Paper Number
1752
Paper Type
Complete Research Paper
Abstract
Understanding user experience presents challenges for researchers since gathering user feedback can be costly and time-consuming. This study introduces a novel semi-automated framework for analyzing user experiences with a topic on Reddit. The framework identifies relevant subreddits based on a seed list and then explores users’ histories to find topically related communities. A heterogeneous graph is constructed to represent interactions between users and subreddits, with users’ activities captured as BERT-generated textual node vectors. The proposed framework is evaluated using a use case of Pay-per-Click technology. Furthermore, the versatility of the proposed data collection method is demonstrated in a news analytics application related to the Russia-Ukraine conflict using the New York Times articles and comments. Results show that the inclusion of neighborhood subreddits significantly broadens the scope of user topic analysis. Furthermore, the proposed graph structure-based framework outperformed four alternatives considered providing evidence that it can effectively predict missing user-subreddit and user-user interactions.
Recommended Citation
Gupta, Shelly; Hai, Ameen Abdel; Alharbi, Abdulrahman; Otudi, Hussain; and Obradovic, Zoran, "Exploring Topic-Related User Experiences Through Social Graph" (2024). ECIS 2024 Proceedings. 14.
https://aisel.aisnet.org/ecis2024/track09_coghbis/track09_coghbis/14
Exploring Topic-Related User Experiences Through Social Graph
Understanding user experience presents challenges for researchers since gathering user feedback can be costly and time-consuming. This study introduces a novel semi-automated framework for analyzing user experiences with a topic on Reddit. The framework identifies relevant subreddits based on a seed list and then explores users’ histories to find topically related communities. A heterogeneous graph is constructed to represent interactions between users and subreddits, with users’ activities captured as BERT-generated textual node vectors. The proposed framework is evaluated using a use case of Pay-per-Click technology. Furthermore, the versatility of the proposed data collection method is demonstrated in a news analytics application related to the Russia-Ukraine conflict using the New York Times articles and comments. Results show that the inclusion of neighborhood subreddits significantly broadens the scope of user topic analysis. Furthermore, the proposed graph structure-based framework outperformed four alternatives considered providing evidence that it can effectively predict missing user-subreddit and user-user interactions.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.