FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2766339513> ?p ?o ?g. }

• W2766339513 abstract "Measuring Gradience in Speakers’ Grammaticality Judgements Jey Han Lau, Alexander Clark, and Shalom Lappin jeyhan.lau@gmail.com, alexander.clark@kcl.ac.uk, shalom.lappin@kcl.ac.uk Department of Philosophy, King’s College London Abstract dark). To sustain a categorical view of grammaticality it is necessary to show that when gradience does arise in speak- ers’ judgements, it is entirely the result of extra-grammatical influences, such as processing factors, semantic acceptabil- ity, or real world knowledge. To the best of our knowledge, the advocates of the categorical view have not managed to demonstrate this in a convincing way over large amounts of grammatically varied experimental data. Grammaticality is a theoretical concept, while acceptabil- ity can be experimentally tested. To maintain the categori- cal view of grammaticality in the face of pervasive gradience in acceptability judgements across a wide range of syntactic structures, one must provide an independent, empirically vi- able criterion for identifying grammaticality, which does not assume the view that is at issue. 1 In the absence of such a cri- terion, solid experimental evidence for gradience in accept- ability for a large number of speakers, across a wide range of data provides strong prima facie support for the hypothesis that grammaticality is a gradient property. Second, both sides of the debate have generally relied solely on linguistic judgements in order to motivate their con- clusions. In fact the discussion would be advanced if indepen- dent non-linguistic paradigms of binary classifiers and gradi- ent properties were identified and used as benchmarks with which to compare the judgement patterns that speakers ex- hibit with respect to a wide range of grammaticality data. We used Google statistical machine translation to map sen- tences randomly selected from the BNC into a number of languages and then back into English. The errors that were generated across these target languages ranged from mild in- felicities of lexical choice and awkward ordering of modi- fiers, through missing arguments, deleted prepositions, and misplaced subordinate clauses, to word salads. We then tested these sentences on speakers through crowd sourcing with Amazon Mechanical Turk (AMT). Each HIT (Human Intelligence Task) contained an original English sen- tence from the BNC and four translated sentences randomly selected from each of the four target languages. We used three presentation modes for these experiments: a binary classifica- tion task, a four point acceptability scale, and a sliding scale with 100 underlying discrete points. We observed a very high Pearson correlation among the mean judgements across all The question of whether grammaticality is a binary categorical or a gradient property has been the subject of ongoing debate in linguistics and psychology for many years. Linguists have tended to use constructed examples to test speakers’ judge- ments on specific sorts of constraint violation. We applied ma- chine translation to randomly selected subsets of the British National Corpus (BNC) to generate a large test set which con- tains well-formed English source sentences, and sentences that exhibit a wide variety of grammatical infelicities. We tested a large number of speakers through (filtered) crowd sourc- ing, with three distinct modes of classification, one binary and two ordered scales. We found a high degree of correlation in mean judgements for sentences across the three classification tasks. We also did two visual image classification tasks to ob- tain benchmarks for binary and gradient judgement patterns, respectively. Finally, we did a second crowd source experi- ment on 100 randomly selected linguistic textbook example sentences. The sentence judgement distributions for individ- ual speakers strongly resemble the gradience benchmark pat- tern. This evidence suggests that speakers represent grammat- ical well-formedness as a gradient property. Keywords: grammaticality, acceptability, gradient classifiers, binary categories, speakers’ judgements Introduction The question of whether grammaticality is a binary categor- ical or a gradient property has been the subject of a long- standing debate in linguistics and psychology (Keller, 2000; Manning, 2003; Crocker & Keller, 2005; Sorace & Keller, 2005; Fanselow, F´ery, Schlesewsky, & Vogel, 2006; Sprouse, 2007; Ambridge, Pine, & Rowland, 2012). While it has been recognized since the very beginning of modern linguistics (Chomsky, 1965) that there are degrees of grammaticality, in practice grammaticality is standardly taken in theoretical lin- guistics to be dichotomous: a binary division between gram- matical and ungrammatical sentences. Most grammar for- malisms are specified so that the grammars that they allow generate sets of sentences, with a binary set membership cri- terion corresponding to a categorical notion of grammatical- ity. Advocates of a categorical view of grammaticality have tended to limit themselves to experimental results involving a small number of constructed examples. These examples ap- pear to show the inviolability of specific kinds of syntactic constraints (such as wh-island conditions). While this work is interesting and important, it suffers from at least two prob- lems. First, it is difficult to see how the existence of a number of cases in which speakers’ judgements are robustly binary in itself entails the categorical nature of grammaticality, even when these cases exhibit clearly identifiable syntactic errors that are well described by a particular theory of syntax. Gra- dient judgments will inevitably appear to be sharp for clear paradigm cases (very tall vs. very short, very light vs very 1 For example, Sprouse (2007) presents some evidence that cer- tain types of syntactic island violations fail to show syntactic prim- ing effects under experimental conditions. He takes this to motivate a categorical view of grammaticality on the grounds that priming is only possible for grammatical sentences. He does not demon- strate that all types of ungrammaticality fail to exhibit priming ef- fects. Also, his experimental results suggest that speakers assign different levels of acceptability to distinct types of island violations." @default.
• W2766339513 created "2017-11-10" @default.
• W2766339513 creator A5032767467 @default.
• W2766339513 creator A5047100763 @default.
• W2766339513 creator A5074294896 @default.
• W2766339513 date "2014-01-01" @default.
• W2766339513 modified "2023-09-22" @default.
• W2766339513 title "Measuring Gradience in Speakers’ Grammaticality Judgements - eScholarship" @default.
• W2766339513 hasPublicationYear "2014" @default.
• W2766339513 type Work @default.
• W2766339513 sameAs 2766339513 @default.
• W2766339513 citedByCount "0" @default.
• W2766339513 crossrefType "journal-article" @default.
• W2766339513 hasAuthorship W2766339513A5032767467 @default.
• W2766339513 hasAuthorship W2766339513A5047100763 @default.
• W2766339513 hasAuthorship W2766339513A5074294896 @default.
• W2766339513 hasConcept C105795698 @default.
• W2766339513 hasConcept C111472728 @default.
• W2766339513 hasConcept C138885662 @default.
• W2766339513 hasConcept C15744967 @default.
• W2766339513 hasConcept C159717818 @default.
• W2766339513 hasConcept C189950617 @default.
• W2766339513 hasConcept C26022165 @default.
• W2766339513 hasConcept C2779525943 @default.
• W2766339513 hasConcept C33923547 @default.
• W2766339513 hasConcept C41895202 @default.
• W2766339513 hasConcept C5274069 @default.
• W2766339513 hasConceptScore W2766339513C105795698 @default.
• W2766339513 hasConceptScore W2766339513C111472728 @default.
• W2766339513 hasConceptScore W2766339513C138885662 @default.
• W2766339513 hasConceptScore W2766339513C15744967 @default.
• W2766339513 hasConceptScore W2766339513C159717818 @default.
• W2766339513 hasConceptScore W2766339513C189950617 @default.
• W2766339513 hasConceptScore W2766339513C26022165 @default.
• W2766339513 hasConceptScore W2766339513C2779525943 @default.
• W2766339513 hasConceptScore W2766339513C33923547 @default.
• W2766339513 hasConceptScore W2766339513C41895202 @default.
• W2766339513 hasConceptScore W2766339513C5274069 @default.
• W2766339513 hasIssue "36" @default.
• W2766339513 hasLocation W27663395131 @default.
• W2766339513 hasOpenAccess W2766339513 @default.
• W2766339513 hasPrimaryLocation W27663395131 @default.
• W2766339513 hasRelatedWork W107913646 @default.
• W2766339513 hasRelatedWork W1983199214 @default.
• W2766339513 hasRelatedWork W2020637909 @default.
• W2766339513 hasRelatedWork W2022496929 @default.
• W2766339513 hasRelatedWork W2054248262 @default.
• W2766339513 hasRelatedWork W2408311675 @default.
• W2766339513 hasRelatedWork W2483145526 @default.
• W2766339513 hasRelatedWork W2527961130 @default.
• W2766339513 hasRelatedWork W2725280480 @default.
• W2766339513 hasRelatedWork W2766468100 @default.
• W2766339513 hasRelatedWork W2801043313 @default.
• W2766339513 hasRelatedWork W2889192287 @default.
• W2766339513 hasRelatedWork W2966544378 @default.
• W2766339513 hasRelatedWork W2971078141 @default.
• W2766339513 hasRelatedWork W3004705763 @default.
• W2766339513 hasRelatedWork W3127937479 @default.
• W2766339513 hasRelatedWork W3163811323 @default.
• W2766339513 hasRelatedWork W3171945451 @default.
• W2766339513 hasRelatedWork W611164951 @default.
• W2766339513 hasRelatedWork W654876267 @default.
• W2766339513 hasVolume "36" @default.
• W2766339513 isParatext "false" @default.
• W2766339513 isRetracted "false" @default.
• W2766339513 magId "2766339513" @default.
• W2766339513 workType "article" @default.