
| "middle": [], | |
| "last": "Lee", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2012, | |
| "venue": "Proceedings of the 50th Annual Meeting of the Association for Computational Linguistics: Long Papers", | |
| "volume": "1", | |
| "issue": "", | |
| "pages": "892--901", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Cristian Danescu-Niculescu-Mizil, Justin Cheng, Jon Kleinberg, and Lillian Lee. 2012. You had me at hello: How phrasing affects memorability. In Pro- ceedings of the 50th Annual Meeting of the Association for Computational Linguistics: Long Papers-Volume 1, pages 892-901. Association for Computational Lin- guistics.", | |
| "links": null | |
| }, | |
| "BIBREF7": { | |
| "ref_id": "b7", | |
| "title": "Detecting visual text", | |
| "authors": [ | |
| { | |
| "first": "Jesse", | |
| "middle": [], | |
| "last": "Dodge", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Amit", | |
| "middle": [], | |
| "last": "Goyal", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Xufeng", | |
| "middle": [], | |
| "last": "Han", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Alyssa", | |
| "middle": [], | |
| "last": "Mensch", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Margaret", | |
| "middle": [], | |
| "last": "Mitchell", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Karl", | |
| "middle": [], | |
| "last": "Stratos", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Kota", | |
| "middle": [], | |
| "last": "Yamaguchi", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yejin", | |
| "middle": [], | |
| "last": "Choi", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Hal", | |
| "middle": [], | |
| "last": "Daum\u00e9", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Alexander", | |
| "middle": [ | |
| "C" | |
| ], | |
| "last": "Berg", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Tamara", | |
| "middle": [ | |
| "L" | |
| ], | |
| "last": "Berg", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2012, | |
| "venue": "Human Language Technologies: Conference of the North American Chapter of the Association of Computational Linguistics, Proceedings", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "762--772", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Jesse Dodge, Amit Goyal, Xufeng Han, Alyssa Men- sch, Margaret Mitchell, Karl Stratos, Kota Yamaguchi, Yejin Choi, Hal Daum\u00e9 III, Alexander C. Berg, and Tamara L. Berg. 2012. Detecting visual text. In Human Language Technologies: Conference of the North American Chapter of the Association of Com- putational Linguistics, Proceedings,, pages 762-772.", | |
| "links": null | |
| }, | |
| "BIBREF8": { | |
| "ref_id": "b8", | |
| "title": "Nearsynonymy and lexical choice", | |
| "authors": [ | |
| { | |
| "first": "Philip", | |
| "middle": [], | |
| "last": "Edmonds", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Graeme", | |
| "middle": [], | |
| "last": "Hirst", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2002, | |
| "venue": "", | |
| "volume": "28", | |
| "issue": "", | |
| "pages": "105--144", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Philip Edmonds and Graeme Hirst. 2002. Near- synonymy and lexical choice. Computational linguis- tics, 28(2):105-144.", | |
| "links": null | |
| }, | |
| "BIBREF9": { | |
| "ref_id": "b9", | |
| "title": "Connotation lexicon: A dash of sentiment beneath the surface meaning", | |
| "authors": [ | |
| { | |
| "first": "Song", | |
| "middle": [], | |
| "last": "Feng", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Jun Seok", | |
| "middle": [], | |
| "last": "Kang", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Polina", | |
| "middle": [], | |
| "last": "Kuznetsova", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Yejin", | |
| "middle": [], | |
| "last": "Choi", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2013, | |
| "venue": "Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics, ACL", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "1774--1784", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Song Feng, Jun Seok Kang, Polina Kuznetsova, and Yejin Choi. 2013. Connotation lexicon: A dash of senti- ment beneath the surface meaning. In Proceedings of the 51st Annual Meeting of the Association for Com- putational Linguistics, ACL, pages 1774-1784.", | |
| "links": null | |
| }, | |
| "BIBREF10": { | |
| "ref_id": "b10", | |
| "title": "PPDB: The paraphrase database", | |
| "authors": [ | |
| { | |
| "first": "Juri", | |
| "middle": [], | |
| "last": "Ganitkevitch", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Benjamin", | |
| "middle": [], | |
| "last": "Van Durme", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Chris", | |
| "middle": [], | |
| "last": "Callison-Burch", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2013, | |
| "venue": "Proceedings of NAACL-HLT", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "758--764", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Juri Ganitkevitch, Benjamin Van Durme, and Chris Callison-Burch. 2013. PPDB: The paraphrase database. In Proceedings of NAACL-HLT, pages 758- 764, Atlanta, Georgia, June. Association for Compu- tational Linguistics.", | |
| "links": null | |
| }, | |
| "BIBREF11": { | |
| "ref_id": "b11", | |
| "title": "Corpus statistics approaches to discriminating among near-synonyms", | |
| "authors": [ | |
| { | |
| "first": "Mary", | |
| "middle": [], | |
| "last": "Gardiner", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Mark", | |
| "middle": [], | |
| "last": "Dras", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2007, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Mary Gardiner and Mark Dras. 2007. Corpus statistics approaches to discriminating among near-synonyms.", | |
| "links": null | |
| }, | |
| "BIBREF12": { | |
| "ref_id": "b12", | |
| "title": "More than words: Syntactic packaging and implicit sentiment", | |
| "authors": [ | |
| { | |
| "first": "Stephan", | |
| "middle": [], | |
| "last": "Greene", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Philip", | |
| "middle": [], | |
| "last": "Resnik", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2009, | |
| "venue": "Proceedings of human language technologies: The 2009 annual conference of the north american chapter of the association for computational linguistics", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "503--511", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Stephan Greene and Philip Resnik. 2009. More than words: Syntactic packaging and implicit sentiment. In Proceedings of human language technologies: The 2009 annual conference of the north american chapter of the association for computational linguistics, pages 503-511. Association for Computational Linguistics.", | |
| "links": null | |
| }, | |
| "BIBREF13": { | |
| "ref_id": "b13", | |
| "title": "The manually annotated sub-corpus: A community resource for and by the people", | |
| "authors": [ | |
| { | |
| "first": "Nancy", | |
| "middle": [], | |
| "last": "Ide", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Christiane", | |
| "middle": [], | |
| "last": "Fellbaum", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Collin", | |
| "middle": [], | |
| "last": "Baker", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Rebecca", | |
| "middle": [], | |
| "last": "Passonneau", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2010, | |
| "venue": "Proceedings of the ACL 2010 conference short papers", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "68--73", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Nancy Ide, Christiane Fellbaum, Collin Baker, and Re- becca Passonneau. 2010. The manually annotated sub-corpus: A community resource for and by the peo- ple. In Proceedings of the ACL 2010 conference short papers, pages 68-73. Association for Computational Linguistics.", | |
| "links": null | |
| }, | |
| "BIBREF14": { | |
| "ref_id": "b14", | |
| "title": "Gender and animacy knowledge discovery from web-scale n-grams for unsupervised person mention detection", | |
| "authors": [ | |
| { | |
| "first": "Heng", | |
| "middle": [], | |
| "last": "Ji", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Dekang", | |
| "middle": [], | |
| "last": "Lin", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2009, | |
| "venue": "Proceedings of the 23rd Pacific Asia Conference on Language, Information and Computation, PACLIC", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "220--229", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Heng Ji and Dekang Lin. 2009. Gender and animacy knowledge discovery from web-scale n-grams for un- supervised person mention detection. In Proceedings of the 23rd Pacific Asia Conference on Language, In- formation and Computation, PACLIC, pages 220-229.", | |
| "links": null | |
| }, | |
| "BIBREF15": { | |
| "ref_id": "b15", | |
| "title": "Predicting the readability of short web summaries", | |
| "authors": [ | |
| { | |
| "first": "Tapas", | |
| "middle": [], | |
| "last": "Kanungo", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "David", | |
| "middle": [], | |
| "last": "Orr", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2009, | |
| "venue": "Proceedings of the Second ACM International Conference on Web Search and Data Mining", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "202--211", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Tapas Kanungo and David Orr. 2009. Predicting the readability of short web summaries. In Proceedings of the Second ACM International Conference on Web Search and Data Mining, pages 202-211. ACM.", | |
| "links": null | |
| }, | |
| "BIBREF16": { | |
| "ref_id": "b16", | |
| "title": "What makes writing great? first experiments on article quality prediction in the science journalism domain", | |
| "authors": [ | |
| { | |
| "first": "Annie", | |
| "middle": [], | |
| "last": "Louis", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Ani", | |
| "middle": [], | |
| "last": "Nenkova", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2013, | |
| "venue": "TACL", | |
| "volume": "1", | |
| "issue": "", | |
| "pages": "341--352", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Annie Louis and Ani Nenkova. 2013. What makes writ- ing great? first experiments on article quality predic- tion in the science journalism domain. TACL, 1:341- 352.", | |
| "links": null | |
| }, | |
| "BIBREF17": { | |
| "ref_id": "b17", | |
| "title": "Name tagging with word clusters and discriminative training", | |
| "authors": [ | |
| { | |
| "first": "Scott", | |
| "middle": [], | |
| "last": "Miller", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Jethran", | |
| "middle": [], | |
| "last": "Guinness", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Alex", | |
| "middle": [], | |
| "last": "Zamanian", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2004, | |
| "venue": "HLT-NAACL 2004: Main Proceedings", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "337--342", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Scott Miller, Jethran Guinness, and Alex Zamanian. 2004. Name tagging with word clusters and discrimi- native training. In HLT-NAACL 2004: Main Proceed- ings, pages 337-342.", | |
| "links": null | |
| }, | |
| "BIBREF18": { | |
| "ref_id": "b18", | |
| "title": "Biber redux: Reconsidering dimensions of variation in american english", | |
| "authors": [ | |
| { | |
| "first": "Rebecca", | |
| "middle": [ | |
| "J" | |
| ], | |
| "last": "Passonneau", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Nancy", | |
| "middle": [], | |
| "last": "Ide", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Songqiao", | |
| "middle": [], | |
| "last": "Su", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Jesse", | |
| "middle": [], | |
| "last": "Stuart", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2014, | |
| "venue": "Proceedings of COLING 2014, the 25th International Conference on Computational Linguistics: Technical Papers", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "565--576", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Rebecca J. Passonneau, Nancy Ide, Songqiao Su, and Jesse Stuart. 2014. Biber redux: Reconsidering di- mensions of variation in american english. In Pro- ceedings of COLING 2014, the 25th International Conference on Computational Linguistics: Techni- cal Papers, pages 565-576, Dublin, Ireland, August. Dublin City University and Association for Computa- tional Linguistics.", | |
| "links": null | |
| }, | |
| "BIBREF19": { | |
| "ref_id": "b19", | |
| "title": "Linguistic models for analyzing and detecting biased language", | |
| "authors": [ | |
| { | |
| "first": "Marta", | |
| "middle": [], | |
| "last": "Recasens", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Cristian", | |
| "middle": [], | |
| "last": "Danescu-Niculescu-Mizil", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Dan", | |
| "middle": [], | |
| "last": "Jurafsky", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2013, | |
| "venue": "ACL (1)", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "1650--1659", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Marta Recasens, Cristian Danescu-Niculescu-Mizil, and Dan Jurafsky. 2013. Linguistic models for analyz- ing and detecting biased language. In ACL (1), pages 1650-1659.", | |
| "links": null | |
| }, | |
| "BIBREF20": { | |
| "ref_id": "b20", | |
| "title": "Using information content to evaluate semantic similarity in a taxonomy", | |
| "authors": [ | |
| { | |
| "first": "Philip", | |
| "middle": [], | |
| "last": "Resnik", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1995, | |
| "venue": "Proceedings of the Fourteenth International Joint Conference on Artificial Intelligence, IJCAI", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "448--453", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Philip Resnik. 1995. Using information content to eval- uate semantic similarity in a taxonomy. In Proceed- ings of the Fourteenth International Joint Conference on Artificial Intelligence, IJCAI, pages 448-453.", | |
| "links": null | |
| }, | |
| "BIBREF21": { | |
| "ref_id": "b21", | |
| "title": "Discovering implicit discourse relations through brown cluster pair representation and coreference patterns", | |
| "authors": [ | |
| { | |
| "first": "Attapol", | |
| "middle": [], | |
| "last": "Rutherford", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Nianwen", | |
| "middle": [], | |
| "last": "Xue", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2014, | |
| "venue": "Proceedings of the 14th Conference of the European Chapter of the Association for Computational Linguistics, EACL", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "645--654", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Attapol Rutherford and Nianwen Xue. 2014. Discov- ering implicit discourse relations through brown clus- ter pair representation and coreference patterns. In Proceedings of the 14th Conference of the European Chapter of the Association for Computational Linguis- tics, EACL, pages 645-654.", | |
| "links": null | |
| }, | |
| "BIBREF22": { | |
| "ref_id": "b22", | |
| "title": "The viability of web-derived polarity lexicons", | |
| "authors": [ | |
| { | |
| "first": "Leonid", | |
| "middle": [], | |
| "last": "Velikovich", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Sasha", | |
| "middle": [], | |
| "last": "Blair-Goldensohn", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Kerry", | |
| "middle": [], | |
| "last": "Hannan", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Ryan", | |
| "middle": [ | |
| "T" | |
| ], | |
| "last": "Mcdonald", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2010, | |
| "venue": "Human Language Technologies: Conference of the North American Chapter of the Association of Computational Linguistics, Proceedings", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "777--785", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Leonid Velikovich, Sasha Blair-Goldensohn, Kerry Han- nan, and Ryan T. McDonald. 2010. The viability of web-derived polarity lexicons. In Human Lan- guage Technologies: Conference of the North Amer- ican Chapter of the Association of Computational Lin- guistics, Proceedings,, pages 777-785.", | |
| "links": null | |
| }, | |
| "BIBREF23": { | |
| "ref_id": "b23", | |
| "title": "Recognizing contextual polarity in phrase-level sentiment analysis", | |
| "authors": [ | |
| { | |
| "first": "Theresa", | |
| "middle": [], | |
| "last": "Wilson", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Janyce", | |
| "middle": [], | |
| "last": "Wiebe", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Paul", | |
| "middle": [], | |
| "last": "Hoffmann", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2005, | |
| "venue": "Proceedings of the Conference on Human Language Technology and Empirical Methods in Natural Language Processing, HLT '05", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "347--354", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Theresa Wilson, Janyce Wiebe, and Paul Hoffmann. 2005. Recognizing contextual polarity in phrase-level sentiment analysis. In Proceedings of the Conference on Human Language Technology and Empirical Meth- ods in Natural Language Processing, HLT '05, pages 347-354.", | |
| "links": null | |
| }, | |
| "BIBREF24": { | |
| "ref_id": "b24", | |
| "title": "Paraphrasing for style", | |
| "authors": [ | |
| { | |
| "first": "Wei", | |
| "middle": [], | |
| "last": "Xu", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Alan", | |
| "middle": [], | |
| "last": "Ritter", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Bill", | |
| "middle": [], | |
| "last": "Dolan", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Ralph", | |
| "middle": [], | |
| "last": "Grishman", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Colin", | |
| "middle": [], | |
| "last": "Cherry", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 2012, | |
| "venue": "COL-ING", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "2899--2914", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Wei Xu, Alan Ritter, Bill Dolan, Ralph Grishman, and Colin Cherry. 2012. Paraphrasing for style. In COL- ING, pages 2899-2914.", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "FIGREF0": { | |
| "uris": null, | |
| "text": "Several paraphrases for money ranked according to automatically learned style dimensions.", | |
| "type_str": "figure", | |
| "num": null | |
| }, | |
| "FIGREF1": { | |
| "uris": null, | |
| "text": "Authors reveal style by choosing casual terms or formal terms for the same concept. Shown is a casual sentence (left) and a formal sentence (right) on the same topic. Alternative paraphrases are ordered casual (top) to formal (bottom).", | |
| "type_str": "figure", | |
| "num": null | |
| }, | |
| "TABREF1": { | |
| "content": "<table/>", | |
| "html": null, | |
| "text": "", | |
| "type_str": "table", | |
| "num": null | |
| }, | |
| "TABREF2": { | |
| "content": "<table><tr><td>agreed</td><td>\u2192 great</td><td>\u2192 sure</td><td>\u2192 yeah</td></tr><tr><td>assumes</td><td>\u2192 implies</td><td>\u2192 imagine</td><td>\u2192 guess</td></tr><tr><td>currently</td><td>\u2192 today</td><td>\u2192 now</td><td>\u2192 nowadays</td></tr><tr><td colspan=\"2\">most beautiful \u2192 very nice</td><td>\u2192 really nice</td><td>\u2192 really pretty</td></tr><tr><td>following a</td><td/><td/><td/></tr></table>", | |
| "html": null, | |
| "text": "\u2192 in the aftermath \u2192 in the wake \u2192 right after the man who \u2192 one who \u2192 the one that \u2192 the guy that", | |
| "type_str": "table", | |
| "num": null | |
| }, | |
| "TABREF3": { | |
| "content": "<table><tr><td/><td colspan=\"2\">Spearman \u03c1</td></tr><tr><td/><td colspan=\"2\">Formality Complexity</td></tr><tr><td>Inter-annotator</td><td>0.654</td><td>0.657</td></tr><tr><td>Log-ratio score</td><td>0.655</td><td>0.443</td></tr></table>", | |
| "html": null, | |
| "text": "Groups of paraphrases ordered from most formal (left) to least formal (right), as described in Section 4.1.", | |
| "type_str": "table", | |
| "num": null | |
| }, | |
| "TABREF4": { | |
| "content": "<table/>", | |
| "html": null, | |
| "text": "First row: mean correlation of each rater's scores with the average of the others. Second row: correlation of our automatic style score with the average human score.", | |
| "type_str": "table", | |
| "num": null | |
| }, | |
| "TABREF6": { | |
| "content": "<table><tr><td>: Pairwise accuracy for paraphrase pairs at varying</td></tr><tr><td>levels confidence. Top 10% refers to the 10% of pairs</td></tr><tr><td>with largest difference in log-ratio style score. Random</td></tr><tr><td>guessing achieves an accuracy of 0.5.</td></tr></table>", | |
| "html": null, | |
| "text": "", | |
| "type_str": "table", | |
| "num": null | |
| }, | |
| "TABREF7": { | |
| "content": "<table/>", | |
| "html": null, | |
| "text": "FormalLow \u03c3 whereupon, the proceedings were adjourned at 4:20 p.m.Formal High \u03c3 mr. president , you have 90 seconds Casual Low \u03c3 is she, what grade is she in? Casual High \u03c3 they bring to you and your loved ones. Complex Low \u03c3 let me abuse the playwright and dismiss the penultimate scene Complex High \u03c3 revealing to you my family 's secret because my late dad 's burial is over. Simple Low \u03c3 you 're not the only one Simple High \u03c3 facebook can get you fired , dumped , and yes , evicted", | |
| "type_str": "table", | |
| "num": null | |
| }, | |
| "TABREF8": { | |
| "content": "<table/>", | |
| "html": null, | |
| "text": "Style ratings of sentences with high and low levels of human agreement, measured by the variance of the human raters' scores (Low \u03c3 = high agreement).", | |
| "type_str": "table", | |
| "num": null | |
| }, | |
| "TABREF9": { | |
| "content": "<table><tr><td/><td colspan=\"2\">Formality</td><td colspan=\"2\">Complexity</td></tr><tr><td/><td colspan=\"2\">Sent. Genre</td><td colspan=\"2\">Sent. Genre</td></tr><tr><td colspan=\"2\">Inter-anno. 0.47</td><td>-</td><td>0.48</td><td>-</td></tr><tr><td>All words</td><td>0.44</td><td>0.77</td><td>0.43</td><td>0.80</td></tr><tr><td>PP only</td><td>0.18</td><td>0.63</td><td>0.23</td><td>0.45</td></tr></table>", | |
| "html": null, | |
| "text": ").", | |
| "type_str": "table", | |
| "num": null | |
| } | |
| } | |
| } | |
| } |