Below are the final papers you'll be responsible for on the Final Exam, with some notes about each to allow you to focus on the important parts of them (the parts that you may have to answer questions about!). Two of the papers relate to Professor Tarasewich's lecture, plus his PPT slides (in PDF format) and two relate to Professor Aslam's lecture. Another is one Professor Futrelle discussed recently, along with a general introduction to clustering. All the papers below are cached on my CCIS teaching site for this course, so they're trivial to download from the links here. Three of them are web pages, with all images, also cached here.
You will have seen Professor Tarasewich's slides and should have downloaded and read the four papers related to their lectures. If you did not, download them now. Since the total amount of reading here is substantial, I don't expect you to have in-depth knowledge of any of them and I will adjust the Final Exam questions about the papers accordingly.
I will continue to discuss these papers as we wind up the course, so you'll hear more about them and be able to ask questions about them.
Here are Professor Tarasewich's slides (in PDF format). They are best understood in connection with the two papers below.
For this paper on speech user interfaces (2003), you should understand the basic points that there are more phones than internet users, so voice-driven retrieval could be quite important. You should also understand the gist of the system and experiment described, including clustering the documents to return a small enough number of classes that the user can keep track. There are short-term memory limitations, LTM -- spoken responses "evaporate", unlike information that can persist on a screen.
This paper tested three ways of presenting electronic documents for reading. You should understand the basics of the three types of views and carefully read the Discussion and Conclusion sections on page 299.
Here is a one page summary of Professor Aslam's lecture. He explained and compared straight filtering, e.g., for Spam, with collaborative filtering such as the suggestions you receive on Amazon.com about things readers "similar" to you have chosen to buy.
GroupLens is an important system that allows users of information to profit from the ratings that other users give to items. Read the paper to track down the fundamental notion that "people who agreed in the past are likely to agree again". Note that the work described here in this 1994 paper continues to this day, as described on the GroupLens homepage. "grouplens" garners 10,800 hits on Google - not a small number.
This paper on social information filtering takes a similar approach to the previous one, this time for music recommendations, attempting to automate "Word of Mouth".
There was an entire workshop in 2001 devoted to these "recommender systems" that included seven papers. You are not responsible for those papers, but I've included the link for completeness.
This paper on using part-of-speech patterns to reduce query ambiguity was discussed by Professor Futrelle. You only need to understand a few important points from this paper. First, there are different types of things about, say, a boat, that people might want to know, top right of page 307. Second, there are language patterns that indicate the type of information, as in the "party" and "boat" examples of Sec. 4.2. Finally, realize that their system responds to a one-word query such as "boat" by returning additional questions for the user to choose from, as in the "boat" questions, middle right on page 309. A query is then formulated behind the scenes to focus on documents with the specific class of information related to the question chosen.
This brief introduction to clustering contains some important general knowledge that is good to understand for many applications, not just for Information Retrieval. Professor Futrelle will go over this paper too.
Go to ISU535 home page. or RPF's Teaching Gateway or homepage