Conferences and Calls for Papers
Department Reception and Awards
Colloquia and Talks
Spring Semester Reading Groups
Dissertation Defense
Robotics Open House

Midwest Computational Linguistics Colloquium

The Computational Linguistics Program of the Linguistics Department of Indiana University, the Cognitive Science Department of Indiana University, and the Linguistics Department of Indiana University are pleased to sponsor the inaugural meeting of the Midwest Computational Linguistics Colloquium (MCLC). We will build on the topics in formal-theoretical and computational linguistics, including, issues in learnability, complexity and the modeling of lexical and grammatical knowledge of language.

MCLC offers a less formal forum for computational linguistics researchers to get together and present ideas and research results. It is intended to be primarily for researchers from the Midwest region, but applications from all regions are welcome. We particularly welcome work in progress from graduate students and other young researchers. The first meeting of MCLC was successfully held at Indiana University, Bloomington in 2004, and we're pleased to bring it back to Indiana.

Schedule of Talks

Speaker	Title	Time
Benjamin Keil	Generating Descriptions of Motion from Cognitive Representations	9:15
Liu Liu, Gregory Aist, Jack Mostow	Generating Example Contexts for Vocabulary Words: Initial Steps	9:45
Markus Dickinson, Chong Min Lee, Soojeong Eom	Analyzing Korean Learner Particles	10:15
Coffee Break
DJ Hovermale, Dennis N. Mehay	Context-sensitive Spelling Correction for CALL	11:30
C. Anton Rytting, Sarah Wayland, Michael Maxwell, Pamela Kling	Spell Correction for Arabic Dialect Text and Speech Search	12:00
Emad Mohamed	Segment-based Arabic Spell Checker	12:30
Lunch Break
Emily Jamison	Using Online Knowledge Sources for Semantic Noun Clustering	14:30
Thomas Mathew, Graham Katz	Supervised Categorization of Habitual and Episodic Sentences	15:00
Roxana Girju	Mining the Web for Reciprocal Relationships	15:30
Break
Holger Wunsch, Sandra Kuebler, Rachael Cantrell	Instance Sampling Methods for Pronoun Resolution	16:00
Sunday 3 May
Invited Talk: Matthias Scheutz	TBA	9:00
Coffee Break
Nathan Sanders	Comparison of Phonological and Syntactic Distance Measures	10:30
John Pate, Chris Brew, Eric Fosler-Lussier	Assessing the Utility of Prosody for Word Segmentation in Phone Prediction	11:00
Wei-rong Chen	An Interface between Formal Phonology and Speech Technology: Speech Accent Coloring in Speech Synthesis	11:30
Lunch Break
Rodney L. Summerscales, Shlomo Argamon Jordan Hupert, Alan Schwartz	Identifying Treatments, Groups, and Outcomes in Medical Abstracts	13:30
Crystal Nakatsu	Contrastive Coverage Usage in the SPaRKy Restaurant Corpus	14:00

Other Conferences of Interest

Information about a wide range of other conferences of interest to IU Linguists can be found in the Linguistics Calendar Conferences Supplement. Please check this link for weekly updates on opportunities for conference attendance and paper submission in areas of interest to IU Linguists.

Department Reception and Awards

The annual Linguistics Department Reception and Awards Ceremony will be held on Wednesday 6 May 2009 in the department seminar room. Awards for faculty and student achievments will be given out around 12:40. Refreshments are provided, but attendees are encouraged to contribute their own dishes as well.

(Logic Series) What is the vector space content of what we say? A compositional categorical approach to distributed meaning

ABSTRACT: The two antipode approaches to semantics of natural languages are the symbolic and the distributed models of meaning. The former is type-logical and compositional, but does not say anything about meaning of words. The latter provides vector space meaning for words (based on their contexts), but says nothing about meaning of sentences. Developing a compositional distributed model of meaning has been a challenge to the computational linguists. Building on a proposal by linguists Clark and Pulman (Cambridge and Oxford), we merge these two approaches by using the axiomatics of compact closed categories and their diagrammatic toolkit of proofs. The key to this solution is our choice of Lambek's Pregroups as a syntax calculus, and the fact that both a Pregroup seen as a posetal category and the category of finite dimensional vector spaces are compact closed. We compute the meaning of a sentence by pre-composing the tensor product of the vectors of the words therein with the map of its syntactic structure. The pre-composition is a prescription on how to apply the units and counits of the adjunctions of the compact closed category. Surprisingly enough, the categorical Quantum Mechanics language of Abramsky and Coecke (Oxford) provides a nice intuitive explanation for the application of these maps: they create entangled Bell states (function abstraction) and take inner products (function application) and as such allow the information to flow among the words within a sentence. The use of two or more Bell states in the meaning map of negative sentences resembles the entanglement swapping protocols of quantum information.

(joint work with S. Clark, B. Coecke, and A. Preller)

(Computer Science) Aximo: implementation of a decision procedure for information update

ABSTRACT: In scenarios of multi-agent systems, agents communicate with each other to acquire new information. All of us take part in such scenarios when talking, emailing, and bidding or shopping on the net. The complexity of information flow in these scenarios, makes the development of an automated reasoner for them a must. Challenges arise in the closer-to-real-life versions of these scenarios, when the agents are not honest or the communication channel is not safe. These cause misinformation flow and increase the complexity of reasoning. I shall present a syntax and an automated decision procedure for proving epistemic properties of interactive scenarios of multi-agent systems. The novelty of the decision procedure is its use of the categorical rules of adjunction to reduce the epistemic and dynamic modalities. The implementation of the decision procedure, a C++ program called Aximo written by S. Richards, consists of a rewrite system and a recursive reasoner. I go through the termination and complexity of the program, show its soundness with regard to an algebraic axiomatics, and demo its applicability by proving properties of honest and also dishonest versions of the the muddy children puzzle as well as a coin toss scenario. I may hint on how the same procedure can be extended to reason about classical and quantum security protocol.

Computational Linguistics Lunch

Rehj Cantrel and Anthony Meyer will present on their semester projects for Markus Dickinson's fall semester L700 course. Additionally, Markus Dickinson and Marwa Hussein will present on the topic Annotating a learner corpus for SLA research. Their abstract follows:

ABSTRACT: We will presenting work on annotating a learner corpus of English in such a way as to make it useful for both second language acquisition (SLA) and computational linguistic (CL) research. We propose to annotate learner language with POS tags reflecting both linguistic form and use in the sentence, as well as dependency relations. The annotation is based on both native language corpora (e.g., SUSANNE) and first language acquisition corpora (e.g., CHILDES).

The architecture of grammar in artificial grammar learning: Formal biases in the acquisition of morphophonology and the nature of the learning task

ABSTRACT: The shape of the grammar that the learner of a language induces from the data s/he is exposed to has been assumed to be constrained by innate knowledge about what types of generalizations are found in grammars of natural human languages. By contrast, the possibility that the shape of the grammar also depends on the nature of the learning task faced by a person acquiring a language has been underexplored. This thesis exposes learners to the same artificial languages in two different training paradigms and shows that the nature of training can bias learners in favor of two different types of generalizations: product-oriented generalizations, which specify typical shapes of words belonging to a particular grammatical category, e.g., “plural nouns end in –s”, or source-oriented generalizations, which specify a mapping from one grammatical category to another, e.g., ‘add –s to the singular to form the plural’. However, we can also identify characteristics of grammars that hold across tasks and are also supported by natural language data. First, learners induce both source-oriented and product-oriented generalizations, for instance learning both what typical plurals are like and what segments of the singular must be retained in the plural. Second, learners rely on observations about what form classes and mappings are frequent in the data rather than relying on observations about which form classes and mapping are observed less often than expected. Third, competing generalizations are weighted relative to each other and obeyed in proportion to how much support each competitor has in the data, rather than the best supported competitor being obeyed 100% of the time. Fourth, learners do not induce the most specific generalizations consistent with the training data but rather overgeneralize in ways that are predictable if we make the Bayesian assumption that the output of perception is a probability distribution over possible percepts rather than the identity of the most likely percept.

This defense is in partial completion of the requirements for the degree of Doctor of Philosophy in Linguistics and Cognitive Science. It was prepared under the supervision of Drs. Kenneth DeJong of Linguistics and David Pisoni of Cognitive Science and Psychology. Vsevolod Kapatsinski came to IU in 2005 with an MA and BA in Linguistics from the University of New Mexico. More information about Mr. Kapatsinski can be found on his student website.

(Cognitive Science) Robotics Open House

The Cognitive Science Robotics groups led by Professor Randall Beer and Professor Matthias Scheutz invite you to a Robotics Open House, to be held on May 1, from 4:00-7:00pm on the 8th floor of the West Wing of Eigenmann Hall. The Open House will feature presentations about research going on in both groups and will include various live demonstrations on robots and in computer simulations. In particular, you will see CRAMER, the friendly interactive humanoid, an autonomous Segway platform moving to natural language commands, an autonomous robot photographer taking pictures, an autonomous walking hexapod controlled by a neural network, a robotic vocal tract producing natural sounds, a robotic head with attached arm showing binocular vision-guided reaching, a gantry robot, as well as various computer simulations of artificial agents performing tasks in simulated environments, and demos of gesture recognition and natural language processing. The exhibition is free and open to all, so please feel free to bring your family and friends. Refreshments will be served.