Analyzing Question Quality through Intersubjectivity: World Views and Objective Assessments of Questions on Social Question-Answering

摘要

Social question-answering (SQA) allows people to askquestions in natural language and receive answers fromothers. While research on SQA has focused on the qualityof answers provided with implications for system-basedinterventions, few studies have examined whether thequestions asked to elicit these answers accurately depict anasker’s information need. To address this gap, the currentstudy explores the viability for system based interventionsto improve questions by comparing human, non-textualassessments of question quality to automatic, textualfeatures extracted from the questions’ content in order todetermine whether there is a significant relationshipbetween subjective judgments on one hand, and objectiveones on the other. Findings indicate that not only is there asignificant relationship between human-based ratings ofquestion quality criteria and extracted textual features, butalso that distinct textual features contribute to explainingthe variability of each human-based rating. These findingsencourage further study of the relationship between thereasons for why a question might be of poor quality andtextual features that can be extracted from the question.This relationship can ultimately inform design ofintervention-based systems that can not only automaticallyassess question quality, but also provide reasons that can beunderstood by the asker as to why the quality of his or herquestion is poor and suggest how to revise the question.