Start Date
12-13-2015
Description
Popular information systems such as community-based question answering sites (CQAs) and electronic networks of practice (ENPs) rely on accurate assessment of user contributed contents to ensure effective knowledge creation and exchange. We study the issue of content quality assessment through a context-free linguistic analysis, and hypothesize that, based on psycholinguistic theories and the politeness theory, the use of pronouns and other function words is correlated with the content’s perceived quality, moderated by the specific quality measure adopted. We empirically test our hypotheses through a random coefficient logit model and a fixed-effect negative binomial model with data obtained from StackExchange, a popular CQA platform, and the results are largely consistent with our hypotheses. Besides demonstrating how linguistic analyses can be used in content quality assessment, our analysis also uncovers potential limitations posed by an overreliance on subjective quality measures. This study contributes to the literature in knowledge management and strategic communication.
Recommended Citation
Lee, Shun-Yang; Rui, Huaxia; and Whinston, Andrew, "Content Quality Assessment through Context-Free Linguistic Features: Application to Community-Based Question Answering Platforms" (2015). ICIS 2015 Proceedings. 6.
https://aisel.aisnet.org/icis2015/proceedings/ConferenceTheme/6
Content Quality Assessment through Context-Free Linguistic Features: Application to Community-Based Question Answering Platforms
Popular information systems such as community-based question answering sites (CQAs) and electronic networks of practice (ENPs) rely on accurate assessment of user contributed contents to ensure effective knowledge creation and exchange. We study the issue of content quality assessment through a context-free linguistic analysis, and hypothesize that, based on psycholinguistic theories and the politeness theory, the use of pronouns and other function words is correlated with the content’s perceived quality, moderated by the specific quality measure adopted. We empirically test our hypotheses through a random coefficient logit model and a fixed-effect negative binomial model with data obtained from StackExchange, a popular CQA platform, and the results are largely consistent with our hypotheses. Besides demonstrating how linguistic analyses can be used in content quality assessment, our analysis also uncovers potential limitations posed by an overreliance on subjective quality measures. This study contributes to the literature in knowledge management and strategic communication.