LBSC 708A Final Exam June 17, 1999 This exam is open book, open notes, and you may even use the Internet to search for information if you like. You may not receive help from any person, however. You have three hours to complete the exam, but each question can be answered adequately in one hour so you should not need all of that time. Do not spend more than half of your time on one question! I will be available to answer questions about this exam if you have any. 1. Answer ONE of the following questions: a. (Retrospective Retrieval) You have been asked to design a new system for retrospective retrieval of document images from a collection of old journal articles in English that have been scanned. You are free to design any type of system that you feel will meet the needs of professional researchers that wish to look for articles in the old journals. Explain each stage of the system model, including which features it will index, how the user will pose queries, how it will detect the documents that should be displayed to the user, how it will display that set of documents to the user in a way that facilitates selection, how users can examine individual documents, and how it will deliver documents desired by the user to them. b. (Information Filtering) In the coming years, people at ISTIC may receive more email than they have time to read. Some of this email will be in English and some will be in Chinese. You have been asked to design a system that can be used to help users cope with this flood of email by building one or more ranked lists for the user each morning that will allow them to examine their most important email first. Explain each stage of the system model for the system that you will build, including which features it will index, how the system will obtain and update each user's profile, how it will detect the documents that should be displayed to the user highest in the list, how it will display that set of documents to the user in a way that facilitates selection, how users can examine individual documents, and how it will deliver documents desired by the user to them. 2. Answer ONE of the following questions: a. (Overall System Evaluation) You have been asked to compare the performance of an existing Boolean text retrieval system with a new ranked retrieval system based on the vector space model that your organization is considering purchasing. Both systems handle the same documents (which are represented as electronic text), and both allow the user to search free text. State which aspects of each system will be important to evaluate, and explain how you would evaluate each in a way that would allow the results of your evaluation of each system to be meaningfully compared. b. (TREC-like Evaluation) You have been asked to design a TREC-like evaluation for ranked retrieval systems that are designed to find photographic images. Your evaluation will be expected to evaluate only the detection component of each system, and it is important that people be able to easily replicate the evaluation so that they can tell whether changes that they make have improved their system. State the measure (or measures) of effectiveness that you will adopt, and explain their strengths and weaknesses. Then explain how you will construct the queries, the collection of images, and the relevance assessments that you will use, and how the measure(s) of effectiveness that you chose will be computed. ---------------------------------- End -------------------------------------