Magda Oiry will give the talk “How Children Deal with Contextually Canceled Presupposition” in LARC’s meeting this week: Wednesday, February 10, at 12:15 in ILC N451. This is the talk she will present at the meeting of DGfS at the end of this month. An abstract can be found here.
07 February 2016
Sam Bowman of Stanford University will be giving the talk “Modeling Natural Language Semantics with Learned Representations” on Thursday, February 11, at 4PM in CS room 151. An abstract of his talk follows
The last few years have seen many striking successes from artificial neural network models on hard natural language processing tasks. These models replace complex hand-engineered systems for extracting and representing the meanings of sentences with learned functions that construct and use their own internal vector-based representations. Though these learned representations are effective in many domains, they aren't interpretable in familiar terms and their ability to capture the full range of meanings expressible in natural language is not yet well understood.
In this talk, I argue that neural network models are capable of learning to represent and reason with the meanings of sentences. First, I use entailment experiments over artificial languages to show that existing models can learn to reason logically over clean language-like data. I then introduce a large new corpus of entailments in English and use experiments on that corpus to show that these abilities extend to natural language as well. Finally, I briefly present ongoing work on a new model that uses the semantic principle of compositionality to more efficiently and effectively learn to understand natural language.
Leland Kusmer writes:
The results of the poll indicate that we're going to be reading Williams 2003 this semester! We'll start discussing the book this week — please try to read pp. 5-23 (up to section 1.5) for this meeting.
Our next meeting is this coming Thursday, the 11th; at that meeting, we'll hear from our visitor Polina and also discuss the part of the reading mentioned above. Location is TBA.
Variable Selection is Hard
Who: Howard Karloff
When: Feb 9, 1-2pm
Where: Computer Science Building, room 150/151
Consider the task of a machine-learning system faced with voluminous data on m individuals. There may be p=10^6 features describing each individual. How can the algorithm find a small set of features that “best” describes the individuals? People usually seek small feature sets both because models with small feature sets are understandable and because simple models usually generalize better.
We study the simple case of linear regression, in which a user has an m x p matrix B and a vector y, and seeks a p-vector x *with as few nonzeroes as possible* such that Bx is approximately equal to y, and we call it SPARSE REGRESSION. There are numerous algorithms in the statistical literature for SPARSE REGRESSION, such as Forward Selection, Backward Elimination, LASSO, and Ridge Regression.
We give a general hardness proof that (subject to a complexity assumption) no polynomial-time algorithm can give good performance (in the worst case) for SPARSE REGRESSION, even if it is allowed to include more variables than necessary, and even if it need only find an x such that Bx is relatively far from y.
This is joint work with Dean Foster and Justin Thaler and was done when all coauthors were at Yahoo Labs.
Bio: After receiving a PhD from UC Berkeley, Howard Karloff taught at the University of Chicago and Georgia Tech before leaving Georgia Tech to join AT&T Labs–Research in 1999. He left ATT Labs in 2013 to join Yahoo Labs in New York, where he stayed till February, 2015. Now he does data science for Goldman Sachs in New York.
A fellow of the ACM, he has served on the program committees of numerous conferences and chaired the 1998 SODA program committee. He is the author of numerous journal and conference articles and the Birkhauser book “Linear Programming.” His interests include data science, machine learning, algorithms, and optimization.
AIX SUMMER SCHOOL ON PROSODY 2016:
Methods in Prosody and Intonation Research: Data, Theories, Transcription
Laboratoire de Parole et Langage, Aix Marseille Université, France.
September 6-9, 2016
We are pleased to announce the "Aix Summer School on Prosody 2016 " tobe held from September 6-9, 2016 in Aix en Provence (France).
The Aix Summer School on Prosody 2016 will bring together experts ontheoretical and practical aspects of the research on prosody. The school will be organized around morning lectures and afternoon tutorials (where participants will practice concepts and skills discussed in lectures) and data clinics (where participants can bring together their own data and questions for discussion).
The school is intended for post-graduate students and researchers interested in all the theoretical and practical aspects of the research on prosody and intonation. The school will be suitable bothfor researchers already working on intonation and prosody, and wishingto learn more about specific topics, and for researchers who wish tobetter understand how to incorporate and control prosody in their own work.
Topics will include (but not limited to): theoretical models on prosody and intonation; perception of intonation; prosody and language pathologies; prosody, semantics, and discourse; prosody and L2;prosody and neurolinguistics; transcription of intonation and prosody;statistical methods in prosody research; and preparation of stimulifor perception studies.
The confirmed invited speakers are:
Amalia Arvaniti (Kent University): Theoretical models of prosody
Mariapaola D'Imperio (Aix-Marseille Université - IUF): Tonal alignment: perception and production
Laura Dilley (Michigan State University): Prosody and perception methods
Janet Fletcher (The University of Melbourne): Transcription of under-described languages
James Sneed German (Aix-Marseille Université): Prosody and discourse
Kiwako Ito (Ohio State University): Prosody and acquisition, prosody and eye tracking
Jelena Krivokapić (University of Michigan): Prosody and articulatory phonology
Oliver Niebuhr (University of Southern Denmark): Prosody and experimental methods in acoustics
Bert Remeijsen and Otto Gwado Ayoker (University of Edimburgh):Typology and field methods in prosody research: Shilluk prosody workshop
Annie Tremblay (Kansas University): Prosody and L2
Application deadline: May 31st 2016
For more information, please visit http://aixprosody2016.weebly.com/
A number of scholarships will be offered for PhD students andpostdocs. For more information about registration and scholarships, go to http://aixprosody2016.weebly.com/registration.html
*2nd Call for Papers*
*ESSLLI 2016 STUDENT SESSION*
Held during the 28th European Summer School in Logic, Language andInformation
Bolzano, Italy, August 15-26, 2016
*Deadline for submissions: March 10th, 2016
The Student Session of the 28th European Summer School in Logic,Language, and Information (ESSLLI) will take place in Bolzano, Italy, August 15th to 26th (http://esslli2016.unibz.it). We invite submissions of original, unpublished work from students in any area at theintersection of Logic & Language, Language & Computation, or Logic &Computation. Submissions will be reviewed by several experts in thefield, and accepted papers will be presented orally or as posters andselected papers will appear in the Student Session proceedings bySpringer. This is an excellent opportunity to receive valuable feedbackfrom expert readers and to present your work to a diverse audience.
Note that there are two separate kinds of submissions, one for oral presentations and one for posters. This means that papers are directly submitted either as oral presentations or as poster presentations. Reviewing and ranking will be done separately. We particularly encourage submissions for posters, as they offer an excellent opportunity to present smaller research projects and research in progress.
Authors must be students, and submissions may be singly or jointly authored. Each author may submit at most one single and one jointly authored contribution. Submissions should not be longer than 8 pages foran oral presentation or 4 pages for a poster presentation (including examples and references). Submissions must be anonymous, without any identifying information. More detailed guidelines regarding submissioncan be found on the Student Session website: http://www2.sfs.unituebingen.de/esslli-stus-2016/.
*SPONSORSHIP AND PRIZES*
As in previous years, Springer has kindly agreed to sponsor the ESSLLI student session. The best poster and best talk will be awarded Springerbook vouchers of 500€ each.
Please direct inquiries about submission procedures or other matters relating to the Student Session to firstname.lastname@example.org email@example.com.
ESSLLI 2016 will feature a wide range of foundational and advanced courses and workshops in all areas of Logic, Language, and Computation. For further information, including registration information and courselistings, and for general inquiries about ESSLLI 2016, please consult the main ESSLLI 2016 page: http://esslli2016.unibz.it.
There are a number of vacancies for PhD students and Postdocs in the inquisitive semantics group in Amsterdam. The projects are briefly described below. The deadline for applications is 15 March 2016, and the projects will start in the fall.
If you are interested in one of these positions, please get in touch with:
- Floris Roelofsen (firstname.lastname@example.org)
- Ivano Ciardelli (email@example.com)
Project 1 (PhD, 4 years): First-order inquisitive logic
This project is intended for a PhD student with a background in logic.Intended starting date: 1 September 2016.Apply through the UvA website.
One important novelty of the inquisitive approach is that it brings out the fact that questions have an important role to play in logic. For instance, suppose we know that Alice and Bob live in the same city. Then, information about Alice’s city of residence yields information about Bob’s city of residence. In inquisitive logic, this takes the form of an entailment:
Alice and Bob live in the same city, where Alice lives |= where Bob lives
Thus, in inquisitive logic we can reason not only, as usual, with particular pieces of information (e.g., that Alice and Bob live in the same city), but also with information types (e.g., Alice’s city of residence), which may be instantiated by multiple pieces of information (that Alice lives in Amsterdam, that Alice lives in Paris, etc). Thus, bringing questions into play leads to an exciting generalization of the fundamental notions of classical logic.
The aim of the project is to explore the consequences of this new perspective, carrying out a thorough investigation of first-order inquisitive logic. What makes this especially urgent is that an extremely broad range of questions are expressible by means of a first-order logical language.
Understanding how to reason with these questions is not only important from a mathematical and philosophical point of view, but also stands to have impact in applications, since question entailment generalizes the notion of functional dependency which is central to database theory and dependence logic. Thus, an axiomatization result would provide effective methods to reason about first-order dependencies.
Project 2 (PhD, 4 years): An inquisitive perspective on quantification in natural language
This project is intended for a PhD student with a background in formal semantics.
Intended starting date: 1 September 2016.Apply through the UvA website.
On the standard view, quantifiers like every man and many students are treated as operators that map properties to truth-values. While this approach allows for an insightful characterization of quantifiers’ truth-conditional contribution, it also has some important limitations. First, interrogative quantifiers like which men and how many students are squarely beyond its scope. Second, even quantifiers like every man are problematic when they occur in questions. For instance, to interpret the question What did every man eat? as ‘for every man x, what did x eat?’, the quantifier needs to scope out of the question; but this is impossible, if its argument is required to be a property.
Inquisitive semantics suggests a simple shift in perspective that simultaneously addresses both these issues. It becomes possible to pursue an account of quantifiers that captures (i) their potential to generate inquisitive content, which is necessary to deal with interrogative quantifiers like which men, and (ii) their ability to propagate inquisitive content, which is necessary to deal with quantifiers scoping out of questions.
We also expect inquisitive semantics to enable us to decompose quantifiers into semantically more primitive parts. A key observation in this regard is that, across languages, quantifiers are often built up from question words. We aim to explain this connection and to show how the meaning of quantifiers can be derived in a systematic way from the inquisitive elements they contain.
Project 3 (Postdoc, ~2.5 years): Applications in computer science and experimental ratification
This project is intended for one or two Postdocs with a background in logic and/or formal semantics/pragmatics, preferably with a particular interest in one or both of the following: (i) computational applications of logic in computer science, in particular in database systems, (ii) experimental semantics/pragmatics.
Intended starting date: 1 September 2016.
Apply through the UvA website.
Inquisitive semantics has so far been used to shed new light on theoretical issues in semantics, pragmatics, logic, and philosophy of language. The aim of this project is to broaden this scope in two directions. First, we will explore practical applications of inquisitive logic in computer science, in particular in database systems. Suppose we have a database and we get a new query. Then instead of looking at the database itself, we may look at the already stored queries, and see whether they entail the new query (in the inquisitive sense). If so, it may be possible to resolve the new query without having to re-consult the database itself, with a substantial gain in efficiency. This is a widely used technique in database systems, but inquisitive logic may make it possible to deal with a much wider range of queries than current systems do.
Second, we want to carry out a number of experiments to corroborate the empirical predictions that are made by linguistic theories formulated in inquisitive semantics, and to further develop these theories guided by the obtained empirical results. Relevant phenomena include (but are not restricted to) the interpretation of disjunction, questions, question-embedding verbs, and other operators that arguably generate and/or propagate inquisitive content, such as quantifiers, modals, and conditionals.
Project 4 (Postdoc, ~2.5 years): A fully compositional dynamic inquisitive semantics
This project is intended for a Postdoc with a background in formal semantics, preferably with a particular interest in compositionality and dynamic semantics.
Intended starting date: 1 October 2016.
Apply through the UvA website.
Most work on inquisitive semantics so far has focused on InqB, the inquisitive counterpart of classical first-order logic. In order to obtain a full-fledged inquisitive framework for natural language semantics, this basic system needs to be extended in two ways. First, we need to allow for meaning composition at the sub-sentential level. Second, we need to capture dynamic aspects of meaning, e.g., cross- and inter-sentential anaphoric dependencies and presuppositions. The goal of this project is to develop a fully compositional dynamic inquisitive semantics, and to apply this framework to a number of linguistic phenomena involving inquisitiveness below and beyond the sentence boundary.