Spatial and temporal-based query disambiguation for improving web search

• Main Author: Kamal, Shahid
• Format: Thesis
• Language: English
• Published: 2016
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://eprints.utm.my/id/eprint/79030/1/ShahidKamalPFC2016.pdf; http://eprints.utm.my/id/eprint/79030/; http://dms.library.utm.my:8080/vital/access/manager/Repository/vital:106883

Queries submitted to search engines are ambiguous in nature due to users’ irrelevant input which poses real challenges to web search engines both towards understanding a query and giving results. A lot of irrelevant and ambiguous information creates disappointment among users. Thus, this research proposes an ambiguity evolvement process followed by an integrated use of spatial and temporal features to alleviate the search results imprecision. To enhance the effectiveness of web information retrieval the study develops an enhanced Adaptive Disambiguation Approach for web search queries to overcome the problems caused by ambiguous queries. A query classification method was used to filter search results to overcome the imprecision. An algorithm was utilized for finding the similarity of the search results based on spatial and temporal features. Users’ selection based on web results facilitated recording of implicit feedback which was then utilized for web search improvement. Performance evaluation was conducted on data sets GISQC_DS, AMBIENT and MORESQUE comprising of ambiguous queries to certify the effectiveness of the proposed approach in comparison to a well-known temporal evaluation and two-box search methods. The implemented prototype is focused on ambiguous queries to be classified by spatial or temporal features. Spatial queries focus on targeting the location information whereas temporal queries target time in years. In conclusion, the study used search results in the context of Spatial Information Retrieval (S-IR) along with temporal information. Experiments results show that the use of spatial and temporal features in combination can significantly improve the performance in terms of precision (92%), accuracy (93%), recall (95%), and f-measure (93%). Moreover, the use of implicit feedback has a significant impact on the search results which has been demonstrated through experimental evaluation.SHAHID KAMAL