StatNLP

Logistics

Course Details

  • Course: COMPGI19, COMPM083.
  • Lecture: Friday 2PM - 5PM, MPEB 1.03
  • Lecturer: Sebastian Riedel
  • Lab: Tuesdays 8AM to 10AM, MPEB 1.03
  • Teaching Assistants: Tim Rocktäschel, Matko Bošnjak, George Spithourakis, Johannes Welbl
  • Office Hours: Friday 5:30PM - 6:30PM (and the walk from here to my office)

Course Details (cont'd)

  • I am not a linguist
    • I probably don’t know your favourite linguistic framework or terminology
  • I am not a cognitive scientist
    • I don’t know how humans process language
  • This course has no exams!
    • You should learn how to construct stat NLP systems
  • Requirements
    • You should know how to code, base knowledge in probability (Bayes Rule, MLE)

Scala?

  • Lectures, lab exercises and assignments focus on Scala
  • Scala
    • is typesafe, concise and efficient
    • has amazing support for processing of collections
    • encourages work with immutable data structures, but allows you to write procedural code where needed.
  • This is an opportunity to learn highly relevant technology (e.g. Scala backs Apache Spark and Twitter)

Really, Scala?

  • You are not required to be(come) an expert Scala programmer or master the OO/functional programming paradigm.
  • If you know functions, for loops, if clauses and linked lists you have almost all you need.
  • If you know classes and interfaces that's a plus, but we do expect you to pick basic knowledge during the course.

Assignments

  • Require programming, math, and analysis.
  • 3 Assignments, weigthed 30/40/30.
  • One project will require team-work.
  • Usual rules of plagiarism apply.
  • You need to submit ACL style PDF documents, no hand-writing

Course Material

  • We will be using the "stat-nlp-book" project.
  • Includes both lecture notes, and slides.
    • slides are summaries of the lecture notes, using the same source document.
  • References to other material are given in context.
  • This is work in progress, incomplete and contains errors.
  • Please contribute by adding issues on github when you see errors.
  • You can clone the project and create pull requests with your own contributions.

Discussion Forum

  • Our moodle page has a discussion forum.
  • Please post questions there (instead of private emails)
  • We aim to answer within 24 hours (except for weekends)
  • We give low priority to questions already answered in previous lectures, tutorials and posts.
  • We expect you to online-search for answers before.
  • You are highly encouraged to participate and help each other on the forum.

Recommended Book