Benue State University Institutional Repository >
Faculty of Science >
Mathematics / Computer Science >
Please use this identifier to cite or link to this item:
|Title: ||IMPLICIT FEEDBACK SYSTEM FOR THE RECOMMENDATION OF RELEVANT WEB DOCUMENTS|
|Authors: ||AKUMA, STEPHEN S.C.|
|Keywords: ||RECOMMENDATION, DOCUMENTS|
IMPLICIT FEEDBACK SYSTEM, RELEVANT WEB
|Issue Date: ||1-Apr-2021|
|Abstract: ||The web is constantly updated with information, leading to the problem of information overload. The generic information retrieval systems retrieve a relatively large amount of irrelevant web resources, forcing information seekers to spend a significant amount of time searching for their information needs. Learners of a particular domain are usually faced with common problems and they visit similar sources of information when searching for their information needs.
The key idea of this research is to capture and record previous web documents visited by a homogeneous group of learners with their associated user behaviour which is inferred from the reading time, mouse and ikey activity, and to utilise this information to optimise the recommendation of relevant documents to users.
Several user studies were conducted to investigate relevance feedback parameters. An investigation was carried out to examine the relationship between user-generated implicit indicators and user perception of relevance. Thirteen users were given fifteen web documents to read and rate according to their perception of relevance based on given tasks. The results show a positive co-relationship between explicit relevance feedback such as user ratings and implicit relevance feedback such as reading time.
The second study was focused on user searching behaviour and it builds on the results of the preliminary study. A plugin in Firefox browser was used to capture and log several implicit relevance feedback indicators, explicit ratings of document familiarity, difficulty and relevance from 77 users. A number of implicit relevance feedback indicators were correlated with user explicit relevance feedback such as ratings. A predictive function model was developed based on the captured implicit and explicit relevance feedback. The effect of task type, document familiarity and document difficulty on user behaviour was also examined. The predictive function model was validated through an eye gaze study and standard evaluation metrics. The results of the eye gaze study indicate that the predictive model derived from implicit indicators can be used in place of an eye gaze.
Furthermore, a prototype system for domain-specific implicit feedback was developed and evaluated. The results show that supplementing user queries with implicit feedback considerably improves the relevancy of returned results from, a domain-specific search engine.|
|Description: ||RECOMMENDATION OF RELEVANT WEB DOCUMENTS|
|Appears in Collections:||Mathematics / Computer Science|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.