David Corne ( home)

Regarding the Web Intelligence and OAB exams

This is the first run of the Web Intelligence and OAB modules with my involvement (in half of the WI module and 1/3 of the OAB module). Since you have no resource of past papers containing questions about my material (and since everyone keeps asking me) this page contains some helpful (but not too helpful) information.

In each case, the exam has 4 questions, and you have to answer any three of them. On the WI paper, since I did half of the lectures, you can reasonably expect that there will often be TWO questions from me. On the OAB paper, you can reasonably expect that ONE of the questions will be on my material.

General notes about examinable material: 

Everything in the following list is EXAMINABLE:

• my powerpoint slides  ("DWC weeks" 1, 2 and 3  for both WI and OAB, weeks 4 and 5 for WI only)
• the SWANSON paper (WI and OAB)
• the SINKA paper (WI and OAB)
• the ADDITIONAL NOTES that accompanied assignment 3 (WI and OAB)
• the HITS paper (but only the parts of it indicated in the lecture slides) (WI only)

This means that my questions will assume that you have read and basically understand all of these things. HOWEVER, the understanding I expect concerning the papers is not complete and detailed. I just assume and expect a general and high-level understanding of these, and a grasp of the main points. For example, example part questions concerning the papers might be:

• Regarding the Sinka and Corne paper, what were the general findings concerning the use of stemming? [5 marks]   In answer to this, no need to remember the datasets involved, details of results etc ...; this asks about one of the general conclusions of the paper. Basically, stemming is often used in practice, but it does not always help -- it seems that stemming can sometimes be unhelpful when your system is trying to make distinctions between very similar types of documents.
• Regarding the Swanson et al paper, describe in broad terms the roles of Procedure I and Procedure II in the Arrowsmith system. [8 marks] Note that I would not ask for fine detail, such as system diagrams and pseudocode (UNLESS SUCH DETAIL WAS INCLUDED IN SLIDES); the question is asking for yur understanding of what basically happens in these procedures.

All other papers are not examinable, they are just recommended reading. As for these papers (the google paper, etc...), if you use knowledge gained from them appropriately within an answer to an exam question, that will probably impress me -- however this is the same for any appropriate general knowledge and understanding that you exhibit, which was not given in the examinable material.

Example part-questions: 

• Draw the complete graph on 5 nodes; what is its average node degree, and what is it's diameter? [5 marks]
• Briefly explain why the study of complex networks is important. [5 marks]
• [I might provide an example graph with about 10 nodes, and then ask]:  What is the degree distribution of this graph? What is the diameter of this graph? Etc ...
• Give the equation for the cluster coefficient of a network, and calculate it for the network below [here I might give a simple example network, like the ones in the slides] - [7 marks]
• Write clear  pseudocode for a procedure that can find the shortest path (i.e the shortest number of hops) between two nodes in a graph. [6 marks]
• In connection with a graph's degree distribution, write down the simple equation for the power law, and explain what can be inferred about the graph if you know the exponent of the power law distribution. [10 marks]
• Explain what `transitive closure' is with regard to articles on the web, and briefly discuss three challenges that must be solved before we can make progress towards achieving it. [15 marks]

     (WI only, of course):

• Write down and explain the equation for working out the PageRank of a web page. [7 marks]
• State and briefly explain/discuss three important design goals for a search engine. [8 marks]
• Describe the role of the Barrels in google's design, and explain why there are two types of barrels. [9 marks]
• Write pseudocode for the HITs algorithm [10 marks]
• Explain the assumptions that underpin Axelrod's model for Cultural Dissemination, and write pseudocode for the simple model itself. [10 marks]

`Pseudocode' just means any clear description of an algorithm or method. What I have in the examinable material (e.g. for HITS, for Google query evaluation, for graph search, and so on ...) is pseudocode.

 ------------------------------------------------------

This page last updated: just now.