NLP Applications II: Information Extraction, Question Answering, Recommender Systems and Conversational Systems
- Course on most relevant tasks for building NLP applications.
- Allow to understand
- When and how to apply NLP techniques in real-world scenario.
- Not only the use pre-existing NLP libraries,
- But be able to reimplement and adapt own models.
- Provide leads to explore and learn further
- Regular sessions: 13 sessions of 150 minutes
- Each session mixes theoretical and hands-on laboratories
- Extra session for the presentations!
- Material available in egela and google-drive
- Course is divided in three main parts:
- Information Extraction systems (Oier Lopez de Lacalle) 6 sessions in total
- Question Answering systems (Ander Barrena) 4 sessions in total
- Recommender System and Conversational systems (Mikel Larrañaga) 3 sessions in total
-
Information Extraction (Oier)
- Tuesday, March 15, 2022, 3:00 – 5:30pm
- Tuesday, March 22, 2022, 3:00 – 5:30pm
- Thursday, March 24, 2022,3:00 – 5:30pm
- Thursday, March 31, 2022, 3:00 – 5:30pm
- Tuesday, April 5, 2022, 3:00 – 5:30pm
- Wednesday April 6, 2022, 3:00 – 5:30pm
-
Question Answering (Ander)
- Thursday, April 7, 2022, 3:00 – 5:30pm
- Monday, April 25, 2022, 3:00 – 5:30pm
- Thursday, April 28, 2022, 3:00 – 5:30pm
- Monday, May 2, 2022, 3:00 – 5:30pm
-
Conversational Systems (Mikel)
- Wednesday, May 4, 2022, 3:00 – 5:30pm
- Wednesday, May 11, 2022, 3:00 – 5:30pm
- Monday, May 18, 2022, 3:00 – 5:30pm
- Laboratories are focused to put the theory in practice (no submission).
- You need to complete and submit 3 assignments.
- Assignment 1. IE: Intent-classification and Slot-filling
- Assignment 2. QA: QA+IR in open domain
- Assignment 3. CS: Recommender system
- Deadline for the assignments: 1st of June
- Main Project: on any open topic related to NLP application.
- Do the implementation, write-up a technical report (~6 pages), present in class.
- Presentations: 1st of June (to be confirmed)
- Deadline for the final report: 8th of June (to be confirmed)
- Basic programming experience, university-level course in computer science, experience in Python. Basic math skills (algebra or pre-calculus), but not much!
- Knowledge about machine learning or deep learning is required.
- Laboratories:
- Python (scikit-learn, pytorch, tensorflow…) using servers from Google Colaboratory
- Time might be tight => auto-study / finish labs at home / ask for help to lecturers
- Time might be plenty => review slides / do assignments
Class assignments: 50% of the grading
Final project: 50% of the grading
- Each group of student (2/3 people) will propose a subject for the final project to one of the lecturers, depending on his/her interests.
- Project proposal are due to May 19 (note that you will have 2 weeks for finishing!).
- The final project will be graded based on the written report, technicality and presentation, with the following percentages:
- write-up 15%, including features like clarity, structure, background, references, discussion
- technical 20%, incl. features like correctness and depth of the work
- poster presentation 15%, including clarity, structure, discussion