(A) data-driven approach for understanding user satisfaction for question answering in online health communities : a case study on a diabetes community

다국어 초록 (Multilingual Abstract)

This study focuses on understanding the posting and replying patterns of users in online health communities (OHCs). Using large-scale crawled data collected from a globally popular diabetes OHC, we formulate a prediction task where the objective is to predict whether an OHC user who posts a question online becomes satisfied with his/her answer. We apply linguistic features extracted from emotion lexicons with machine learning and deep learning approaches such as sentiment analysis, non-negative matrix factorization, and Word2Vec to obtain features, upon which we build a binary classifier for predicting user satisfaction. We present our experiment results and analyze the importance of different features, both in categorical and individual levels.

목차 (Table of Contents)

• Abstract
• Section 1 Introduction 1
• Section 2 Related Work 3
• 2.1 Social support types in OHCs 3
• 2.2 Data-driven approaches for identifying user types in OHCs 5
• Section 3: Datasets 6
• 3.1 Dataset introduction 7
• 3.2 Data collection 9
• 3.3 Preprocessing steps 11
• 3.4 Feature engineering 13
• Section 4 Experiments 18
• Section 5 Conclusion 24
• References 25