WS 2019/20

Neural approaches to parsing


Neural approaches to parsing

Winter semester 2019/20
Prof. Dr. Alexander Koller
Fri 10-12; C73 Seminar room

Start: Friday, October 18

At first glance, it seems obvious that syntactic parsing requires a grammar which represents the grammatically correct syntactic structures. However, over the past five years, models for syntactic parsing that are based on neural models have become much more accurate than grammar-based parsers.

In this seminar, we will look at current neural parsing models for constituency parsing, dependency parsing, and CCG. We will explore how neural models need to be designed and trained in order to achieve high parsing accuracy and look at recent work that analyzes what neural parsers learn about syntax.

You can find the list of papers for the seminar in this Google Sheet. Please have a look at the papers before you come to the first class to find out whether the seminar is the right level for you and which paper might interest you most.

Prerequisites. This seminar is suitable for MSc students and advanced students in the BSc Computerlinguistik. If non-German-speakers attend the seminar, it will be taught in English. The seminar assumes familiarity with neural networks (including LSTMs) and a solid background in constituency and dependency parsing (e.g. from Mathemathische Grundlagen 3 or Computational Linguistics). Most papers we will read are very recent (from the past 2-3 years), and you may have to read some of the papers they cite in order to understand some of their technical background. If this appeals to you, I look forward to working with you!

Grading. You will give a talk about a paper of your choice (60 minutes). You may also write a seminar paper in the term break.

If you choose to write a seminar paper, your overall grade for the course will consist of 40% talk grade + 40% grade for the seminar paper + 20% grade for the active participation in the in-class discussions and reading group sessions. If you do not write a seminar paper, your grade will be 66% talk grade + 34% participation grade.