10 February 2014

Brendan O'Connor on text analysis today

Brendan O'Connor, CMU, will give a talk today, February 10, at
4 PM in Computer Science Building 150 entitled “Statistical Text Analysis for Social Science.” An abstract follows.

What can text analysis tell us about society? Corpora of news, books, and social media encode human beliefs and culture. But it is impossible for a researcher to read all of today's rapidly growing text archives. My research develops statistical text analysis methods that measure social phenomena from textual content, especially in news and social media data. For example: How do changes to public opinion appear in microblogs? What topics get censored in the Chinese Internet? What character archetypes recur in movie plots? How do geography and ethnicity affect the diffusion of new language? In order to answer these questions effectively, we must apply and develop scientific methods in statistics, computation, and linguistics.

In this talk I will illustrate these methods in a project that
analyzes events in international politics. Political scientists are
interested in studying international relations through *event data*:
time series records of who did what to whom, as described in news
articles. To address this event extraction problem, we develop an
unsupervised Bayesian model of semantic event classes, which learns the verbs and textual descriptions that correspond to types of
diplomatic and military interactions between countries. The model
uses dynamic logistic normal priors to drive the learning of semantic
classes; but unlike a topic model, it leverages deeper linguistic
analysis of syntactic argument structure. Using a corpus of several
million news articles over 15 years, we quantitatively evaluate how
well its event types match ones defined by experts in previous work,
and how well its inferences about countries correspond to real-world
conflict. The method also supports exploratory analysis; for example,
of the recent history of Israeli-Palestinian relations.

Second Call: ESSLLI

Models in formal Semantics and Pragmatics Workshop held at ESSLLI 26, August 18-22, Tuebingen, Germany

* Magdalena Kaufmann, University of Connecticut* Stefan Kaufmann, University of Connecticut

Invited Speakers:
* Michael Glanzberg, Northwestern University* Stanley Peters, Stanford University* Thomas Ede Zimmermann, Frankfurt University

Workshop Information:
Whatever happened to "model-theoretic" semantics?  Since Montague’s groundbreaking work and throughout much of its history, the field of formal semantics (and later pragmatics) was characterized by the use of models - abstract mathematical structures in which linguistic expressions are interpreted and which serve as the backdrop for stating generalizations about their semantic properties and relations.

Over the last couple of decades, however, the once-prominent status ofmodels has been eroding. In the research literature, explicitlydefined fragments and models were the norm in the early days (Partee1975, 1976; Dowty, 1979), but are now the exception rather than therule. In teaching, one of the most widely used textbooks, Heim andKratzer (1998), makes no mention of models, in stark contrast withearly standard works like Dowty, Wall and Peters (1981). Aside fromsuch signs of waning interest, there is a small but formidable body ofwork which actively questions the status of models and finds them tobe of limited use at best (Lepore 1983; Higginbotham 1988; Zimmermann 1999, 2011; Glanzberg, t.a.).
Such explicit reflections are rare, however. The overall decline ofmodels in the field is not driven by a general debate, let aloneconsensus. Nor is the turn away from models a turn towards somenon-model-theoretic alternative. What we do see instead is a tendencyto stay loosely within the model-theoretic framework, but to enrich itwith notions and tools whose formal properties remain largely implicit.

The goal of this workshop is to promote and generate discussion of thepast, present, and future of models in natural-language semantics andpragmatics, specifically the implications of their apparent demise forthe foundations and goals of the field. Topics for discussion include,but are in no way limited to the following:

* What are models, anyway?  Commitments about language, reality, and the nature of meaning that a model-theoretic approach to semantic analysis implies. The (special?) status of possible worlds and their relationship to extensional models.

* What are models good for? Linguistic phenomena or aspects of meaning in whose analysis a model-theoretic approach has been, or would be, crucial or at least beneficial. The (potential) use of models         in treating meaning as variable (e.g., in the analysis of uncertainty about language, or in cross-linguistic and diachronic comparative semantics).

* Where do models get in the way? Desiderata for semantic theory and limitations of the model-theoretic approach. Risks and side effects of specific methods associated with the model-theoretic approach (e.g., meaning postulates).

* Are we safe without models? Advantages and potential pitfalls of innovative uses of formal techniques or metalinguistic expressions, whose repercussions are underexplored (various kinds of states and events, partial functions, etc.)

* What are the alternatives? 

Workshop Format: 
The workshop is part of ESSLLI and open to all ESSLLI participants. It will consist of five 90-minute sessions held over five consecutive days in the second week of ESSLLI. The three invited talks are allotted one hour each, including discussion. On the first day, the workshop organizers will give a 30-minute introduction to thetopic. This leaves room for eight submitted papers of 30 minutes each,including discussion.

Submission Procedure:
Authors are invited to submit an abstract of up to three pages,including examples and/or references (single-spaced, at least 11 pt font, on US letter of A4 paper with margins at least 1in or 2.5cm on all sides, in .pdf, .txt, .doc or .odt format). Abstracts must be submitted by February 15, 2014, electronically at the following address: http://linguistlist.org/easyabs/Models_ESSLLI14