Syllabus

Introduction, Machine Translation

Lecturer: Jan Tore

Presentation

Mandatory reading

• Jurafsky and Martin, Speech and Language Processing, Ch. 25-25.2
• Koehn, Statistical Machine Translation (K:SMT), Ch. 1

Thu
25 Aug

Probabilities

Lecturer: Jan Tore

Presentation

Recommended reading

It is a good idea to repeat the parts from INF1080 Logic on "Kombinatorikk"

Machine translation evaluation

Presentation

Mandatory reading

• Jurafsky and Martin, Speech and Language Processing, Ch. 25.9
• Koehn, Statistical Machine Translation Ch. 8 up to sec. 8.2.3 BLEU

Probabilities and mathematical notation

Presentation

Exercises

• Probability
• Conditional probability
• Bayes¡¯ rule
• Independent events

Wen
7 Sep

MT evaluation, the noisy channel model, language models

Presentation

• Part 1 (revised 8 sep 1130)
• Part 2

Mandatory reading

• Koehn, Statistical Machine Translation Ch. 8, sec. 8.2 and 8.4, with errata page
• Jurafsky and Martin, Speech and Language Processing, Sec. 25.3-25.4
• Koehn, Statistical Machine Translation Ch. 4, sec. 4.3

Recommended reading

• Koehn, Statistical Machine Translation sec 8.3 (for they with sufficient background)
• Kishore Papineni et. al.: "Bleu: a Method for Automatic Evaluation of Machine Translation" in Fulltext.
• Chris Callison-Burch, Miles Osborne and Philipp Koehn: "Re-evaluation the Role of Bleu in Machine Translation Research". Fulltext

Work on obligatory assignment 1

Word-based models and alignment

Presentation

• pdf
• PowerPoint show with animations

Mandatory reading

• Jurafsky and Martin, Speech and Language Processing, Sec. 25.4-25.6
• Koehn, Statistical Machine Translation Ch. 4:
  • Sec 4.1-4.2 (except the technical details of 4.2.4)
  • Sec 4.4 (except technical details, to be explained)
  • Sec 4.5-4.6

More on alignment, higher-order models

Presentation

Completing obligatory assignment 1

Obligatory assignment 1 due 1800

Phrase-based alignment

Presentation

• First part
• Slides from K: SMT
  • Chapter 4, slides 49-51, 53-54
  • Chapter 5

Mandatory reading

• Jurafsky and Martin, Speech and Language Processing, Sec. 25.4, 25.6
• Koehn, Statistical Machine Translation Ch. 5, except
  • The technical details of sec. 5.3.3-5.3.6
  • Sec. 5.5

Work on obligatory assignment 2

Decoding

Presentation

• Slides from K: SMT
  • Chapter 6

Mandatory reading

• Jurafsky and Martin, Speech and Language Processing, Sec. 25.8
• Koehn, Statistical Machine Translation Ch. 6: sec 6.0-6.3

Program

Refinements

Presentation

• Main part
• Slides from K: SMT
  • Chapter 5, slides 17-21

Mandatory reading

• Koehn, Statistical Machine Translation Ch2,
  • Sec 2.1.1 Tokenization,
  • Sec. 2.3 Corpora
• Koehn, Statistical Machine Translation Ch9,
  • Sec. 9.0
  • Sec. 9.2, except 9.2.4
  • Sec. 9.3 up to "finding threshold points", p. 266
• Koehn, Statistical Machine Translation Ch10,
  • Sec. 10.0
  • Sec. 10.1.1 and 10.1.2
  • Sec. 10.2

Recommended reading

• Koehn, Statistical Machine Translation Ch2,

Alternative translation strategies

Presentation

Distributional Semantics: Extracting Meaning from Data

Wen

26 Oct

Introduction: linguistic foundations of distributional semantics

Lecturer: Andrei Kutuzov

Presentation

• Screen version
• Print version

Mandatory Reading

1. (Optional) Distributional Structure. Zellig Harris, 1954.
2. The Distributional Hypothesis. Magnus Sahlgren, 2008.
3. Speech and Language Processing. Daniel Jurafsky and James Martin. 3rd edition draft of April 9, 2016. Chapter 15, 'Vector Semantics'.
4. From Frequency to Meaning: Vector Space Models of Semantics. Peter Turney and Patrick Pantel, 2010. Skip Section 5.

Thu

27 Oct

Exercises on MT

Wed

2 Nov

Distributional and distributed: inner mechanics of modern word embedding models (including word2vec)

Lecturer: Andrei Kutuzov

Presentation

• Screen version
• Print version

Mandatory reading

1. (Optional) A neural probabilistic language model. Bengio, Yoshua, et al., 2003
2. Extracting semantic representations from word co-occurrence statistics: A computational study. Bullinaria, John A., and Joseph P. Levy, 2007
3. Distributed representations of words and phrases and their compositionality. Mikolov, Tomas, et al., 2013.
4. Word2vec parameter learning explained. Rong, Xin, 2014
5. Speech and Language Processing. Daniel Jurafsky and James Martin. 3rd edition draft of April 11, 2016. Chapter 16 `Semantics with dense vectors'.
6. (Optional) Glove: Global Vectors for Word Representation. Pennington, Jeffrey, Richard Socher, and Christopher D. Manning, 2014.

Thu

3 Nov

Wed

9 Nov

Practical aspects of training and using distributional models

Presentation

• Screen version
• Print version

Mandatory reading

1. Don't count, predict! A systematic comparison of context-counting vs. context-predicting semantic vectors. Marco Baroni, Georgiana Dinu, and Germ¨¢n Kruszewski. ACL 2014.
2. Improving Distributional Similarity with Lessons Learned from Word Embeddings. Omer Levy, Yoav Goldberg, and Ido Dagan. TACL 2015.
3. SimLex-999: Evaluating Semantic Models with (Genuine) Similarity Estimation. Felix Hill, Roi Reichart and Anna Korhonen. Computational Linguistics. 2015
4. (Optional) Correlation-based Intrinsic Evaluation of Word Vector Representations. Yulia Tsvetkov, Manaal Faruqui, and Chris Dyer. RepEval, ACL 2016
5. Word2vec in Gensim tutorial
6. (Optional) Vector representation of words in TensorFlow

Thu

10 Nov

Work on obligatory assignment 3.

Slides on setting up your isolated Python environment at IFI cluster.

Wed

16 Nov

Beyond words: distributional representations of texts

Presentation

• Screen version
• Print version

Mandatory reading

1. Distributed Representations of Sentences and Documents. Quoc Le, Tomas Mikolov. ICML 2014
2. Learning Distributed Representations of Sentences from Unlabelled Data. Felix Hill, Kyunghyun Cho, Anna Korhonen. arXiv:1602.03483, 2016
3. (Optional) Composition in Distributional Models of Semantics. Jeff Mitchell, Mirella Lapata. Cognitive Science, 2010
4. (Optional) Document Classification by Inversion of Distributed Language Representations. Matt Taddy. ACL 2015
5. (Optional) An Empirical Evaluation of doc2vec with Practical Insights into Document Embedding Generation. Jey Lau, Timothy Baldwin. ACL 2016

Thu

17 Nov

Wed

23 Nov

Kings and queens, men and women: semantic relations between word embeddings

Presentation

• Screen version
• Print version

Mandatory reading

1. Exploiting similarities among languages for machine translation. Tomas Mikolov, Quoc Le, Ilya Sutskever. arXiv:1309.4168, 2013
2. Diachronic Word Embeddings Reveal Statistical Laws of Semantic Change. William Hamilton, Jure Leskovec, Dan Jurafsky. ACL 2016
3. (Optional) Do Supervised Distributional Methods Really Learn Lexical Inference Relations? Omer Levy et al. NAACL 2015

Thu

24 Nov

Wed

30 Nov

What's going on: recent advances and trends in the word embeddings world

(+ exam information and discussion of the obligatory assignment)

Presentation

• Screen version
• Print version

Mandatory reading

1. (Optional ) Defining words with words: beyond the distributional hypothesis. Pontus Stenetorp et al. RepEval, ACL 2016.

Thu

1 Dec

Discussion of exam-like problems.

Attention: the room has changed to Datastue Fortress!

Thu

8 Dec

Discussion of exam problems in MT.

Attention: the room has changed to seminar room Pascal!