Effects of Multiple Representations in Questions on Q&A Sites
Saori Suzuki (AY 2010)
The use of Q&A sites has increased recently. There are many studies on Q&A sites, but few studies on how to formulate effective questions. On the other hand, there is the concept of multiple representations in information retrieval. Multiple representation is a way of representing information about a single object from different perspectives. Although some studies have shown that such multiple representations are effective in query generation, there is no research on the use of multiple representations for questions on Q&A sites. However, there are no studies using multiple expressions for questions on Q&A sites where background information about the questioner can be added to the question. However, we believe that the use of multiple expressions may be effective on Q&A sites where background information about the questioner can be added to the question. Therefore, the purpose of this study was to show an effective way to express questions using multiple expressions, and a human subject experiment was conducted to investigate this.
In the experiment, each subject was asked to think of six questions to post on a Q&A website. The questions used were information-type questions, which asked for facts, and discussion-type questions, which asked for opinions. To examine the effect of multiple expressions, the background information was classified into five categories: "task," "idea," "situation," "attribute," and "constraint," and subjects were asked to add them to the basic questions to determine which background elements were effective when combined as multiple expressions. The questions created by the subjects were posted on a Q&A website, and the collected responses were shown to the subjects who created the questions, and they were asked to rate each response according to the rating index.
A total of 930 responses were obtained for 276 questions created by 46 subjects. Analysis of these responses showed that multiple expressions with the addition of "attributes" among the five background elements increased novelty ratings for both question types. In addition, the effect of background factors differed between the information and discussion types. The information type was affected differently by the background elements than the discussion type. In the information type, adding "idea" to the attribute increased the rating of helpfulness, while in the discussion type, adding "task" decreased the ratings of satisfaction and thematicity, and adding "situation" decreased the ratings of thematicity and length. Thus, the experiment confirmed the effect of multiple expressions in questions on Q&A sites, and also showed that the effect varied by question type.
Based on the above analysis, this study was able to indicate elements that should be added or avoided in questions on Q&A sites. As a future direction, interaction design that provides input support for easier and more effective description of background elements could be developed to facilitate question creation.
(Translated by DeepL)