Lecture 6: 21 Sept.
Neural networks, Language models, word2wec
Slides
The slides have been corrected 22 Sept. after the recording:
- The bias node in the diagram is now 1 everywhere (and not -1 somewhere).
- A typo in slides 11 and 12 has been corrected: (y^-y) vs (y-y^).
Recording
Mandatory reading
Jurafsky and Martin, Speech and Language Processing, 3. ed. (edition of Oct. 2019)
- Ch. 3, "N-gram Language Models"
- Sections 3.0-3.4
- Ch. 6, "Word Semantics and Embeddings"
- Sections 6.8 and 6.9-6.12
- Ch. 7, "Neural Networks and Neural Language Models"
Lab-session 5, Tuesday 22 Sept. at Sed
Exercises