2026-09-19T18:19:11Z https://www.ideals.illinois.edu/oai-pmh

oai:www.ideals.illinois.edu:2142/108068 2023-07-11 col_2142_5131 col_2142_10761 com_2142_5130 com_2142_10755 com_2142_234

Chang, Kevin Mohit Vyas, - 2020-08-26T21:58:09Z 2020-08-26T21:58:09Z 2020-05-15 2020-05 Integrating database (DB) and information retrieval (IR) technologies has long been an important problem. Recent growth in tagged textual data, which is a combination of structured knowledge bases and unstructured text data, has made principled index design for IR applications even more desirable. In this work, we propose a framework for designing IR applications using DB concepts. We model an IR application as a specialized DB system that is required to support only a fixed predefined query workload rather than general SQL queries. The physical design of an information retrieval system is optimized for a prespecified target query workload. We then identify IR indexes as basic building blocks of the design of an IR application. An IR index is formalized as a look-up function built over a materialized view. The overall ”index design problem” is then to find an index design with lowest cost from a space of candidate designs. Since the space of candidate designs can be doubly exponential in the size of query workload, we give polynomial time heuristic algorithm with provable guarantees for a weighted set cover based relaxation of the original problem. This is then combined with an overarching branch-and-bound based optimality preserving state-space search algorithm that efficiently prunes the state-space by using the above heuristic algorithm. Finally, we show via experiments how the proposed framework can be used to obtain index designs for different kinds IR applications in practice. Submission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2020-08-25 without embargo terms The student, - Mohit Vyas, accepted the attached license on 2020-05-14 at 16:54. The student, - Mohit Vyas, submitted this Thesis for approval on 2020-05-14 at 17:17. This Thesis was approved for publication on 2020-05-15 at 09:10. DSpace SAF Submission Ingestion Package generated from Vireo submission #15401 on 2020-08-25 at 17:14:51 Made available in DSpace on 2020-08-26T21:58:09Z (GMT). No. of bitstreams: 2 MOHITVYAS-THESIS-2020.pdf: 731191 bytes, checksum: 92ce098b27ad7ac162a0609bc0df49d1 (MD5) LICENSE.txt: 4209 bytes, checksum: a00f0ced8ea2eb69db426972b5c6cb60 (MD5) Previous issue date: 2020-05-15 application/pdf http://hdl.handle.net/2142/108068 en Copyright 2020 - Mohit Vyas Information Retrieval Databases Index Design Entity Search Index design for information retrieval applications using database concepts text Thesis Computer Science Computer Science University of Illinois at Urbana-Champaign Thesis M.S.