1. Subject (Exam) Review WSTA 2015 Trevor Cohn

2. Exam Structure Worth 50 marks Parts: – A: short answer [14] – B: method questions [18] – C: algorithm questions [10] – D: short essay [8] 2 hours in duration … 2 minutes 24 seconds / mark

3. Short answer (14 marks) Several short questions – 1 or 2 sentence answers for each – look at the # marks, either 1 = very brief, 2 = longer/two-part question – answer both parts of two-part questions Often – definitional – conceptual – call for an example illustrating a technique/problem

4. Method questions (18 marks) Longer answer – larger questions 4-7 marks each – broken down into parts Focus on analysis and understanding, e.g., – contrast different methods – outline or analyze an algorithm – motivate a modelling technique – derive a mathematical equation

5. Algorithmic questions (10 marks) Perform algorithmic computations – numerical computations for algorithm on some given example data – present an outline of an algorithm on your own example Each question work 5 marks.

6. Essay question (8 marks) Expect to write 1 page Several topics given, you should select one – only one, no marks given for attempting many Provide – motivation: why it’s important, why it’s hard – present some approaches – contrast their strengths and weaknesses

7. Example questions Short answer: – What is an information need and how does this relate to a query? Use an example to justify your answer. [1 mark] Expect to give a concise answer, e.g., – The information need characterises what kinds of documents are relevant to the user (e.g., how many words there are in Inuit languages for ‘snow’), while the query is an approximate means of expressing this need to a search engine (‘eskimo snow words’).

8. Example Method Questions (B) Each question individually larger – may be broken down into steps Questions around 5 marks each, e.g., – Give the basic steps involved in positional index construction and querying … [2 marks] – What is the space and time complexity … Justify your answer. [2 marks] – Identify one disadvantage and one advantage a … has over a …. [1mark]

9. Example Algo Questions (C) Asked to do compute algorithm results, e.g., – numerical computation – illustration (drawing, examples) E.g., compute cosine, Rocchio – for example term/doc matrix – you won’t be required to simplify maths, i.e., you can leave things as fractions Conceptually easier, but can be slow going

10. Essay Question (D) Overview of a broad research topic As with the projects, requires – motivating the research problem, why it’s important and why it’s hard – describing some techniques – comparing and contrasting the methods – describing strengths and weaknesses Choice of 3 different topics

11. What’s fair game? Material covered in – the lectures – the core reading (including research papers cited) – invited lectures – the workshops – the projects

12. What’s not fair game? A few advanced topics, e.g., algorithms for – Lempel-Ziv compression – Unsupervised training for hidden Markov models – MEMMs & CRFs – Machine Translation But… you still need to understand – general concepts, motivation, broad outline

13. What’s not fair game? (cont) Guest lecture in final week – recommender systems (as this was not covered in workshops) Knowledge Technology content – despite overlap in subject content, there won’t be much overlap in examined content

14. What to expect Roughly half on each of – information retrieval / web search – text analysis Greater focus on concepts that have not yet been assessed by projects