Alumni Dissertations and Theses

 
 

Alumni Dissertations and Theses

Filter Dissertations and Theses By:

 
 
  • Prosodic Phrasing and Modifier Attachment in Standard Arabic Sentence Processing

    Author:
    Hala Abdelghany
    Year of Dissertation:
    2010
    Program:
    Linguistics
    Advisor:
    Janet Fodor
    Abstract:

    This dissertation investigates the syntax-prosody interface in Standard Arabic, focusing on the ambiguity of a modifier (relative clause or adjective phrase) in relation to the two nouns in a complex noun phrase. Ambiguity resolution tendencies for this construction differ across languages, contrary to otherwise universal parsing tendencies. One explanation proposed is Fodor's (2000) Implicit Prosody Hypothesis: that readers mentally project onto a text a default prosodic phrasing (possibly different between languages), which then influences their syntactic ambiguity resolution. Since implicit (silent) prosody cannot be directly observed, previous research has had to infer it by analogy with overt prosody. But the phonology and orthography of SA permit use of novel methods for tapping into the silent prosody of readers. Liaison phenomena sensitive to prosodic boundaries make phonological phrasing in SA very easy to detect. Also, liaison is indicated by diacritics in the `vowelized' version of SA orthography. Thus, clear data on prosodic phrasing patterns in SA complex nominals can be related to their preferred syntactic/semantic interpretations. Six experiments are reported: three production experiments and three perception experiments. Participants in Experiment 4 silently read sentences in unvowelized orthography, and added diacritics as they thought appropriate. The inserted diacritics gave evidence of their implicit prosodic phrasing of the sentence. Experiments 5 and 6 investigated Arabic speakers' preferred overt prosodic phrasing when the modifier was forced to attach to either the lower or the higher noun, providing standards for comparison with the prosodic phrasing preferences in silent reading in Experiment 4. The orthography was put to a different use in assessing modifier interpretation under varying prosodic conditions. In Experiments 2 and 3, vowelized text was presented, establishing one or other of two relevant prosodic patterns. Participants read aloud, and then indicated their interpretation of the sentence. This provided standards for comparison with modifier interpretation (attachment preferences) in silent reading of unvowelized texts (lacking prosodic disambiguation) in Experiment 1. Results obtained from these experiments provide new information concerning the constraints that apply at the syntax-prosody interface in SA, and also support the hypothesis of an effect of implicit prosody on syntactic interpretation during silent reading.

  • Consonantal voicing effects on vowel duration in Italian-English bilinguals

    Author:
    Ylana Beller-Marino
    Year of Dissertation:
    2014
    Program:
    Linguistics
    Advisor:
    Dianne Bradley
    Abstract:

    This project reported in this dissertation analyzes phonetic details of the speech patterns in one of New York's bilingual communities, asking whether a bilingual speaker can attain native-like proficiency in both languages and the extent to which authenticity — maintenance of language-specific settings — is sustainable. Researchers have established that Italian and English differ strikingly in their characteristic time settings for vowel durations: durations are greater for vowels preceding voiced consonants, e.g., cab, rather than voiceless, e.g., cap. This duration difference, termed the consonantal voicing effect (CVE), is notably greater for English than for Italian. The greater magnitude of the CVE found with English is considered to be a phonological enhancement of a basic phonetic process. Utilizing a speech production task, the study reported compares the performance of Italian-born bilinguals for whom English was acquired in adulthood, as a second language, with that of U.S.-born speakers who experienced simultaneous acquisition of their languages (albeit in an English-dominant setting). In separate sessions for each language, speakers produced utterances in which the target word, situated inside a carrier phrase, contrasted in [voice] value for the post-vocalic consonant, e.g., Say the word « ___ » to me. Stimuli were familiar words selected to sample the vowel inventories for each language and for which the voicing contrast was realized through the inventory of stops common to both languages. Analyses revealed no evidence of influence of the second language on the CVE for the first language for either group, despite an extended immersion period in an English-language environment for the foreign-born speakers and simultaneous exposure to both languages from birth for the U.S.-born speakers. But crucially, there was evidence of an influence of the first language in the timing settings found for the CVE in the second language, for both speaker groups: the foreign-born speakers managed to increase the magnitude of the CVE-English but failed to fully implement the phonological mechanism consistent with larger CVE values for that language; and the U.S.-born speakers managed to reduce the magnitude of the CVE-Italian but failed to fully suppress that same mechanism. Results are discussed in relation to language-specific timing patterns and the extent to which a dominant language may influence production in the non-dominant language.

  • Techniques for Automatic Normalization of Orthographically Variant Yiddish Texts

    Author:
    Yakov Blum
    Year of Dissertation:
    2015
    Program:
    Linguistics
    Advisor:
    Martin Chodorow
    Abstract:

    Yiddish is characterized by a multitude of orthographic systems. A number of approaches to automatic normalization of variant orthography have been explored for the processing of historic texts of languages whose orthography has since been standardized. However, these approaches have not yet been applied to Yiddish. Using a manually normalized set of 16 Yiddish documents as a training and test corpus, four techniques for automatic normalization were compared: a hand-crafted set of transformation rules, an off-the-shelf spell checker, edit distance minimization with manually set weights, and edit distance minimization with weights learned through a training set. Performance was evaluated by calculating the proportion of correctly normalized words in a test set, and by measuring precision and recall in a test of information retrieval. For the given test corpus, normalization by minimization of edit distance with multi-character edit operations and learned weights was found to perform best in all tests.

  • Systematic Asymmetries in Perception and Production of L2 Inflections in Mandarin L2 Learners of English: The Effects of Phonotactics, Salience, and Processing Pressure on Inflectional Variability

    Author:
    Timothy Bonner
    Year of Dissertation:
    2013
    Program:
    Linguistics
    Advisor:
    Gita Martohardjono
    Abstract:

    The study of language production by adults who are learning a second language (L2) has received a good deal of attention especially when it comes to omission of inflectional morphemes within L2 utterances. Several explanations have been proposed for these inflectional errors. One explanation is that the L2 learner simply does not have the L2 syntactic or prosodic representation in his grammar leading to omission of surface inflections (Hawkins & Liszka, 2003; Goad, White, & Steele, 2003, respectively). Others attribute L2 errors to mapping problems between the lexicon and syntax (Prévost & White, 2000; Lardiere, 1998, 2003). Another potential explanation for the variable production of inflectional endings is that it may be due to performance factors as in Hopp (2009) and Martohardjono, Valian, and Klein (2012) or to "Extra-syntactic" factors as proposed in Klein (2004) or to syllable repairs due to L1 phonotactic interference as proposed in Davidson (2005, 2006a, 2006b). This dissertation claims that when L2 morphosyntactic representations are shown to be available in the L2 learner's grammar, L2 inflectional variability can be attributed to L1 phonotactic interference, salience of the L2 inflection, and performance factors leading to systematic, but asymmetrical patterns of perception and production of the allomorphs that represent the surface L2 inflections. The results revealed that the target inflections were not omitted across the board as would be expected under deficit accounts. On the contrary, repairs of the final target coda clusters (i.e., schwa epenthesis before and after the final inflectional consonant and devoicing of the word-final consonants) revealed patterns that are consistent with the degree of syllabicity (e.g., [Vd] vs. [t] and [d]) and sonority (e.g., [s] vs. [t]) of the allomorph or coda and are not indicative of morphosyntactic deficits. Importantly, schwa epenthesis was applied asymmetrically (i.e., particularly to stops [t] and [d]) in clusters that contained target codas and inflectional allomorphs in real, nonce, and monomorphemic items, and thus, this repair pattern is contra the Prosodic Transfer Hypothesis of Goad et al. (2003). Overall, this dissertation presents an alternative explanation for L2 inflectional errors outside of the morphosyntactic and prosodic deficit arguments.

  • The Influence of Pseudo-relatives on Attachment Preferences in Spanish

    Author:
    David Branco-Moreno
    Year of Dissertation:
    2014
    Program:
    Linguistics
    Advisor:
    Marcel den Dikken
    Abstract:

    This paper presents the results from an off-line experiment on the extent to which the availability of pseudo-relatives modulates attachment preferences in Spanish. Participants were presented with sentences in which different syntactic and semantic factors had been manipulated to allow for either both a pseudo-relative (PR) and a relative-clause (RC) reading or a RC reading only. All the experimental items included two potential antecedents with which the constituents of interest could be associated. The experimental items can be divided into four groups: group 1 consists of stimuli allowing for a double reading in direct object position, and groups 2, 3 and 4 consist of stimuli containing RCs in prepositional complement position, preverbal subject position, and postverbal subject position, respectively. A stronger preference for the "higher" antecedent was expected in the first group of experimental items. The results indicate that the availability of pseudo-relatives seems to influence attachment preferences; however, the results ensuing from the statistical comparison of groups 3 and 4 need further investigation.

  • THE USE OF LE BY L1 CHINESE SPEAKERS AND THE ACQUISITION OF LE BY L2 CHINESE LEARNERS

    Author:
    CHI CHEN BREDECHE
    Year of Dissertation:
    2011
    Program:
    Linguistics
    Advisor:
    GITA MARTOHARDJONO
    Abstract:

    The perfective marker V-le is claimed to be one of the most problematic items in the acquisition of L2 Chinese, perhaps because no unified and comprehensive treatment of it exists in the literature. Although much has been written on this topic, the semantic and pragmatic functions of V-le have remained elusive. While linguists and grammarians all agree that V-le performs multiple functions, there has been no consensus on its meaning and usage. Adding to this complex situation is that V-le is not always syntactically obligatory in Mandarin Chinese. Even though scholars are well aware of the phenomenon of "optionality" with Chinese aspect markers, disagreement and ambiguity prevail. Smith (1997) claims that V-le, like other aspect markers in Chinese, is always syntactically optional. Li and Thompson (1981) claim that the use of V-le depends largely on the speaker's viewpoint as to whether an event is "bounded" or not. Both seem to suggest global optionality. In contrast, textbooks written for Chinese L2 learners discuss V-le as obligatory in various different contexts. Having found no established and unified treatment of V-le that reflects native speaker's knowledge regarding its obligatory and optional use, we conducted a larger survey of 482 native speakers, 316 adolescents and 166 adults, in an attempt to capture some generalizations on the obligatory and optional use of -le in various contexts. Our results show a range of frequencies, from very high (98% to 100%) in the context of accomplishment predicates and activity predicates as the first event in a sequence; to high (67% to 84%) in the context of achievement predicates; to variable (31% to 64%) in the context of resultative verb complements. We argue that this pattern of V-le suppliance can be derived by positing a hierarchy of boundedness in the predicate and that it follows a redundancy principle in discourse. The results from learners' data suggest that they acquired the knowledge of the perfective marking in the obligatory context after 300-400 hours of classroom instruction. They consistently used -le with accomplishment verbs and activity verbs as the first event in a sequence. They also consistently omitted -le with resultative verb complements, a hint of their implicit understanding of the semantic cues given by the predicates. In sum, the learners had a good understanding of the semantic properties of the verb class and had acquired a good, but not yet native-like, knowledge about the interaction between the perfective marker V-le and the lexical and semantic properties of different verb type categories.

  • The emergence of L1 innovations in Spanish-English bilinguals: Evidence from cross-linguistic structural priming

    Author:
    Agustina Carando
    Year of Dissertation:
    2015
    Program:
    Linguistics
    Advisor:
    Eva Fernández
    Abstract:

    This dissertation explores the hypothesis that structural priming is an internal mechanism motivating processes of convergence in bilinguals. The focus of the investigation is linguistic innovations in Spanish produced by Spanish-English bilinguals. Innovations involve both changes in the frequency of alternative constructions and existing patterns produced in new contexts modeled on English equivalents. From structural priming techniques that model convergence, the data assess the extent of English influence on Spanish, in a contact setting (New York, United States) and a non-contact setting (Córdoba, Argentina). In the field of language contact, convergence may manifest itself as an increase in the use of native language patterns shared with the contact language. Another outcome of convergence is grammatical replication, where native language structures acquire a new context of use resembling the contact language. Structural priming is the tendency for speakers to repeat previously processed structures. Cross-linguistic priming has been shown to increase the use of shared constructions; this investigation tests the applicability of priming to the study of grammatical replication. Three experiments examine the voice, reciprocal, and dative alternations. First, a picture description task in Spanish and English establishes baseline frequencies: the voice and reciprocal alternations have a similar distribution in English and Spanish; the dative alternation, however, differs between the two languages. Second, a within-language priming task (Spanish-to-Spanish) confirms strong priming effects for all three alternations and yields extremely low rates of grammatical replication. Third, a cross-language priming task demonstrates that English primes Spanish and increases grammatical replication rates, only with the alternations that have similar cross-linguistic distributions (voice, reciprocal). The priming effect did not differ between the contact and non-contact groups, but the bilinguals in the contact setting had higher grammatical replication rates. The data support the view that structural priming could be a catalyst facilitating language change in bilingual communities. We argue that this process is better explained with priming as implicit learning and suggest additional considerations. The data also support models of contact as an accelerant of processes already in motion in the native language, rather than as a trigger of the creation of completely new patterns.

  • Temporal Information Extraction and Knowledge Base Population

    Author:
    Taylor Cassidy
    Year of Dissertation:
    2014
    Program:
    Linguistics
    Advisor:
    Heng Ji
    Abstract:

    Temporal Information Extraction (TIE) from text plays an important role in many Natural Language Processing and Database applications. Many features of the world are time-dependent, and rich temporal knowledge is required for a more complete and precise understanding of the world. In this thesis we address aspects of two core tasks in TIE. First, we provide a new corpus of labeled temporal relations between events and temporal expressions, dense enough to facilitate a change in research directions from relation classification to identification, and present a system designed to address corresponding new challenges. Second, we implement a novel approach for the discovery and aggregation of temporal information about entity-centric fluent relations.

  • Non-standard Italian Dialect Heritage Speakers' Acquisition of Clitic Placement in Standard Italian

    Author:
    Lionel Chan
    Year of Dissertation:
    2014
    Program:
    Linguistics
    Advisor:
    Gita Martohardjono
    Abstract:

    This dissertation examines the acquisition of object clitic placement in Standard Italian by heritage speakers (HSs) of non-standard Italian dialects. It compares two different groups of Standard Italian learners--Northern Italian dialect HSs and Southern Italian dialect HSs--whose heritage dialects contrast with each other in clitic word order. The syntactic constructions tested include restructuring contexts (i.e., constructions in which clitic climbing can take place), and negative first- and second-person informal imperatives. The overarching research question guiding this pilot study is to determine what influences non-standard Italian dialect HSs' clitic placement when learning these constructions in Standard Italian. Three possible sources that may motivate these speakers' clitic placement in Standard Italian are considered: heritage non-standard Italian dialects; universal principles and dominant language transfer (English). A secondary research question of this study investigates whether there is a universal preference for encliticization. Participants completed two experimental tasks. The first was an Oral Elicited Imitation task that focused HSs' usage of clitics, whereas the second was a Grammaticality Judgment task that examined HSs' explicit knowledge of this property. The overall findings of this pilot study suggest that HSs parallel their heritage dialect clitic word order in their usage of Standard Italian, even though they are aware that another structure is possible in the standard dialect. The results also show only weak evidence to support a universal preference for encliticization, as suggested by the data gathered in previous studies (Bruhn-Garavito & Montrul 1996; Duffield & White 1999; Montrul 2010a; 2010b). A pedagogical implication based on this pilot study's findings is that when teaching standard dialect syntax, pedagogues should differentiate instruction based on learners' heritage non-standard dialectal background.

  • The Inner Workings of Text Summarization Systems

    Author:
    Hope Cotton
    Year of Dissertation:
    2014
    Program:
    Linguistics
    Advisor:
    Virginia Teller
    Abstract:

    THE INNER WORKINGS OF AUTOMATIC TEXT SUMMARIZATION SYSTEMS: MEAD AND SWESUM