A task-based scientific paper recommender system for literature review and manuscript preparation

In the domain of scholarly communication lifecycle, recommender systems have been built to provide research papers for researchers’ explicit and implicit information needs. Previous studies (Jardine, 2014; Mcnee, 2006) have employed an algorithmic approach of providing solutions to researcher’s task...

Full description

Saved in:
Bibliographic Details
Main Author: Sesagiri Raamkumar, Aravind
Other Authors: Foo Shou Boon, Schubert
Format: Theses and Dissertations
Language:English
Published: 2018
Subjects:
Online Access:https://hdl.handle.net/10356/89406
http://hdl.handle.net/10220/46243
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
id sg-ntu-dr.10356-89406
record_format dspace
institution Nanyang Technological University
building NTU Library
country Singapore
collection DR-NTU
language English
topic DRNTU::Library and information science::Libraries::Digital libraries and information portals
DRNTU::Library and information science::Libraries::Information retrieval and analysis
spellingShingle DRNTU::Library and information science::Libraries::Digital libraries and information portals
DRNTU::Library and information science::Libraries::Information retrieval and analysis
Sesagiri Raamkumar, Aravind
A task-based scientific paper recommender system for literature review and manuscript preparation
description In the domain of scholarly communication lifecycle, recommender systems have been built to provide research papers for researchers’ explicit and implicit information needs. Previous studies (Jardine, 2014; Mcnee, 2006) have employed an algorithmic approach of providing solutions to researcher’s tasks. The characteristics of the tasks, their inter-relationships and intra-relationships with algorithms have been largely ignored since the focus has mainly been to propose different recommendation techniques on top of a variety of algorithms. Driven by these research gaps, the overarching goal of this research is to build an assistive system for helping researchers in finding papers for key Literature Review (LR) and Manuscript Preparatory (MP) tasks. To achieve this goal, two research objectives are proposed. The first objective is to identify an appropriate method to map the identified LR and MP tasks to relevant algorithms. The deliverable for this objective is a prototype assistive system that provides recommendations for three tasks. The second research objective is to evaluate whether the performance of the proposed recommendation techniques and the overall system are at the expected level. To address the research objectives, the research is divided into two interrelated studies. In Study I, a university-wide survey was conducted on the topic of Inadequate and Omitted Citations in manuscripts (IOC). The 207 survey respondents were classified into manuscript reviewer and author groups. Survey results indicated that manuscript authors frequently miss citing seminal and topically-similar papers in journal manuscripts. The lack of experience in a specific research area was perceived as a major reason for IOC, followed by lack of overall research experience and the scenario of working in interdisciplinary research projects. Authors frequently needed external assistance in finding interdisciplinary and topically-similar papers for LR. Based on the findings, two LR search tasks of building reading list and finding topically similar papers were shortlisted. A third task meant to help researchers in identifying unique and important papers from their final reading list was selected, thereby making it a total of three tasks for the assistive system. A prototype called Rec4LRW (Recommendations for Literature Review and Writing) system was developed for providing recommendations for the shortlisted three tasks. The system development was guided by a threefold intervention framework comprising of (i) task redesign for addressing the algorithmic improvements, (ii) task interconnectivity addressing the management of papers between the tasks and (iii) informational display features in the system’s user-interface for expediting researcher’s relevance judgment decisions. The second research objective is addressed in Study II. As a part of Study II, an offline evaluation experiment and a user evaluation study were conducted. An extract of papers from the ACM Digital Library was used as the corpus for the evaluations. A total of 119 researchers who had experience in authoring research papers, participated in the user study. Predictors and correlates for the output quality measures were identified for each task. This study established the effectiveness of the three interventions in providing relevant recommendations. Graduate students and novice researchers found the recommendations and the overall system to be more useful and effective.
author2 Foo Shou Boon, Schubert
author_facet Foo Shou Boon, Schubert
Sesagiri Raamkumar, Aravind
format Theses and Dissertations
author Sesagiri Raamkumar, Aravind
author_sort Sesagiri Raamkumar, Aravind
title A task-based scientific paper recommender system for literature review and manuscript preparation
title_short A task-based scientific paper recommender system for literature review and manuscript preparation
title_full A task-based scientific paper recommender system for literature review and manuscript preparation
title_fullStr A task-based scientific paper recommender system for literature review and manuscript preparation
title_full_unstemmed A task-based scientific paper recommender system for literature review and manuscript preparation
title_sort task-based scientific paper recommender system for literature review and manuscript preparation
publishDate 2018
url https://hdl.handle.net/10356/89406
http://hdl.handle.net/10220/46243
_version_ 1681056742838894592
spelling sg-ntu-dr.10356-894062020-06-22T05:46:34Z A task-based scientific paper recommender system for literature review and manuscript preparation Sesagiri Raamkumar, Aravind Foo Shou Boon, Schubert Wee Kim Wee School of Communication and Information Pang Lee San, Natalie DRNTU::Library and information science::Libraries::Digital libraries and information portals DRNTU::Library and information science::Libraries::Information retrieval and analysis In the domain of scholarly communication lifecycle, recommender systems have been built to provide research papers for researchers’ explicit and implicit information needs. Previous studies (Jardine, 2014; Mcnee, 2006) have employed an algorithmic approach of providing solutions to researcher’s tasks. The characteristics of the tasks, their inter-relationships and intra-relationships with algorithms have been largely ignored since the focus has mainly been to propose different recommendation techniques on top of a variety of algorithms. Driven by these research gaps, the overarching goal of this research is to build an assistive system for helping researchers in finding papers for key Literature Review (LR) and Manuscript Preparatory (MP) tasks. To achieve this goal, two research objectives are proposed. The first objective is to identify an appropriate method to map the identified LR and MP tasks to relevant algorithms. The deliverable for this objective is a prototype assistive system that provides recommendations for three tasks. The second research objective is to evaluate whether the performance of the proposed recommendation techniques and the overall system are at the expected level. To address the research objectives, the research is divided into two interrelated studies. In Study I, a university-wide survey was conducted on the topic of Inadequate and Omitted Citations in manuscripts (IOC). The 207 survey respondents were classified into manuscript reviewer and author groups. Survey results indicated that manuscript authors frequently miss citing seminal and topically-similar papers in journal manuscripts. The lack of experience in a specific research area was perceived as a major reason for IOC, followed by lack of overall research experience and the scenario of working in interdisciplinary research projects. Authors frequently needed external assistance in finding interdisciplinary and topically-similar papers for LR. Based on the findings, two LR search tasks of building reading list and finding topically similar papers were shortlisted. A third task meant to help researchers in identifying unique and important papers from their final reading list was selected, thereby making it a total of three tasks for the assistive system. A prototype called Rec4LRW (Recommendations for Literature Review and Writing) system was developed for providing recommendations for the shortlisted three tasks. The system development was guided by a threefold intervention framework comprising of (i) task redesign for addressing the algorithmic improvements, (ii) task interconnectivity addressing the management of papers between the tasks and (iii) informational display features in the system’s user-interface for expediting researcher’s relevance judgment decisions. The second research objective is addressed in Study II. As a part of Study II, an offline evaluation experiment and a user evaluation study were conducted. An extract of papers from the ACM Digital Library was used as the corpus for the evaluations. A total of 119 researchers who had experience in authoring research papers, participated in the user study. Predictors and correlates for the output quality measures were identified for each task. This study established the effectiveness of the three interventions in providing relevant recommendations. Graduate students and novice researchers found the recommendations and the overall system to be more useful and effective. Doctor of Philosophy 2018-10-08T06:41:39Z 2019-12-06T17:24:48Z 2018-10-08T06:41:39Z 2019-12-06T17:24:48Z 2018 Thesis Aravind, S. R. (2018). A task-based scientific paper recommender system for literature review and manuscript preparation. Doctoral thesis, Nanyang Technological University, Singapore. https://hdl.handle.net/10356/89406 http://hdl.handle.net/10220/46243 10.32657/10220/46243 en 344 p. application/pdf