Identification of re-finding tasks and search difficulty

Sadeghi, S 2015, Identification of re-finding tasks and search difficulty, Doctor of Philosophy (PhD), Computer Science and Information Technology, RMIT University.

Document type: Thesis
Collection: Theses

Attached Files
Name Description MIMEType Size
Sadeghi.pdf Thesis application/pdf 1.63MB
Title Identification of re-finding tasks and search difficulty
Author(s) Sadeghi, S
Year 2015
Abstract We address the problem of identifying if users are attempting to re-find information and estimating the level of difficulty of the re-finding task. Identifying re-finding tasks and detecting search difficulties will enable search engines to respond dynamically to the search task being undertaken. To this aim, we conduct user studies and query log analysis to make a better understanding of re-finding tasks and search difficulties. Computing features particularly gathered in our user studies, we generate training sets from query log data, which is used for constructing automatic identification (prediction) models. Using machine learning techniques, our built re-finding identification model, which is the first model at the task level, could significantly outperform the existing query-based identifications. While past research assumes that previous search history of the user is available to the prediction model, we examine if re-finding detection is possible without access to this information. Our evaluation indicates that such detection is possible, but more challenging. We further describe the first predictive model in detecting re-finding difficulty, showing it to be significantly better than existing approaches for detecting general search difficulty. We also analyze important features for both identifications of re-finding and difficulties.

Next, we investigate detailed identification of re-finding tasks and difficulties in terms of the type of the vertical document to be re-found. The accuracy of constructed predictive models indicates that re-finding tasks are indeed distinguishable across verticals and in comparison to general search tasks. This illustrates the requirement of adapting existing general search techniques for the re-finding context in terms of presenting vertical-specific results. Despite the overall reduction of accuracy in predictions independent of the original search of the user, it appears that identifying “image re-finding” is less dependent on such past information. Investigating the real-time prediction effectiveness of the models show that predicting ``image'' document re-finding obtains the highest accuracy early in the search. Early predictions would benefit search engines with adaptation of search results during re-finding activities. Furthermore, we study the difficulties in re-finding across verticals given some of the established indications of difficulties in the general web search context. In terms of user effort, re-finding “image” vertical appears to take more effort in terms of number of queries and clicks than other investigated verticals, while re-finding “reference” documents seems to be more time consuming when there is a longer time gap between the re-finding and corresponding original search. Exploring other features suggests that there could be particular difficulty indications for the re-finding context and specific to each vertical.

To sum up, this research investigates the issue of effectively supporting users with re-finding search tasks. To this end, we have identified features that allow for more accurate distinction between re-finding and general tasks. This will enable search engines to better adapt search results for the re-finding context and improve the search experience of the users. Moreover, features indicative of similar/different and easy/difficult re-finding tasks can be employed for building balanced test environments, which could address one of the main gaps in the re-finding context.
Degree Doctor of Philosophy (PhD)
Institution RMIT University
School, Department or Centre Computer Science and Information Technology
Subjects Information Retrieval and Web Search
Human Information Behaviour
Computer-Human Interaction
Keyword(s) re-finding
search task
behavioral feature
predictive models
Version Filter Type
Access Statistics: 190 Abstract Views, 253 File Downloads  -  Detailed Statistics
Created: Fri, 14 Aug 2015, 12:41:01 EST by Keely Chapman
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us