What can we learn about the mental lexicon from non-prototypical cases on compounding
Time: 12:10pm - 01:10pm
Place: Psychology 128
Réka Benczes (Eotvos Lorand University)
English is remarkably abundant in nominal compounds whose meaning is based upon some sort of metaphor and metonymy. Examples include both lexicalized ones, such as couch potato (denoting a person who spends too much time before the television snacking on unhealthy food), and novel ones alike, such as muffin top (denoting the roll of spare flesh which cascades over the top of low-slung jeans). Nevertheless, such compounds have often been dismissed in morphological literature as semantically opaque – non-compositional – phenomena that are not formed on the basis of productive patterns. This bias can be traced back to the widely acknowledged and applied endocentric–exocentric distinction, which is still the dominant approach toward the semantics of compounds.
Cognitive linguistics, however, has demonstrated that these “exocentric” compounds are indeed analysable with the application of conceptual metaphor and metonymy on the one hand and blending theory on the other. Through the analysis of numerous examples, the talk will focus on how the everyday creativity of language that is inherent in metaphor- and metonymy-based compounds can complement a cognitive semantics-based word formation theory in order to have a better understanding of the structure of the mental lexicon.
In category: Morphosyntax and semantics
Perception of impoliteness in ‘Why don’t you shut up?’: Analysis of comments via Computer-Mediated Communication in the YouTube community
Time: 01:30pm - 02:30pm
Place: Woodburn Hall 002
Percepción de la descortesía en ‘¿por qué no te callas?’: Análisis de comentarios vía
CMC en la comunidad YouTube
[Perception of impoliteness in ‘Why don’t you shut up?’: Analysis of comments via Computer-Mediated Communication in the YouTube community]
Rosa Maria Piqueres
In category: Sociolinguistics and pragmatics
SWIFT Aligner: A Tool for the Visualization and Correction of Word Alignment and for Cross Language Transfer
Time: 03:00pm - 04:00pm
Place: Ballantine Hall (BH) 015
Olga Scrivner and Tim Gilmanov
It is well known that parallel corpora are valuable linguistic resources. One of the benefits of such corpora is that they allow for the building an annotated corpus for resource-poor languages via cross-language transfer. That is, given accurate alignment between a word from a source language and its equivalent in a target language, some linguistic information, such as part-of-speech tags or syntactic annotation, can be projected to the aligned word. While there are several state-of-the-art word-aligners, such as GIZA++ and Berkeley, there is no simple visual tool that would enable correcting and editing aligned corpora of different formats. We have developed Swift Aligner, a free portable software that facilitates the visual representation corpora, the correction of alignment and finally the transfer of morphological information and syntactic relations from an annotated source language into an unannotated target language, by means of word-alignment. In addition, this tool is flexible, as it imports corpora in various formats, such as GIZA++, Berkeley, and LIHLA. Finally, we have also shown that by using cross-language transfer, we would need only an estimated 30% of correction by human annotator, compared to 100% of manual annotation.
In category: Computational linguistics
Auditory free classification of nonnative speech
Time: 11:15am - 12:15pm
Place: Speech and Hearing Building, Room C141
Recent research on speech variability has found that listeners encode and integrate indexical features of speech (e.g., talker’s gender, age, dialect) with the linguistic information in speech. Furthermore, through repeated encoding, listeners build categories of indexical features (e.g., male/female, child/adult), similar to the categories of linguistic variables (e.g., phonemes and semantic classes). For English, a language that now has more nonnative speakers than native speakers across the world, foreign accents are indexical features that frequently introduce a significant amount of between-talker variability. The auditory free classification task—a task in which listeners freely group talkers based on audio samples—has been a useful tool for examining listeners’ perceptual representations of regional dialects, and is employed in the current studies to examine native and nonnative listeners’ representations of nonnative speech. In this talk, I present three experiments that address the following questions. (1) What are the salient features of nonnative speech for native listeners, and how stable is perception across different stimulus sets? (2) Does listeners’ perception change depending on whether they are asked about the general similarity of talkers or about the talkers’ native languages? (3) How does nonnative listeners’ perception of nonnative speech compare to that of native listeners? Results indicate that listeners—both native and nonnative—find nonnative talkers’ foreign accent strength to be a central organizational principle. Both listener and stimulus factors, however, also play important roles in further shaping listeners’ perception of nonnative speech. Specifically, I will discuss listeners’ attention to the talkers’ native language, listeners’ prior linguistic experience, and variability in the stimulus set, as relevant factors when perceiving nonnative speech.
In category: Second language acquisition
Computational cycles in (second) language processing: Cyclic versus non-cyclic integration in French
Time: 02:30pm - 04:00pm
Place: Ballantine Hall 205
Laurent Dekydtspotter, Mark Black, Rodica Frimu, Amber Panwitz, and Amandine Lorente-Lapole
Computational cycles promote efficiency by chunking information (Chomsky, 2005; Rizzi, 2013). However, the ability of (even advanced) second-language (L2) learners to compute wh-movement dependencies in real time is disputed (e.g., Clahsen & Felser, 2006a, b; Felser & Roberts, 2007; cf. Pliatsikas & Marinis, 2012). This questions the degree to which learners’ real-time processing benefits from efficiencies provided by language design. The role of language architecture in (L2) processing was therefore investigated in terms of the degree to which computational cycles (if any) guide the processing of noun-complements versus noun-modifiers along wh-movement dependencies as in (1).
(1) Quelle décision [à propos de lui(i,j) /le(i,j) concernant] est-ce que Paul(i) a dit pendant une semaine que Lydie avait prise sans hésitation?
‘Which decision about/concerning him did Paul say for a week that Lydie had made without hesitation?’
Despite similar referential possibilities, noun-complements and noun-modifiers interact with cyclic movement differently. The à-propos-de-lui (‘about-him’) noun-complement is computed on each cycle (2); whereas the le-concernant (‘concerning-him’) noun-modifier can be adjoined post-cyclically (3) (Lebeaux, 1988; Chomsky, 1995). Cyclic integration (2) induces V-domain binding of lui by Principle B. Non-cyclic integration (3) means that le is free and is contextually valued. Semantically, binding requires abstracting over assignments to variables. V-domain binding in chain-structure (2), therefore, delivers processing costs in the semantics. Our experiment, thus, targeted semantic processing load asymmetries in anaphora resolution induced by (non-)cyclic integration of noun-modifiers versus complements.
(2) [CP est-ce que Paul(i) a dit pendant une semaine [CP que [TP Lydie avait prise sans hésitation]]]
(3) [CP est-ce que Paul(i) a dit pendant une semaine [CP que [TP Lydie avait prise sans hésitation]]]
Binding-induced processing loads in the semantics were investigated in a priming experiment involving picture-probe classifications during a forced-paced reading aloud task (Miller, 2011). Visual and language processing interact within the semantic component (Jackendoff, 1987). Classification times on picture-probes processed concomitantly with relevant conceptual representations can reflect processing loads in the semantics. Written texts are familiar to L2-learners and reading aloud avoids comprehension questions that add semantic processing load. Reading-pace was set in DMDX (Forster & Forster, 2003) with 418-ms default duration per word plus an additional 16.7 ms per letter after 4 letters. Repeated measures were taken in two counter-balanced blocks each with 20 experimental items on a Latin Square and 30 fillers. Counterbalanced [± animate] picture-probes appeared for 500ms at the offset of pendant (control), que (C-domain) and envoyé (V-domain) and in varied positions. Sixteen advanced L1-English learners and sixteen native speakers were tested in the US and France respectively. In both groups, complement/modifier structures induced different classification times on V-domain probes, yielding structure-by-position interactions as expected (p<.05). Complement structures resulted in increased classification times for V-domain probes, reflecting binding-induced processing loads. Thus, these asymmetries suggest that (non-)cyclic integration of noun-modifiers and noun-complements regulated semantic anaphora resolution in French learners and native speakers.
In category: Second language acquisition