CSE 591: Advanced Topics in Computational Linguistics

Instructor: Niranjan Balasubramanian

Class Hours : Tuesday and Thursday 8:30 - 9:50 am

Office Hours: Tuesday 10:00 - 12:00 pm

Email: niranjan at cs dot stonybrook dot edu

Class Structure

The class will be a mix of presentations from me, paper presentations by the students and class discussions. Every class we will read two or three papers.

For the first lecture on each topic, I will introduce the topic and discuss one paper. This will be followed by two presentations by students (20 minutes each).

In the final lecture, I will lead a discussion that either:

• Summarizes other related work on the topic, open issues, and future directions (or)
• Provides an in-depth look at an algorithm and work out examples.

Presentations

These presentations are critical to the quality of the course and therefore, I would like to use some quality control and incentive mechanism. Here are the details:

• A team of two students will present each paper.
• Each presentation should be about 20 minutes with an additional 10 minutes for questions/discussions.
• We will follow a common high-level template. I will send this out in the first week of classes.
• All presentations will need to be sent to me x (TBD) days before the presentation date, so I have the time to give you feedback and suggest changes.
• The presentations will be graded and included for the final course grade.

Paper Readings

You are expected to turn in reports summarizing two papers before class on Thursday. Only the top 10 scores out of X reports you submit will be counted towards grade. This means you can turn in just ten summaries if you so choose.

Schedule

Date Topic	Papers to Read [N] - Niranjan, [S] - Student	Assignment/Project
Tue, Aug 25 Course Introduction + Syntactic Parsing Introduction	[N] [Slides] Introduction to Parsing.
Thu, Aug 27 Syntactic Parsing	[N] Introduction continued.
Tue, Sep 01 Dependency Parsing (Tree, Search, and Joint)	[N] [Slides] Accurate Unlexicalized Parsing [ACL 2003 Best Paper!] [N] [Slides] Transition-based dependency parsing with rich non-local features [P][S] [Slides from Shay Cohen's course.]Non-projective Dependency Parsing using Spanning Tree Algorithms [EMNLP 2005 Best Paper!]
Thu, Sep 03 Dependency Parsing (Neural)	[N] [Slides] Deep Learning for Efficient Discriminative Parsing [S] Neural CRF Parsing	Summary Report Due. Report to include all papers marked [S].
Tue, Sep 08 No Classes
Thu, Sep 10	[P][S] Parsing with Compositional Vector Grammars. [N] Summary + Discussion.
Wed, Sep 16 Semantic Role Labeling (Intro)	[N] Introduction [N] The Proposition bank: An annotated corpus of semantic roles. [N] Automatic Labeling of Semantic Roles (Cited by 1487!) [N] Semantic Role Labeling Via Generalized Inference Over Classifiers	[N] will issue a list of projects
Wed, Sep 23 Semantic Role Labeling (Structure + Scaling)	[S] Semi-supervised Semantic Role Labeling [S] Unsupervised Semantic Role Induction with Graph Partitioning.	Summary Report Due. Form teams for project. Propose project: A title and a two sentence description of the idea.
Wed, Sep 30 Semantic Role Labeling (Neural)	[S] Semantic Role Labeling with Neural Network Factors [S] End-to-end learning of semantic role labeling using recurrent neural networks [N] Introduction of Frame-Semantic parsing.	Summary Report Due. [N] will post Assignment 1.
Oct 7th, Wed	No class. N out of town.
Wed, Oct 14th Relation and Event Extraction (Intro)	[N] Introduction [N] Knowledge-Based Weak Supervision for Information Extraction of Overlapping Relations [N] Probabilistic Frame Induction
Wed, Oct 21st Relation and Event Extraction (Neural)	[N] Summarizing discussion for SRL. [N] Event Schema Induction with a Probabilistic Entity-Driven Model [S] A Dependency-Based Neural Network for Relation Classification [N] Mid-term discussion.	Summary Report Due.
Wed, Oct 28 Relation and Event Extraction (Neural)	[N] Exploiting Parallel News Streams for Unsupervised Event Extraction [S] Generative Event Schema Induction with Entity Disambiguation [S] Classifying Relations by Ranking with Convolutional Neural Networks	Summary Report Due.
Mon, Nov 2nd 5:30 -- 7:00 pm	Mid-term Exam 1 hour and 30 minutes.
Wed, Oct Nov 4th Relation and Event Extraction	[S] Relation Extraction with Matrix Factorization and Universal Schemas [S] Combining Distant and Partial Supervision for Relation Extraction [N] Summarizing relation and event extraction.	Summary Report Due.
Wed, Nov 11 Semantic Parsing (CCG Based)	[N] Introduction + CCG [N] Learning Synchronous Grammars for Semantic Parsing with Lambda Calculus [ACL 2007 Best Paper!] [S] Inducing probabilistic CCG grammars from logical form with higher-order unification
Wed, Nov 18 Semantic Parsing (Joint Models)	[N] Lexical generalization in CCG grammar induction for semantic parsing. [S] Joint Syntactic and Semantic Parsing with Combinatory Categorial Grammar [S] A-Star CCG Parsing with a Supertag-factored Model	Summary Report Due. Assignment 1 -- Part B due.
Wed, Nov 25 Textual Entailment and Question Answering	[S] Efficient Markov Logic Inference for Natural Language Semantics [N] Exploring Markov Logic Networks or Question Answering [N] Summarizing discussion.	Summary Report Due.
Wed, Dec 2 Project Presentations
Thu, Dec 3 Project Presentations [Last class.]
Tue, Dec 8		Project Report Due.
Thu, Dec 10
Tue, Dec 15
Thu, Dec 17