Answering Communities Research Articles

Profiling Users for Question Answering Communities via Flow-Based Constrained Co-Embedding Model

In this article, we study the task of user profiling in question answering communities (QACs). Previous user profiling algorithms suffer from a number of defects: they regard users and words as atomic units, leading to the mismatch between them; they are designed for other applications but not for QACs; and some semantic profiling algorithms do not co-embed users and words, leading to making the affinity measurement between them difficult. To improve the profiling performance, we propose a neural Flow-based Constrained Co-embedding Model, abbreviated as FCCM. FCCM jointly co-embeds the vector representations of both users and words in QACs such that the affinities between them can be semantically measured. Specifically, FCCM extends the standard variational auto-encoder model to enforce the inferred embeddings of users and words subject to the voting constraint, i.e., given a question and the users who answer this question in the community, representations of the users whose answers receive more votes are closer to the representations of the words associated with these answers, compared with representations of whose receiving fewer votes. In addition, FCCM integrates normalizing flow into the variational auto-encoder framework to avoid the assumption that the distributions of the embeddings are Gaussian, making the inferred embeddings fit the real distributions of the data better. Experimental results on a Chinese Zhihu question answering dataset demonstrate the effectiveness of our proposed FCCM model for the task of user profiling in QACs.

Evolution of Experts in Question Answering Communities

Community Question Answering (CQA) services thrive as a result of a small number of highly active users, typically called experts, who provide a large number of high quality useful answers. Understanding the temporal dynamics and interactions between experts can present key insights into how community members evolve over time. In this paper, we present a temporal study of experts in CQA and analyze the changes in their behavioral patterns over time. Further, using unsupervised machine learning methods, we show the interesting evolution patterns that can help us distinguish experts from one another. Using supervised classification methods, we show that the models based on evolutionary data of users can be more effective at expert identification than the models that ignore evolution. We run our experiments on two large online CQA to show the generality of our proposed approach.

How do Virtual Badges Incentivize Voluntary Contributions to Online Communities?

Online communities rely heavily on badges to incentivize users to participate in voluntary activities. We examine the relationship between earned badges/pursued badges and contributions in the Stack Overflow question and answer community. Our main results reveal that different levels of earned and pursued badges increase the amount of subsequent answering activity. In addition, earned gold badges are the most influential in the motivation of users while earned bronze badges are the least impactful. Furthermore, the effect of badges involves positive spillovers from higher levels of earned badges to lower ones. Our findings offer important theoretical and managerial implications for online communities.

Anonymity, User Engagement, Quality, and Trolling on Q&A Sites

In online question and answer (Q&A) communities, people ask questions and share answers at all levels of topic sensitivity. Identity options within these communities range from anonymity to real name. The amount of engagement, and the quality of engagement on Q&A sites may differ depending on the identity options available. In this paper, we investigate the relationship between the amount of engagement, the quality of engagement, and different types of identity by analyzing three Q&A sites with different identity policies. We find that highly sensitive questions are more likely to be asked anonymously. Furthermore, allowing anonymity does not affect answer quality and only has a weak, negative indirect effect on engagement. On the other hand, anonymity leads to more trolling. We suggest online communities provide a way for users to ask highly sensitive questions anonymously and pair this with moderation mechanisms to reduce trolling

Measuring topic network centrality for identifying technology and technological development in online communities

Online communities are a rapidly growing knowledge repository that provides scholarly research, technical discussion, and social interactivity. This abundance of online information increases the difficulty of keeping up with new developments difficult for researchers and practitioners. Thus, we introduced a novel method that analyses both knowledge and social sentiment within the online community to discover the topical coverage of emerging technology and trace technological trends. The method utilizes the Weibull distribution and Shannon entropy to measure and link social sentiment with technological topics. Based on question-and-answer and social sentiment data from Zhihu, which is an online question and answer (Q&A) community with high-profile entrepreneurs and public intellectuals, we built an undirected weighting network and measured the centrality of nodes for technology identification. An empirical study on artificial intelligence technology trends supported by expert knowledge-based evaluation and cognition provides sufficient evidence of the method's ability to identify technology. We found that the social sentiment of hot technological topics presents a long-tailed distribution statistical pattern. High similarity between the topic popularity and emerging technology development trends appears in the online community. Finally, we discuss the findings in various professional fields that are widely applied to discover and track hot technological topics.

Hot question prediction in Stack Overflow

Stack Overflow is a very popular programming question and answer community. Some questions become hot, and receive high views, which are of widespread concern to developers. Finding hot questions early can give priority to recommend potential hot questions to answers, thereby shortening the response time. Besides, the hot question prediction is also helpful for making advertising plan, planning advertising campaigns and estimating costs. Therefore, it is important to predict hot questions. The authors propose the VSAF method which analyses the View amount changes, Answer amount changes and Score changes soon after questions' creation based on Fully convolutional neural network. The performance of the VSAF method based on a training set and two different test sets has been evaluated. The training set has 1600 hot questions and 1600 cold questions. The random test set has 381 hot questions and 2819 cold questions, while the balanced test set has 400 hot questions and 400 cold questions. The experimental results show that using the balanced test set, VSAF achieves Accuracy, F1hot and F1cold of 80%, 77.77% and 81.81%, which outperforms the baseline approach by 25.59%, 21.52% and 29.04%, respectively. Using the random test set for evaluation, VSAF achieves Accuracy, F1hot and F1cold of 84.91%, 53.96% and 90.97%, which outperforms the baseline approach by 31.83%, 84.16% and 19.35%, respectively. The VSAF method significantly outperforms the state-of-the-art approach on hot question prediction.

Recommended System for Enhancing Tag Popularity in a Question Answering Community through Splay-net Techniques

Collaborative filtering filters information by using the recommendations of peer participants. The long tail problem states users with higher points obtain a high reputation compared to less scored users. In popular community question answering websites, like stack exchange network sites, users with unanswered or ignored questions for a long time get a tumbleweed badge without considering their past history. This deteriorates their further contribution to the website. Mostly new or low-reputation people ask the tumbleweed questions. The popularity of the tags follows a long tail theory. The focus of this research work is to design a recommendation system that prevents participants from tumbleweed badge with tag suggestion method to add new or non-popular tags to the existing popular tag list. The splay-net has a self-balancing graph which brings the recently accessed item to the top of the tree. In this paper, we use the splay-net technique to represent users’ reputation along with their tags.

Discovering America: Unpacking Popular Social Q&A for Prospective Chinese International Students

The global pandemic and recent shift to online learning has heightened the need to better understand how to support international students, especially across online and virtual platforms. However, a review of literature reveals a paucity of studies dedicated to international student use of internet-based platforms and services for seeking information and sharing experiences. One avenue for further research is Social Question & Answer Communities (SQACs), which are vast conduits of shared experiences and knowledge connecting hundreds of thousands of pre-application and pre-arrival students. In this study, content analysis was utilized to qualitatively observe the contents of the “Overseas Studies in The United States” section of Zhihu, and quantitatively count their features and characteristics. The study found that 58% of the questions and answers were devoted to Academic issues such as testing, admissions, learning and research, with another 13.0% on Crime, Law and Safety, and the remaining 29% of the questions were associated to a diverse array of topics associated with living and working in American society. The most popular answers were made up of mainly 4 main types: Sharing One’s Experience (32.0%), Advice (26.0%), Opinion (22.0%) and Critique (15.0%). Content analysis of three main answer features, namely the use of Imagery, Digital Resources, and Social feature, indicated that the Advice and Critique answer types contain the richest variety of features and that question context, textual styles and use of digital resources are important factors for understanding the answer popularity in SQACs.

Automatic Voter Recommendation Method for Closing Questions in Stack Overflow

Stack Overflow is the most popular programming question and answer community that continuously receives a large number of questions every day. To ensure the quality of questions, the community grants privileges for the moderators and a group of experienced users to review the quality of questions and close the low-quality ones (e.g. duplicate or irrelevant questions). The review process is a typical crowdsourcing job that relies on users’ volunteer participation, and the current practices of closing questions in Stack Overflow face two aspects of challenges: (1) an obvious increase in both the absolute number and the percentage of “closed” questions; (2) a considerable decrease in participation willingness of experienced users to close questions. In order to solve the problem, we present a novel model of user willingness for reviewing and voting questions by incorporating four types of user activity history, including questions, answers, comments and votes of closing questions. Then we propose an automatic recommendation method based on the model to assign experienced users proper questions, to utilize the forces of them to close questions. The evaluation shows that the successful recommendation probability in the top 5, top 10, top 20, top 30, top 40, top 50 users are 48.23%, 58.93%, 68.83%, 74.27%, 78.13% and 81%, respectively.

Bridging Semantic Gaps between Natural Languages and APIs with Word Embedding

Developers increasingly rely on text matching tools to analyze the relation between natural language words and APIs. However, semantic gaps, namely textual mismatches between words and APIs, negatively affect these tools. Previous studies have transformed words or APIs into low-dimensional vectors for matching; however, inaccurate results were obtained due to the failure of modeling words and APIs simultaneously. To resolve this problem, two main challenges are to be addressed: the acquisition of massive words and APIs for mining and the alignment of words and APIs for modeling. Therefore, this study proposes Word2API to effectively estimate relatedness of words and APIs. Word2API collects millions of commonly used words and APIs from code repositories to address the acquisition challenge. Then, a shuffling strategy is used to transform related words and APIs into tuples to address the alignment challenge. Using these tuples, Word2API models words and APIs simultaneously. Word2API outperforms baselines by 10-49.6 percent of relatedness estimation in terms of precision and NDCG. Word2API is also effective on solving typical software tasks, e.g., query expansion and API documents linking. A simple system with Word2API-expanded queries recommends up to 21.4 percent more related APIs for developers. Meanwhile, Word2API improves comparison algorithms by 7.9-17.4 percent in linking questions in Question&Answer communities to API documents.

HOW BEHAVIOR OF USERS IMPACTS THE SUCCESS OF ONLINE Q&A COMMUNITIES

What makes Question & Answer (Q&A) communities productive? In this paper, we look into how the diversity of behavioral types of agents impacts the performance of Q&A communities using different performance metrics. We do this by developing an agent-based model informed by insights from previous studies on Q&A communities. By analyzing the different strategies for how questions are selected to answer, we find that there are mixtures of strategies leading to the best outcomes for different performance conditions. Particularly, Q&A communities that encourage participants to focus on answering the new questions reach the best performance in answering the questions, creating the long term value, and improving the competence of solving difficulty. In conclusion, we find that the current strategies of question selection on Stack Overflow are in line with the high performance of producing public benefit from the collective attention available.

The influence of the community climate on users’ knowledge-sharing intention: the social cognitive theory perspective

ABSTRACT The purpose of this paper is to investigate the underlying mechanism of the community climate on knowledge-sharing intention in virtual communities. Based on social cognitive theory, a theoretical model was developed and empirically tested using a web-based survey of 525 members from an online question and answer (Q&A) community in China. Results showed that knowledge-sharing self-efficacy and outcome expectation played a chain mediation role in the relationship between the community climate and knowledge-sharing intention. This study is among the first to examine whether and how community climate influences knowledge sharing. Our research findings unpack the mechanism of the community climate on knowledge-sharing intention, and provide practical insights on how to use community climate to facilitate users to share knowledge in online Q&A communities.

For a few diversities more: A voting‐attack‐immune voting scheme for social question and answer websites

SummarySocial question and answer (Q&A) communities or websites, such as Quora, StackOverflow, become popular from the last decade. They provide collected answers in a crowdsourcing way by encouraging the interactive between users. The ranking of answers in such Q&A websites depends on many factors, such as received upvotes, downvotes, publishing time, and so on. When there exist malicious users, they can manipulate the ranking of answers by launching voting spamming attacks. In this article, we propose a metric, diversity of visibility, to characterize the balance between rankings of answers with conflicting opinions and propose a time‐diversitybased voting scheme to reach a high diversity of visibility, so as to minimize the impact of voting spamming attacks. We evaluate the proposed metric and voting scheme in the case study of a Chinese Q&A website, Zhihu, and show that the proposed voting scheme consistently shows a good balance of answers with conflicting opinions in both actual scenarios and synthetic scenarios.

Understanding the formation mechanism of high-quality knowledge in social question and answer communities: A knowledge co-creation perspective

The social question and answer (Q&A) community provides people with an effective tool to obtain high-quality information. From the perspective of reciprocal determinism and value co-creation, this study aims to investigate the formation mechanism of high-quality knowledge in the community. We develop a model to investigate how cognitive factors and community technological factors influence users’ knowledge co-creation behavior, thereby influencing knowledge quality in the community. A survey of 382 knowledge contributors in a social Q&A community shows that knowledge self-efficacy, topic richness, personalized recommendation, and social interactivity have a positive impact on users' knowledge sharing and integration behavior, which subsequently affect the community’s knowledge quality. Moreover, users' ratings moderate the influence of knowledge sharing on knowledge quality. This research demonstrates the synergistic effect of people and technology in knowledge co-creation, thus advances literature about value co-creation and content quality in online communities.

Why do participants continue to contribute? Evaluation of usefulness voting and commenting motivational affordances within an online knowledge community

Participants are increasingly using online knowledge communities to access and share information and to collaboratively solve problems. However, in online communities that address technical questions (i.e., utilitarian, rather than hedonic or supportive, in nature), a key problem is encouraging sufficient, ongoing knowledge contribution. While many may use the community to post and obtain information periodically, fewer take the time to consistently contribute knowledge to the community. Interestingly, research has yet to comprehensively consider the impact of motivational affordances, in the forms of voting and commenting features, to address this challenge of under-contribution in online knowledge communities. To fill this research gap, this study proposes a research framework that integrates consideration of usefulness voting and the collaboration tool of commenting to explain variation in individuals' online knowledge contribution behaviors. The research model is estimated with a fixed-effects Poisson regression applied to a longitudinal panel dataset collected from a technical online question and answer community. We find that positive votes motivate participants' knowledge contributions, while received negative votes significantly decrease their knowledge contributions. Thus, while maintaining content quality is important, negative motivational affordances (i.e., down voting) can reduce sustained contributions. We also find commenting to be an important motivator for online knowledge contribution and, interestingly, find that comments moderate the relationships between voting and knowledge contribution. Overall, this research not only extends our current understandings of knowledge contribution in online communities, but also sheds light on improving online community knowledge exchange.

A supernetwork-based online post informative quality evaluation model

In an era of big data, explosive growth of online posts makes the judgment of their qualities harder while more important. In many cases, people want to quickly identify those most informative posts, which contain details, insights or in-depth criticisms, to help them make better decisions. To meet this demand, a three-stage model is proposed. First, a super-network model is introduced to accommodate the multidimensional attributes of online posts, including keywords, user ID, emotions and the related event. Second, a corpus updating mechanism is introduced to generate event specific corpora, which help to discriminate the informative quality of posts in the next stage. Third, machine learning algorithms are applied, where the posts are first filtered by a linear discriminant classifier and then assessed by a multilayer perceptron neural network. To test the model, we chose six online public opinion events that fell into two categories: major public safety crisis and online controversies about public policies. Experimental results showed the effectiveness of the proposed model, where majority of errors are less than 0.05, on a 0–1 measuring scale. In the future, this model may also be adapted to areas including evaluation of informative quality of websites, product reviews and answers in question and answer communities.

Evaluating Answer Qualities on Q&A Community Sites (StackOverFlow)

Social Networking sites have gained much popularity in the recent years due to the widespread use of internet and availability of huge amounts of information. One kind of social network is Question Answering Community sites like StackOverflow, Quora and many more. These sites facilitate the user to ask their query on such sites using natural language and get the answers to their queries in natural language. There are many challenges to face when achieving above goals. One of the goals is to find the valid answers out of available answers and present it to the user, so as to save the user from going through multiple answers for a particular question and selecting the valid or appropriate answer. In this paper, we propose a system that finds the best answer from all the available answers based on answer content and the relationship between question and answer. We trained the system using three different classifiers and noted the accuracy results of each; the classifier which gave maximum accuracy was selected.

Goal achievement, subsequent user effort and the moderating role of goal difficulty

It is established in the marketing literature that people increase their effort in a recurring goal framework when the goal difficulty increases, because their experience of success upon goal achievement leads to higher self-efficacy. It is not yet established, however, whether this mechanism also applies to a gamified environment without monetary incentives or quasi-monetary benefits. Using data from a popular German Question & Answer community in which goals are represented in the form of badges we empirically investigate the impact of successful goal achievement on the base level of effort to attain the next goal in a hierarchical recurring goal framework. Our findings indicate that after successful badge achievement users increase their subsequent base level of effort to attain the next badge, but only as long as the badge just achieved represented a challenge to the user. We identify self-efficacy as the key driver of this behavior and provide additional experimental evidence in support of this explanation.

Expert Finding in Community Question-Answering for Post Recommendation

Community question answering system is a perfect example of platform where people participate to seek expertise on their topic of interest. But information overload, finding the expertise level of users and trustworthy answers remain key challenges within these communities. Moreover, people do not look for personal advices but expert views on such platforms therefore; expert finding is an integral part of these communities. In order to trust someone's opinion who is not known in person by the users of the community, it is necessary to find the credibility of such person. By determining expertise levels of users, authenticity of their posts can easily be determined. Also, by identifying experts, each expert will be shown relevant posts to indulge in so that he can use his knowledge and skills to give valid and correct answers. For users too, it will be easy to find reliable answers, once they get to know the expertise level of the answerers. Motivated by these facts, we put forward a framework for finding experts in online question answer community (stackoverflow) referred to as Expert Recommender System which uses a well-recognized global-trust metric, PageRankTM for finding experts in the community building a Trust-based system and then uses collaborative filtering to find similar experts based on their level of expertise and their topics of interests to a particular user. Once we have the top- k similar experts to a given expert, that expert is recommended with posts to collaborate upon, based on activities done by his top-k neighbor experts. The framework is evaluated for its performance and it clearly indicates the effectiveness of the system.

Experts and likely to be closed discussions in question and answer communities: An analytical overview

How do important members of online Question & Answer communities (who we call experts) behave? And how do they influence the discussions in which they take part? This work reports on an investigation into these questions, which we answer through analyses exploring metrics, machine learning classifiers, and recommendations. We report on several findings: the degree of expertise correlates to behavioral patterns, whereby experts would rarely ask for help, and instead, predominantly provide help to other community members; the inclusion of an expert results in longer discussions. We propose a metric (the weighted sum), which enables us to better quantify the reputations of expert members of the community. We describe the use of four machine learning classifiers for the identification of both expert users and the most significant conversations within these communities. We propose a novel approach for a recommendation system, which utilizes semantic annotations to identify topical experts and to ascertain their respective area of specialism. We foresee the suitability of our expertise-finding methods and findings to support Learning Analytics, and in scenarios where users may apply lessons learnt from our results to improve their status in a community. Our findings can also inform systems for recommending experts and discussions.