June 10
| 9:00 to 9:15 |
Welcome and Introduction to TAG+10
|
| 9:15 to 10:45 |
Tutorial 1: Rajesh Bhatt (UMass Amherst)
Interpreting TAG Structures - Syntactic and Semantic Issues |
| 10:45 to 11:15 |
Coffee Break |
| 11:15 to 12:45 |
Tutorial 2: Marco Kuhlmann (Uppsala)
Relating TAG and Dependency Grammar. A Tree-Based Approach
|
| 12:45 to 2:15 |
Lunch Break |
| 2:15 to 3:45 |
Tutorial 3: Alexander Clark (Royal Holloway)
Grammar Induction and "Empiricist" Representations
This tutorial will cover recent advances in grammar induction, focusing particularly on distributional learning of context free and context-sensitive languages. Rather than defining a map from grammar to language, we invert the process and define a map from the class of languages to the class of representations, creating representations that depend on objective or "empiricist" features of the data. Once we have made this modelling assumption, the inference algorithms turn out to be quite simple. We will discuss two basic classes of algorithms: first a set of algorithms that assume that the non-terminals of a context free grammar should correspond to the congruence classes of the language; and secondly algorithms that are based on the syntactic concept lattice. Finally, we will discuss the relevance of this work to traditional learnability problems in theoretical linguistics.
|
| 3:45 to 4:15 |
Coffee Break |
| 4:15 to 5:45 |
Tutorial 4: William Schuler (Minnesota/Ohio State)
Incremental Parsing in Bounded Memory
This tutorial will describe the use of a factored probabilistic sequence model for parsing speech and text using a bounded store of three to four incomplete constituents over time, in line with recent estimates of human short-term working memory capacity. This formulation uses a grammar transform to minimize memory usage during parsing. Incremental operations on incomplete constituents in this transformed representation then define an extended domain of locality similar to those defined in mildly context-sensitive grammar formalisms, which can similarly be used to process long-distance and crossed-and-nested dependencies.
|
| 5:45 to 6:45 |
Refreshments |
June 11
|
|