The 50th Language Lunch
Date: 2015-12-10
Location: G.07 Informatics Forum
Northern Arizona: Sound Change and Dialect Contact
Lauren,Hall-Lew; LEL; Lauren.Hall-Lew@ed.ac.uk
Mirjam,Eiswirth; PPLS; s1322502@sms.ed.ac.uk
Mary-Caitlyn,Valentinsson; None; None
William,Cotter; None; None
This poster reports on the English short-a ‘nasal split’ in progress in Northern Arizona. Two subsets of acoustic data from 2002 were analyzed for social predictors of TRAP/BANsplit: MALE/FEMALEwas tested within an urban sample and URBAN/RURALwas testedamong males. TRAP-lowering and BAN-fronting show apparent-time correlations in the urban sample. Women lead in TRAPlowering and favor a backer TRAPand a higher BAN. This evidence suggests that urban Northern Arizona in 2002 was participating in the short-a nasal split, possibly Californian in origin. In contrast, a fronted TRAPvowel and a lack of a nasal split among ranchers is potential evidence that rural Northern Arizona in 2002 was more oriented to a Southern Vowel system than a Californian one. We suggest that Northern Arizona has been a site of dialect contact between these two major US English varieties.rn
Free adjuncts, matrix clauses and temporal coherence relations
James,Reid; None; None
Free adjuncts are considered noteworthy in the thetheoretical and descriptive literature in that the way they relate to their matrix clause must be arrived at by inferential means. The following examples (adapted from Kortmann 1991) demonstrate that the same free adjunct can relate to its matrix clause in multiple ways, depending on the content of the two.(1) Closing the window, John spotted a finch in the garden. [simultaneity](2) Closing the window, John turned to face his mother. [anteriority]Existing research on these constructions adduces a number of different features as being relevant for how they are interpreted temporally, including serial ordering, aktionsarten and world knowledge. In thispresentation, I outline a new accountframed within discourse coherence (specifically, an informal version of Segmented Discourse Representation Theory; Asher and Lascarides 2003 inter alia).This new analysis, which affords itself access to a more sophisticated theory of event structure (outlined in Moens and Steedman 1988), is able to capturethe range of temporal possibilitiesattested for complex sentences containing free adjunctsbutis able to do so in a more principles and explanatory way thanprevious analyses.Discourse coherence as a theoretical framework has been appliedto a number of cross-sentential phenomena, but analyses focusing on sub-sentential phenomena are extremely few.This work constitutes part of a wider research project on how theories of coherence might be used to shed light on sub-sentential pragmatic phenomena.rn
Reduplication facilitates early word segmentation
Barbora,Skarabela; None; None
Mits,Ota; None; None
This study tested the role of adjacent repetition in early lexical acquisition by comparing infants’ word segmentation of reduplicated versus non-reduplicated words. Twenty-four 9-month-olds were familiarized with two passages, one containing a novel reduplicated C1V1C1V1 word (e.g., ‘neenee’) and the other a novel non-reduplicated C1V1C2V2 word (e.g., ‘bolay’). A central fixation paradigm was then used to measure the infants’ looking times in response to four word types: 1) the familiarized reduplicated word, 2) the familiarized non-reduplicated word, 3) a newly introduced reduplicated word (e.g., ‘foofoo’), and 4) a newly introduced non-reduplicated word (e.g., ‘yahdaw’). Looking times were significantly longer for the familiarized reduplicated word compared to all other types of words. These results suggest that words consisting of repeated syllables are preferentially segmented in running speech. We discuss the implications of this finding for the role of perceptual biases in early lexical development and its relation to infant-directed vocabulary.rn
Partner perception and linguistic alignment in dialogue
Alessia,Tosi; None; None
Jarek,Lelonkiewicz; None; None
Holly,Branigan; None; None
In our study, we tested the hypothesis that the level of linguistic alignment in dialogue is modulated by how speakers perceive their conversational partner (e.g., Branigan et al., 2011).Sixteen pairs of English native speakers engaged in a Public-Goods Game, during which we manipulated participants’ perception of their interactional partner. Believing to be playing with each other, they actually played with a pre-set computer programme that was trying to maximize only its gains (individualistic) or the gains of both (collectivist). Subsequently, participants engaged in a joint discussion, this time with the real other.Using cross recurrence quantification analysis (CRQA), we measured the degree to which participants coordinated their language use. Consistently with previous findings in favour ofanaffiliative role forlinguisticalignment, we found some evidence that people who interacted with a collectivist (vs. individualistic) partner exhibited greater linguistic alignment on the subsequent discussion task. We discuss this finding in the context of previous research, as well as discuss some possible factors modulating this effect, i.e. mood and length of conversation.rn
Bilingual Word Alignment with Side Information
Clara,Vania; c.vania@ed.ac.uk
Current SMT systems rely (almost) exclusively on parallel data. There are around 7000 spoken languages in the word and for the vast majority of them there is almost nothing. However, there are various forms of human knowledge available on the Web. In this work, we explore how to improve word alignment quality by incorporating the bilingual dictionaries, word sense inventories, and morphological information which are available on the Web into the alignment model.rn
