Natural Language Processing
Fall 2024
Announcement
- Lectures: Tues/Thurs 12:30–2pm in Wurster 102
- GSI Office Hours: 1–2pm Monday and 4–5pm Thursday, on Zoom (see Edstem for link)
- Professor Office Hours: 2–2:30pm after lecture
- Edstem link (only accessible to Berkeley accounts): https://edstem.org/us/join/RZpSp9 – contains links to bCourses, Gradescope, Kaggle, etc.
- This schedule is tentative, as are all assignment release dates and deadlines.
Schedule
- August 29
- Introduction
- Project 1 released by EOD Friday, August 30
- Sept 3
- Language modeling Statistical models | J&M 3
- Sept 5
- Language modeling Statistical models
- Sept 10
- Language modeling Representation learning and early neural models | E 3.0-3.3; G 1-5
- Sept 12
- Language modeling Neural models | J&M 7; Goldberg 10-11
- Project 1 due EOD Monday, September 16
- Sept 17
- Language modeling Recurrence
- Project 2 released
- Sept 19
- Language modeling Machine translation: neural approaches | J&M 13
- Sept 24
- Language modeling Machine translation: neural approaches pt.2
- Sept 26
- Language modeling Transformers
- Oct 1
- Core NLP tasks, data, and evaluation
- Oct 3
- Panel: philosophy of evaluation and data (slides)
- Project 2 due EOD Monday, October 7
- Oct 8
- Structured prediction Syntactic parsing | J&M 18
- Oct 10
- Structured prediction Syntactic parsing
- Oct 15
- Structured prediction Compositional semantics
- Oct 17
- LLMs LM recap, LLM architectures, inference, training objectives
- Oct 22
- LLMs Training, data, evaluation
- Oct 24
- LLMs Adaptation: RLHF, instruction-tuning, etc.
- Oct 29
- LLMs RLHF, scaling and efficiency, impacts
- Oct 31
- Panel: considerations in building/deploying LLMs (risks, harms, misuse)
- Nov 5
- Beyond text Speech | J&M 16
- Nov 7
- Beyond text Speech
- Nov 12
- Beyond text Pragmatics
- Nov 14
- Beyond text Vision and language
- Nov 19
- Beyond text Embodied interactive agents
- Nov 21
- Panel: NLP beyond text (speech vision, embodiment)
- Nov 26
- Panel: NLP beyond English
- Nov 28
- HOLIDAY
- Dec 3
- Panel: limitations
- Dec 5
- Panel: future of NLP