Computer Laboratory

Course pages 2011–12

Machine Learning for Language Processing

Mark Gales' lectures:

Last year's lecture slides: are here

This years update lecture slides:

  • Lectures 1 and 2
  • Lecture 3
  • Lecture 4
  • Lecture 5
  • Lecture 6

    Instructions, Schedule and Reading List

    See Syllabus for lecture details and Assessment for assessment scheme

    L = lecture, S = Seminar (Three presentations per seminar)

    Please select three papers that you would like to present in order of preference by noon on Friday 20th January and email your selections to I will assign papers by 5pm that day.

    There will be 3 presentations per 50 minute seminar. Your presentations should be about 10-12 minutes allowing at least 3 minutes for questions and 5 minutes at the end of each seminar for general discussion. You should summarise the paper briefly (remember everyone will have read it), explicate any parts you found difficult or innovative, and critically evaluate the work described. For your evaluation you should consider questions like: To what extent have the stated aims of the research been achieved? To what extent is the work replicable given the information provided? In what way does the work advance the state of the art?, etc. You may prepare slides and use the data or overhead projector, and whiteboard. You should liaise with your co-presenters to decide the order in which to make presentations. The first presentation on each topic should briefly define the task, the others should not. You should have all slides for each seminar loaded onto a single laptop set up with the data projector by the beginning of each seminar.

    All students should read all the papers and come to all seminars prepared to discuss each paper after the presentations


    You may write an essay on a topic related to the paper you present, or any of the course material. Alternatively you may undertake a small project on text classification using existing datasets and machine learning software and submit a project report. In both cases, your essay or report should not exceed 5000 words and will be due in around the end of the first week of Easter Term.

    You should discuss and agree your essay topic or project with by email after the division of the Lent Term. Write a proposal of up to 500 words outlining the topic or project giving a preliminary reading list and indicating what resources you plan to use, if relevant. The first draft of your proposal should reach me by Monday, 5th March at the latest.

    Your essay topic should involve an in-depth critical evaluation of a specific machine learning technique and its application to language processing, or of a specific language processing task and machine learning techniques that have been applied to that task. Little credit will be given for summaries of papers. An example of a possible title/topic on named entity recognition might be `To what extent do we need sequential models to achieve accurate NER?' This essay might critically examine the claim made by Ratinov and Roth that NE recognition and classification can be done accurately by conditioning only on the class label assigned to the previous word(s) (as well as other invariant observed features of the context) without (Viterbi) decoding to find the most likely path of label assignments. In doing this, it might review the NER task definition and consider how dealing adequately with conjoined or otherwise complex NEs (Mazur and Dale) might affect their claims. It might also propose an experiment that would resolve the issue empirically and/or identify one that has been published that sheds some light on it.

    Suitable small projects will need to make use of existing labelled datasets and existing machine learning tools that are distributed and documented, so that they can be completed in reasonable time. Some examples of text classificataion tasks and datasets are: spam filtering (lingspam, genspam), sentiment of movie reviews ("sentiment polarity datasets" Pang), named entity recognition (conll shared task ner), hedge (scope) detection (conll shared task hedge scope), language identification (altw 2010 langid dataset), document topic classification (Reuters-21578), genre classification (genre collection repository), and many more. Some examples of (good) machine learning toolkits are SVMlight, WEKA, or Mallet. A project might replicate a published experiment but try different feature types or a different classifier, and describe the experiment and report results in a comparable manner to the relevant (short) paper.

    Week 1: Mark Gales, 23/1 L, 25/1 L, Classification by ML

    Week 2: Ted Briscoe, 30/1 S, 1/2 S, Document Topic Classification


  • First session (Me only): Lewis, Yang, Rose, Li, RCV1: A New Benchmark Collection for Text Categorization Research, JMLR, 2004

  • 1) Nigam & McCallum, A comparison of event models for naive bayes text classification, 1998
  • 2) Rennie, Shih et al. Tackling the poor assumptions of naive bayes text classifiers, ICML, 2003
  • 3) Rogati, Monica and Yang, Yiming, High-Performing Feature Selection for Text Classification, CIKM, 2002

    Week 3: Mark Gales, 6/2 L, 8/2 L, Graphical Models 1 \& 2

    Week 4: ; Ted Briscoe, 13/2 S, Spam Filtering; Mark Gales, 15/2 L, Graphical Models 3


  • 4) Sahami, Mehran et al, A bayesian approach to filtering junk email, AAAI, Wkshp on Text Classification, 1998
  • 5) Andoutsopoulos et al, An experimental comparison of naive Bayesian and keyword-based anti-spam filtering with personal e-mail messages, SIGIR, 2000
  • 6) Medlock, An adaptive, semi-structured language model approach to spam filtering on a new corpus, CEAS 2006

    Week 5: Mark Gales, 20/2 L, Graphical Models 4; Ted Briscoe, 22/2 S, NER 1


  • 7) Zhou & Su, Named Entity Recognition using an HMM-based Chunk Tagger, ACL02
  • 8) Klein et al, Named Entity Recognition with Character-Level Models
  • 9) Ratinov & Roth, Design Challenges and Misconceptions in NER, CoNLL 2009

    Week 6: Ted Briscoe, 27/2 S, NER 2; Mark Gales, 29/2 L, SVMs


  • 10) Mazur & Dale, Disambiguating Conjunctions in Named Entities, 2005
  • 11) Settles, Biomedical Named Entity Recognition Using Conditional Random Fields and Rich Feature Sets, 2004
  • 12) Vlachos, Tackling the BioCreative2 Gene Mention task with Conditional Random Fields and Syntactic Parsing, 2007

    Week 7: Ted Briscoe, 5/3 S, 7/3 S, Relation Extraction


  • 13) Aron Culotta and Jeffrey Sorensen, Dependency tree kernels for relation extraction, ACL04
  • 14) Pyysalo et al, A graph kernel for protein-protein interaction, BioNLP08
  • 15) Kate & Mooney, Joint Entity and Relation Extraction using Card-Pyramid Parsing, CoNLL10
  • 16) Mintz et al, Distant supervision for relation extraction without labeled data, ACL09
  • 17) Banko and Etzioni, The tradeoffs between open and traditional relation extraction, ACL08
  • 18) Greenwood and Stevenson, Improving Semi-Supervised Acquisition of Relation Extraction Patterns, 2006

    Week 8: Mark Gales, 12/3 L, Clustering; Ted Briscoe, 14/3 S, Topic/Term Clustering


  • 19) Griffiths, Steyvers, Finding scientific topics, PNAS 2004
  • 20) Andrezejewski, Zhu, Latent Dirichlet Allocation with Topic-in-Set Knowledge, ACL09 Wkshp Semi-supervised Lrng for NLP
  • 21) Hsu and Glass, Style & Topic Language Model Adaptation Using HMM-LDA, EMNLP 2006