The probabilistic language of thought
Time: 04:00pm - 05:00pm
Place: Psychology 101
Noah Goodman (Stanford University)Logic and probability are key themes of cognitive science that have long had an uneasy coexistence. I will argue that they can be reconciled by using the formalism of stochastic lambda calculus -- the result is a view of mental representation that is both compositional and probabilistic: a probabilistic language of thought. After introducing these formal ideas I will describe how this view helps explain human reasoning and intuitive theories. These case studies will include model words ('certain', 'plausible',..) in property induction, social cognition and natural language pragmatics, and causal reasoning. I will conclude by describing preliminary ideas about the mental process underlying inference in the probabilistic language of thought: sampling-based 'mental simulations' derived from algorithms of modern AI.
In category: Morphosyntax and semantics
A discussion about the acquisition of Chinese verb resultative complements and corresponding teaching strategies
Time: 04:30pm - 06:00pm
Place: Indiana Memorial Union Maple Room
Yongping Zhu (University of Mississippi)The structure of the Chinese verb resultative complement (VRC), which is composed of two verbs in which the first verb indicates the causal activity and the second one indicates the result, can be formed productively and is used extensively in Chinese. Since the semantic references of the verbs in the structure to the subjects and objects in sentences are various and complex, it is always difficult for students who study Chinese as a second language (L2) to acquire this structure. The purpose of this study is to find out the hierarchy of acquiring Chinese VRC and the appropriate teaching strategies for L2 learners through surveys on Chinese VRC and theoretical analysis of the surveys.
In category: Second language acquisition
Formalizing the left-chiastic simply-typed lambda calculus, or: What I did last summer
Time: 12:00pm - 01:00pm
Place: Memorial Hall 401
Wren ThorntonA couple years ago, in order to handle scrambling in Japanese, I introduced a new calculus for the semantic side of CCG derivations. This simple calculus is easy to describe and it can also be used to formalize keyword-arguments in programming languages, to justify notational shorthands in category theory, and numerous other things I keep stumbling upon. The calculus ties together a number of disparate threads in theoretical linguistics, type theory, logic, and category theory; and seems therefore worthy of future study. But one nagging question is: is it a nice place to do work? Does the calculus have the properties we usually desire from computational formalisms, like confluence and strong normalization?This summer I worked on answering these questions. Some have been answered in the affirmative. Others remain open, due to limitations in our usual techniques for solving them--- and perhaps due to limitations in our understanding of the phenomena in question. This talk relays some of that exploration. While it has more to do with type theory than with linguistics, I aim to convey why these questions are important and why formal theoretical linguists should care more about what theoretical computer scientists have been up to lately.
In category: Computational linguistics
Phonotactic selectional restrictions in Russian diminutives
Time: 05:30pm - 07:00pm
Place: Swain East 240
Maria Gouskova (New York University)Some affixes impose selectional restrictions on the phonological shape of the base: an affix can fail to attach to words that have certain phonological characteristics (e.g., in English, the verbalizing suffix -n attaches only to obstruent-final monosyllabic adjectives: stiff-en, black-en vs. *full-en, *calm-en, *aloof-en; Siegel 1974), or it might be realized as one of several allomorphs depending on the phonological context (e.g., the English indefinite determiner is "a" before consonant-initial words vs. "an" before vowel-initial ones; Paster 2004, Bye 2007, Wolf 2008). The exact formulation and role of selectional restrictions at the morphology-phonology interface has been controversial. How are selectional restrictions stated in the grammar? Are they enforced in parallel or derivationally, and if they are enforced derivationally, do they apply before affix realization or after? How exactly does blocking and allomorph selection work? In this talk, I develop the idea that selectional restrictions are phonotactic generalizations about the morphemes in the lexicon that can combine with an affix (Becker and Gouskova 2012). Unlike approaches using Generalized Alignment, this phonotactic approach predicts that properties not local to the affix can affect the choice or realization of the affix. I will also argue that selectional restrictions can be enforced both on the bases of affixation and on the words that result from affixation--that is, selectional restrictions can apply both before and after affixation has taken place. This approach will be tested on the case study of Russian diminutive suffixes, which impose a range of phonotactic constraints on the stress, syllable structure, and segmental makeup of the bases (Polivanova 1967, Kapatsinski 2011). The claims will be supported with evidence from an elicitation study.
In category: Phonetics and phonology
Full text citation analysis for scientific recommendation
Time: 12:30pm - 01:30pm
Place: Wells Library 001
Xiaozhong LiuWhile different citation analysis studies employed various sophisticated network analysis methods for scientific characterization, the basic assumption was easy and straightforward: either Publication1 cites Publication2, or Author1 cites Author2, regardless of sentiment, reason, topic, or motivation. More recent studies have shown, however, that this assumption is oversimplified. For this proposed research, by using citation context (extract from full-text data), we will characterize each scientific publication/venue/author along with each citation relation on a scholarly network differently by using labeled topic modeling method. More importantly, based on our experiment result, we found full-text citation can significantly enhance the scientific recommendation performance.
The Oc/Oil boundary in medieval France: Hard fact or ideological fiction?
Time: 03:00pm - 04:00pm
Place: Maple Room, Indiana Memorial Union
Anthony Lodge (University of St. Andrews) Histories of the French language invariably open with a map of the country neatly divided into three linguistic zones : French (langue d'oil), Occitan (langue d'oc) and Franco-Provençal. Are we to take the boundaries between these languages as reflecting real breaks in the dialect continuum, implying a severe loss of mutual intelligibility, or should we see them as just a convenient way of dividing up the subject, a methodological fiction? Traditional histories of French do not discuss the problem, for the reality of these boundaries, and their impermeability, are axiomatic. The municipal account-books of Montferrand in Basse-Auvergne (13th-14th centuries) provide a helpful base for exploring 'on the ground' the interface between the three languages of Gallo-Romance in the medieval period. Close examination of the language of these documents and of the life of the urban community which produced them shows no evidence of an Oc/Oil boundary. The boundaries so real to modern historians of French had no reality at all for the people of the time. How then are we to explain the tenacity of the three-language model in the minds of modern linguistic historians? The reasons for it are ideological rather than scientific, and they invite us to create a more holistic, less standard-oriented paradigm for studying the history of Gallo-Romance, which takes the fluidity and flexibility of medieval vernaculars more fully into account.
In category: Sociolinguistics and pragmatics
The syntax of modifiers in second language acquisition
Time: 03:15pm - 04:30pm
Place: Ballantine Hall 006
David StringerThis talk presents an overview of a research project concerned with the role of Universal Grammar in second language acquisition, specifically examining the acquisition of prepositional modifiers (e.g. "straight on through into the room"), adjectives (e.g. "a beautiful old wooden box"), and adverbs (e.g. "He probably no longer completely believes her"). Research questions include: (i) To what extent are modifier hierarchies universal? (ii) Does L1 transfer lead to distinct paths of development? (iii) Does natural language input suffice to ensure acquisition or is targeted instruction required? Universal hierarchies have been proposed for all three domains (Cinque, 1999, 2010; Laenzlinger, 2005; Larson, 1999; Scott, 2002; Shlonsky, 2004; Sproat & Shih, 1991; Stringer, 2005), but differences between languages raise learnability issues. The first part of the project, reported in Stringer, Burghardt, Seo, & Wang (2011), explored the question of whether learners of L2 English have knowledge of the syntax of prepositional modifiers even when such elements are not manifest in the L1. Results from a study with 121 ESL learners show awareness of the hierarchy irrespective of proficiency level, and constitute sound evidence for full access to Universal Grammar in L2 acquisition. The second part of the project, currently underway, investigates the acquisition of the more complex Adjective Ordering Restrictions (Stringer, Choi, Dlamini, & Martin, in prep). Experiments involved 204 ESL learners across 5 levels of proficiency. L1 Arabic, Korean, and Chinese were specifically examined for transfer effects, but contrary to expectations, the results suggest a lack of L1 influence. Knowledge of the nonabsolute-absolute distinction appears to be robust for all learners, but more fine-grained distinctions are poorly understood even at advanced proficiency levels. Thus while language universals are in evidence, they are insufficient for learners to converge on the target grammar, highlighting the need for enhanced input in language instruction.
JEvents v3.0.9 Stable
Copyright © 2006-2013
Powered by Joomla!®
Copyright © 2008 The Trustees of Indiana University | Copyright Complaints