[Paper Review] Leveraging Social Foci for Information Seeking in Social Media

Suhas Ranganath, Jiliang Tang+3 more|arXiv (Cornell University)|Feb 23, 2015

Expert finding and Q&A systems41 references7 citations

TL;DR

This paper proposes a framework that leverages social foci—shared interests and affiliations derived from network and content data—to identify suitable answerers for subjective, personal questions on social media platforms like Twitter. By integrating social network structure and user-generated content, the method outperforms baselines across diverse question categories, achieving a MAP@5 of 4.68% for 'Suggestions' and maintaining strong performance (MAP > 3%) across varying ratios of network and content information.

ABSTRACT

The rise of social media provides a great opportunity for people to reach out to their social connections to satisfy their information needs. However, generic social media platforms are not explicitly designed to assist information seeking of users. In this paper, we propose a novel framework to identify the social connections of a user able to satisfy his information needs. The information need of a social media user is subjective and personal, and we investigate the utility of his social context to identify people able to satisfy it. We present questions users post on Twitter as instances of information seeking activities in social media. We infer soft community memberships of the asker and his social connections by integrating network and content information. Drawing concepts from the social foci theory, we identify answerers who share communities with the asker w.r.t. the question. Our experiments demonstrate that the framework is effective in identifying answerers to social media questions.

Motivation & Objective

• To address the challenge of identifying relevant answerers for subjective, personal questions in social media, where traditional Q&A platforms are not designed for such interactions.
• To investigate how social context—specifically shared interests and affiliations—can be leveraged to improve answerer identification for social media questions.
• To develop a framework that integrates network and content information to model soft community memberships and infer shared foci between askers and potential answerers.
• To evaluate the effectiveness of the framework across diverse question categories and varying proportions of network and content data.
• To demonstrate that social foci-based methods outperform baseline approaches in identifying answerers for subjective questions on Twitter.

Proposed method

• Infers soft community memberships for users by combining network structure (social graph) and textual content from user posts.
• Applies social foci theory to model shared interests (e.g., sports, music, academics) between the asker and their social connections.
• Uses a weighted fusion model with parameters α (content weight) and β (network weight) to combine content and network signals.
• Employs a retrieval framework that ranks potential answerers based on shared foci with the asker, using metrics like MRR and MAP@5.
• Leverages user-generated questions on Twitter as instances of information-seeking behavior, treating them as queries to be answered by socially connected users.
• Performs human annotation to categorize questions into types such as 'Suggestions', 'Opinion', 'Favor', and 'Rhetorical' for evaluation.

Experimental results

Research questions

• RQ1 How can network and content information be effectively combined to identify answerers for subjective questions in social media?
• RQ2 To what extent do shared social foci (e.g., interests, affiliations) improve the identification of relevant answerers compared to baseline methods?
• RQ3 How does the framework perform across different categories of social media questions, such as suggestions, opinions, and rhetorical questions?
• RQ4 How robust is the framework to variations in the relative proportions of network and content information?
• RQ5 Can social foci-based methods effectively identify answerers for personal, non-factual questions that are prevalent on platforms like Twitter?

Key findings

• The framework achieves a MAP@5 of 4.68% for 'Suggestions' questions, representing a 1.78% improvement over the nearest baseline.
• For 'Favor' questions, the framework achieves a MAP@5 of 4.39%, a 1.01% improvement over the baseline.
• The framework maintains a MAP value above 3% across all tested combinations of α and β, indicating robustness to varying ratios of network and content information.
• The lowest performance is observed for 'Rhetorical' questions (MAP@5 = 0%), suggesting that such questions may not reflect genuine information needs and are less amenable to foci-based matching.
• A paired t-test confirms that the performance improvements over the baseline are statistically significant across all question categories.
• The framework significantly outperforms random ordering, with the lowest performance being more than twice that of random selection, demonstrating its effectiveness even under suboptimal parameter settings.