Leivada, Fritz & Dentella (2024). Reply to Hu et al: Applying different evaluation standards to humans vs. Large Language Models overestimates AI performance

Autors:

Evelina Leivada, Fritz Günther & Vittoria Dentella

Títol:

Reply to Hu et al: Applying different evaluation standards to humans vs. Large Language Models overestimates AI performance

Editorial: PNAS 121(36), e2406752121 (National Academy of Sciences)
Data de publicació: 26 d'agost, 2024

Text complet

Dentella et al. (DGL) argued that 3 Large Language Models (LLMs) perform almost at chance in grammaticality judgment tasks, while revealing an absence of response stability (1). Hu et al.’s (HEA) “re-evaluation” led to different conclusions (2). HEA argue that i) “LLMs align with human judgments on key grammatical constructions,” ii) LLMs show “human-like grammatical generalization capabilities,” while iii) grammaticality judgments (GJs) are not the best evaluation method because they “systematically underestimate” these capabilities. While HEA’s aim to elucidate the abilities of LLMs is laudable, their claims are fraught with interpretative difficulties.

Salmons & Gavarró (2022). Intervention effects in Catalan agrammatism

Autors:

Io Salmons & Anna Gavarró

Títol:

Intervention effects in Catalan agrammatism

Editorial: Glossa: a journal of general linguistics (Open Library of Humanities)
Data de publicació: 31 de desembre del 2022

Text complet

The goal of the present study is to test the agrammatic comprehension of clitic left dislocation and contrastive focus in Catalan, in order to evaluate the Generalised Minimality hypothesis of Grillo (2009). According to this hypothesis, the comprehension deficit observed in agrammatism is the result of the underspecification of scope-discourse features giving rise to generalised intervention effects. We conducted two sentence-picture matching tasks to assess the comprehension of clitic left dislocation and contrastive focus with nine and seven Broca’s aphasic subjects, respectively, as well as control participants. The results show that the comprehension of SVO sentences and object clitics was preserved, whereas the comprehension of object dislocations and object focalisations was compromised. These findings are consistent with the analysis of the deficit as an instance of generalised intervention effects. Yet, we also examined the prediction that a relevant syntactic feature mismatch between the subject and the object would suffice to block generalised minimality effects; in particular, the number features of subject and object were controlled for. The agrammatic subjects’ performance on mismatched sentences did not differ from their performance on sentences where the subject and the object were matched in number. These findings call the hypothesis into question and stress the need for future research.

Tsiakmakis, Borràs-Comes & Espinal (2023). On the distribution and interpretation of voice in Greek anticausatives

Autors:

Evripidis Tsiakmakis, Joan Borràs-Comes i M.Teresa Espinal

Títol:

On the distribution and interpretation of voice in Greek anticausatives. Frontiers in Psychology

Editorial: Frontiers
Data de publicació: 23 de febrer del 2023
Pàgines: 15

Text complet

This paper provides experimental evidence in support of the view that Greek does not have three productive morphological classes of anticausative verbs, but only two: the class of verbs that bear non-active voice morphology and the class of verbs that are morphologically active. Across two experiments, native Greek speakers are found to prefer for each anticausative verb either non-active or active voice morphological marking, in the presence or absence of explicit contextual information. It is also shown experimentally that native speakers prefer an interpretation that involves a specific cause for all anticausatives, especially when the existence of such a cause is favored by the contextual setting. Our empirical findings are consistent with the view that the Voice Phrase that is realized as non-active voice morphology in Greek anticausatives is expletive. From a theoretical perspective, we analyze the expletiveness of this Voice projection as the result of semantic redundancy: the Voice head of Greek anticausatives combines with a v head that encodes a redundant cause meaning component and is, therefore, interpreted merely as introducing an identity function.

Pons-Moll, Torres-Tamarit & I.Mascaró (2023). Prosodically-driven morpheme non-realization in the Minorcan Catalan DP

Autors:

Clàudia Pons-Moll, Francesc Torres-Tamarit i Ignasi Mascaró

Títol:

Prosodically-driven morpheme non-realization in the Minorcan Catalan DP

Editorial: Linguistics, De Gruyter Mouton
Data de publicació: Abril del 2023

Més informació
Text complet


This article focuses on a case of prosodically-driven morpheme non-realization found in the Minorcan Catalan DP that has not been documented or accounted for in the previous literature. In this variety of Catalan, kinship restrictive appositive phrases show the realization of the masculine personal article if the following personal name starts with a consonant (es conco en Toni, the-def.art.m uncle the-pers.art.m Toni ‘uncle Toni’), but not if it starts with a vowel (es conco Àngel, the-def.art.m uncle Àngel ‘uncle Àngel’). This asymmetrical pattern is entirely unexpected since a preconsonantal coda is generated in the former cases (cf. es conco en Toni [əs.kòŋ.kun.tɔ́.ni]) and an onsetless syllable and a hiatus are generated in the latter (cf. es conco Àngel [əs.kòŋ.ku.án.ʒəl]). The structures obtained are indeed non-optimizing from a strictly syllabic point of view and stand as a clear case of antimarkedness. We argue that this asymmetric behavior is prosodic in nature, and is mainly driven by the need to keep the left edge of the prosodic word free of clitic material, without challenging basic syllabification constraints active in Catalan. We further justify how the cases under scrutiny shed new light on the typology of alignment constraints referred to the alignment between prosodic and lexical categories, on the nature of morpheme realization constraints, and on the morphosyntax-phonology interface.

Hamann & Torres-Tamarit (2023). Merger in Eivissan Catalan: an acoustic analysis of the vowel systems of young native speakers

Autors:

Silke Hamann & Francesc Torres-Tamarit

Títol:

Merger in Eivissan Catalan: an acoustic analysis of the vowel systems of young native speakers

Editorial: Phonetica (De Gruyter)
Data de publicació: 16 juny 2023

Més informació

The vowel system of Catalan has been the focus of many studies, though work on the varieties spoken on the island of Eivissa (Ibiza) are scarce, with a single mention of the possible merger of the mid back vowels /o, ɔ/ (Torres Torres, Marià. 1983. Aspectes del vocalisme tònic eivissenc. Eivissa 14. 22–23). The present article provides the first acoustic analysis of the vowel inventory of 25 young native speakers of Eivissan Catalan, with a focus on the realisations of stressed /ə, ɛ/, and the back mid vowels /o, ɔ/. We employed Pillai scores (Hay, Jennifer, Paul Warren & Katie Drager. 2006. Factors influencing speech perception in the context of a merger-in-progress. Journal of Phonetics 34. 458–484) to compare the possibly merged pairs /ə, ɛ/ and /o, ɔ/ to the fully-contrasting neighbouring pairs /e, ɛ/ and /o, u/. Our results show that all participants had considerable overlap of stressed /ə/ and /ɛ/, and all but one had considerable overlap of the back mid vowels, while the fully contrastive pairs (/e, ɛ/ and /o, u/) showed almost no overlap.