问答系统研究综述

问答系统是人工智能和自然语言处理领域中具有广泛发展前景的研究方向之一.早期的问答系统限定以自然语言形式进行提问和回答,近年来,随着多模态知识图谱、多模态预训练模型的发展,支持文字、图片、音频、视频等多种模态间信息查询的广义问答系统逐渐成为新的研究热点,其以多媒体方式展示结果,更加直观、全面.本文根据问答系统任务对象的变化,将问答系统划分为3种类型:专用问答系统、通用问答系统和多模态问答系统.分析了这3种类型的问答系统发展过程中所面临的问题,着重总结每个阶段所采用的关键技术与方法,同时对问答系统在工业上的应用进行了举例说明,并对未来研究方向进行了展望.     

StackExchange问答社区知识传播

StackExchange是目前最流行的问答社区集结地之一.本文利用StackExchange中具有美国地理信息的用户构建StackExchange问答社区在美国境内的知识传播图谱,对传播网络的统计特征进行了分析,提取出问答社区类网站的传播模式,获取得到网络用户的知识分享方式.我们发现StackExchange中的问答社区在分享知识过程中,传播源往往不止一个.同时,我们为问答社区构建了知识传播图谱,发现这些传播图谱具有相似的统计特征,这意味着不同的问答社区可能具有类似的知识传播模式.     

What are developers talking about? An analysis of topics and trends in Stack Overflow

Programming question and answer (Q&A) websites, such as Stack Overflow, leverage the knowledge and expertise of users to provide answers to technical questions. Over time, these websites turn into repositories of software engineering knowledge. Such knowledge repositories can be invaluable for gaining insight into the use of specific technologies and the trends of developer discussions. Previous work has focused on analyzing the user activities or the social interactions in Q&A websites. However, analyzing the actual textual content of these websites can help the software engineering community to better understand the thoughts and needs of developers. In the article, we present a methodology to analyze the textual content of Stack Overflow discussions. We use latent Dirichlet allocation (LDA), a statistical topic modeling technique, to automatically discover the main topics present in developer discussions. We analyze these discovered topics, as well as their relationships and trends over time, to gain insights into the development community. Our analysis allows us to make a number of interesting observations, including: the topics of interest to developers range widely from jobs to version control systems to C# syntax; questions in some topics lead to discussions in other topics; and the topics gaining the most popularity over time are web development (especially jQuery), mobile applications (especially Android), Git, and MySQL.     

Domain problem-solving expert identification in community question answering

Question-Answering (Q&A) services provide internet users with platforms to exchange knowledge and ideas. The development of Q&A sites, or Community Question Answering (CQA), mainly depends on the high-quality content continuously contributed by users with high-level expertise, who can be recognized as experts. Expert finding is an important task for the authorities of Q&A communities to encourage commitment. In a highly competitive market environment, CQA managers have to take measures to retain and nurture users, especially superior contributors. However, current expertise scoring techniques adopted in CQA often give much credit to very active users and fail to identify real experts. This study aims to develop a robust and practical expert identification framework for Q&A communities, by combining well-designed expertise scoring technique and probabilistic clustering model. With regard to expert identification, a numerical metric of users' expertise is developed as the optimal expert finding strategy, and a clustering algorithm based on Gaussian-Gamma mixture model (GGMM) is proposed to efficiently distinguish experts from nonexperts. In the experiments, the proposed method is applied to real-world datasets collected from subcommunities of Stack Exchange Q&A networks. Results obtained from comparative experiments show that our method achieves better performance than the state-of-the-art methods and demonstrate the effectiveness of the proposed framework. The analysis shows that the framework which combines the proposed expertise scoring technique and Gaussian–Gamma mixture clustering model is capable of detecting excellent domain problem-solving experts who exhibit both domain interest and expertise.     

The structure and dynamics of knowledge network in domain-specific Q&A sites: a case study of stack overflow

Programming-specific Q&A sites (e.g., Stack Overflow) are being used extensively by software developers for knowledge sharing and acquisition. Due to the cross-reference of questions and answers (note that users also reference URLs external to the Q&A site. In this paper, URL sharing refers to internal URLs within the Q&A site, unless otherwise stated), knowledge is diffused in the Q&A site, forming a large knowledge network. In Stack Overflow, why do developers share URLs? How is the community feedback to the knowledge being shared? What are the unique topological and semantic properties of the resulting knowledge network in Stack Overflow? Has this knowledge network become stable? If so, how does it reach to stability? Answering these questions can help the software engineering community better understand the knowledge diffusion process in programming-specific Q&A sites like Stack Overflow, thereby enabling more effective knowledge sharing, knowledge use, and knowledge representation and search in the community. Previous work has focused on analyzing user activities in Q&A sites or mining the textual content of these sites. In this article, we present a methodology to analyze URL sharing activities in Stack Overflow. We use open coding method to analyze why users share URLs in Stack Overflow, and develop a set of quantitative analysis methods to study the structural and dynamic properties of the emergent knowledge network in Stack Overflow. We also identify system designs, community norms, and social behavior theories that help explain our empirical findings. Through this study, we obtain an in-depth understanding of the knowledge diffusion process in Stack Overflow and expose the implications of URL sharing behavior for Q&A site design, developers who use crowdsourced knowledge in Stack Overflow, and future research on knowledge representation and search.     

基于知识图谱、TF-IDF和BERT模型的冬奥知识问答系统

传统信息检索技术已经不能满足人们对信息获取效率的要求,智能问答系统应运而生,并成为自然语言处理领域一个非常重要的研究热点。本文针对中文的冬奥问答领域,提出了基于知识图谱、词频-逆文本频率指数 (term frequency-inverse document frequency,TF-IDF)和自注意力机制的双向编码表示(bidirectional encoder representation from transformers,BERT)的3种冬奥问答系统模型。本文首次构建了冬奥问答数据集,并将上述3种方法集成在一起,应用于冬奥问答领域,用户可以使用本系统来快速准确地获取冬奥内容相关的问答知识。进一步,对3种模型的效果进行了测评,测量了3种模型各自的回答可接受率。实验结果显示BERT模型的整体效果略优于知识图谱和TDIDF模型,BERT模型对3类问题的回答可接受率都超过了96%,知识图谱和TDIDF模型对于复合统计问答对的回答效果不如BERT模型。     

Piazza问答平台与Open edX平台的集成

Piazza问答平台与Open edX学习平台两者相互独立,影响用户使用,并且Piazza问答数据无法被高效利用。针对上述问题 ,对Piazza问答数据进行持久保存, 利用多标签过滤方法提高了Piazza问答数据的查找能力;基于Piazza-Xblock插件,实现了在Open edX平台查找和展现Piazza问答数据,以及通过URL参数直接访问Piazza特定页面的功能,达到Piazza问答平台与Open edX平台有机结合的效果。     

面向专家示例的Stack Overflow本体构造和推理研究

Stack Overflow是一个计算机领域的IT技术问答网站,为了获取问答网站中的专家示例并将其应用于API挖掘中,首先采用Scrapy爬虫框架技术获取Stack Overflow问答网站中的结构化数据,并存储在关系模式中;再使用本体建模工具Protégé构建本体,然后使用D2RQ工具实现对关系数据库的知识抽取,将关系模式转换为三元组形式的本体模型;同时,提出了一个面向专家示例的子本体抽取算法,用于从原本体中抽取出专家示例推理相关的子本体,并提出了若干条专家示例推理规则,能推导出专家所编写的代码示例。实验结果证明,从Stack Overflow本体模型中抽取的专家示例能提高API调用序列挖掘的准确率。     

An SVD-Entropy and bilinearity based product ranking algorithm using heterogeneous data

E-commerce websites, besides selling products and services, pay ample emphasis on providing a platform for consumers to share their opinions about past and potential purchases. They share such opinions as product reviews (star ratings, plain text, etc.) and answering product related questions (Q&A data). There are several machine learning and classification approaches available to scrutinize this review data, e.g., algorithms based on Entropy measures, Bilinear Similarity, stochastic methods, etc. In this paper, we review some of the prevalent review classification techniques and present a hybrid approach, involving Singular Value Decomposition (SVD), Entropy and Bilinear Similarity measures, that uses heterogeneous product data and simultaneously analyze and rank products for customers. With experimental results, we show that our approach effectively ranks products using (1) text reviews (2) Q&A data (3) five-star rating of products and has 10% improved prediction accuracy as compared to the individual approaches. Also, using SVD, we achieve a 35% runtime efficiency for our algorithm while only sacrificing 1% of the prediction accuracy.     

The Impact of Linguistic Complexity on Leadership in Online Q&A communities: Comparing Knowledge Shaping and Knowledge Adding

Understanding how to enhance online leadership in online Q&A communities is important because an online leader plays a role model or knowledge coordinator who can strengthen member commitment in the community. Considering the essential role of communication in establishing leadership, this study aims to understand how the linguistic complexity of two types of knowledge contribution, i.e., knowledge adding (KA) versus knowledge shaping (KS) that are targeted at two types of audience, may influence leadership in online Q&A communities. By analyzing the posting history of members from StackExchange, a massive network of online Q&A communities, our findings suggest that among the three linguistic complexity dimensions, readability and lexical diversity of KA have more positive impacts on online leadership than those of KS. However, the sentiment of KS has a more positive impact than the sentiment of KA. This study contributes to the online leadership research by highlighting the importance of adjusting linguistic styles based on types of communication behaviors (i.e., KA and KS) to earn leadership.     

Gratifications and social network service usage: The mediating role of online experience

This paper incorporates dual theories from communication research (uses and gratifications) and psychology research (online flow) to examine consumer behavior in the use of social network services. In particular, the study proposes that consumers’ online experience of interaction and arousal serves as the mediator of the relationship between social motivations and use behaviors. The empirical results indicate that arousal fully mediates the relationship between social gratifications and problematic social network service use. Furthermore, both interaction and arousal are partial mediators of the relationship between social gratifications and the intention to revisit social networking websites.     

Understanding User Behavior of Asking Location-Based Questions on Microblogs

With the prevalence of mobile social network services, people can post location-based questions on their social networks to satisfy their needs anytime anywhere. In this article, the authors study location-based questions that people post on microblogs, which is a popular form of social network service. The authors collected posts with geo-tags from Sina Weibo and conducted the study based on about a thousand location-based questions. Their results reveal unique characteristics of location-based questions by analyzing what people ask, how they ask, why they ask, and the context when asking. Location-based questions are closely related to people’s offline activities. Spatial restriction, subjectivity, interactivity, and propagation are the main characteristics that people value for choosing social networks to ask location-based questions. People also apply different phrasing skills to different types of questions. The questions people ask in different contexts also have different focuses. Based on their findings, the authors discuss practical design implications for social networks, location-based Q&A systems, and other applications with location-based features.     

Development, Distribution and Evaluation of Online Tourism Services in China

The development of electronic commerce (E-commerce) has led to great changes in the tourism industry in many countries around the world including China. The Chinese tourism industry has invested large amounts of money over last few years in the development of what is known as the 'Golden Tourism Project.' This study sheds more light on this project by investigating online tourism service development in China from three perspectives: the tourism website, the tourism website user and the tourism website provider. The results show that the majority of tourism website providers are regional tourism destination organizations that mainly provide comprehensive local tourism information and online services. The results also show the level of regional economic development has a significant impact on the construction of these local tourism websites. Through conducting a questionnaire survey, this paper identifies the types of web users and their evaluation for tourism websites. It assesses the level of current user satisfaction and discusses the principal barriers of implementation of online tourism services in China from a technical, financial and organizational point of view respectively. It is found that obtaining information is still the main aim of web users, however, the difficulties are slow Internet access and high fees. In conclusion, this paper proposes possible approaches to improve the quality of online tourism services in China.     

Topic extraction from extremely short texts with variational manifold regularization

Machine Learning - With the emerging of massive short texts, e.g., social media posts and question titles from Q&A systems, discovering valuable information from them is increasingly...     

Concept-based search and questionnaire systems

World Wide Web search engines including Google, Yahoo and MSN have become the most heavily-used online services (including the targeted advertising), with millions of searches performed each day on unstructured sites. In this presentation, we would like to go beyond the traditional web search engines that are based on keyword search and the Semantic Web which provides a common framework that allows data to be shared and reused across application. For this reason, our view is that “Before one can use the power of web search the relevant information has to be mined through the concept-based search mechanism and logical reasoning with capability to Q&A representation rather than simple keyword search”. In this paper, we will first present the state of the search engines. Then we will focus on development of a framework for reasoning and deduction in the web. A new web search model will be presented. One of the main core ideas that we will use to extend our technique is to change terms-documents-concepts (TDC) matrix into a rule-based and graph-based representation. This will allow us to evolve the traditional search engine (keyword-based search) into a concept-based search and then into Q&A model. Given TDC, we will transform each document into a rule-based model including it’s equivalent graph model. Once the TDC matrix has been transformed into maximally compact concept based on graph representation and rules based on possibilistic relational universal fuzzy-type II (pertaining to composition), one can use Z(n)-compact algorithm and transform the TDC into a decision-tree and hierarchical graph that will represents a Q&A model. Finally, the concept of semantic equivalence and semantic entailment based on possibilistic relational universal fuzzy will be used as a basis for question-answering (Q&A) and inference from fuzzy premises. This will provide a foundation for approximate reasoning, language for representation of imprecise knowledge, a meaning representation language for natural languages, precisiation of fuzzy propositions expressed in a natural language, and as a tool for Precisiated Natural Language (PNL) and precisation of meaning. The maximally compact documents based on Z(n)-compact algorithm and possibilistic relational universal fuzzy-type II will be used to cluster the documents based on concept-based query-based search criteria. This Paper is dedicated to Prof. Lotfi A. Zadeh, father of Fuzzy Logic “Zadeh Logic”.     

Internet Reference Services for Distance Education

SUMMARY

Published guidelines for distance learning library services provide a framework for distance education librarians to use in planning services for off-campus students. Other literature in the arena of distance education librarianship provides concrete examples of how reference services have been offered in real settings. This paper attempts to synthesize these two types of literature in order to offer models of reference service for distance learners.     

Impact of Tools and Techniques for Securing Consultancy Services

In a digital world moving at a breakneck speed, consultancy services have emerged as one of the prominent resources for seeking effective, sustainable and economically viable solutions to a given crisis. The present day consultancy services are aided by the use of multiple tools and techniques. However, ensuring the security of these tools and techniques is an important concern for the consultants because even a slight malfunction of any tool could alter the results drastically. Consultants usually tackle these functions after establishing the clients’ needs and developing the appropriate strategy. Nevertheless, most of the consultants tend to focus more on the intended outcomes only and often ignore the security-specific issues. Our research study is an initiative to recommend the use of a hybrid computational technique based on fuzzy Analytical Hierarchy Process (AHP) and fuzzy Technique for Order Preference by Similarity to Ideal Solutions (TOPSIS) for prioritizing the tools and techniques that are used in consultancy services on the basis of their security features and efficacy. The empirical analysis conducted in this context shows that after implementing the assessment process, the rank of the tools and techniques obtained is: A7 > A1 > A4 > A2 > A3 > A5 > A6 > A7, and General Electric McKinsey (GE-McKinsey) Nine-box Matrix (A7) obtained the highest rank. Thus, the outcomes show that this order of selection of the tools and techniques will give the most effective and secure services. The awareness about using the best tools and techniques in consultancy services is as important as selecting the most secure tool for solving a given problem. In this league, the results obtained in this study would be a conclusive and a reliable reference for the consultants.     

Website features that gave rise to social commerce: a historical analysis

Social commerce is a form of commerce mediated by social media and social network services (SNS). As a multifaceted phenomenon, social commerce can be studied from different angles and analyzed through the lens of various disciplines. This article examines website technical features to depict the transformation of e-commerce into social commerce. We first develop a conceptual framework to capture three emphases of e-commerce: transactional, relational and social. Then, we use the framework to conduct an historical analysis of the actual website screen captures for five top e-commerce companies since their websites were established. We were able to identify and classify a total of 174 emerging technical features. Our results show that: (1) all three emphases were expressed in the websites and have been reshaping their business and marketing strategies over the years; (2) there was a clear blooming of social features in 2007; and (3) there has been a significant effort to strengthen customer and merchant ties through relational features. Our findings signal that there still is room for further exploration of the social emphasis.     

基于Web2.0的教学多媒体共享平台应用探讨

论述了多媒体共享平台如何应用于教学中,介绍了Web2.0的基本概念与其在教育中的应用发展,分析了网络资源的分类与其作为开放学习的特性,随后介绍了网络教学多媒体共享平台的机制,进而分析其在教育与学习应用的相关实例。最后,提出对于教学多媒体共享平台在图书管理领域的可能应用建议。希望通过分析论述及创新观点的提出,能为相关研究者提供研究与应用的参考。     

What are mobile developers asking about? A large scale study using stack overflow

The popularity of mobile devices has been steadily growing in recent years. These devices heavily depend on software from the underlying operating systems to the applications they run. Prior research showed that mobile software is different than traditional, large software systems. However, to date most of our research has been conducted on traditional software systems. Very little work has focused on the issues that mobile developers face. Therefore, in this paper, we use data from the popular online Q&A site, Stack Overflow, and analyze 13,232,821 posts to examine what mobile developers ask about. We employ Latent Dirichlet allocation-based topic models to help us summarize the mobile-related questions. Our findings show that developers are asking about app distribution, mobile APIs, data management, sensors and context, mobile tools, and user interface development. We also determine what popular mobile-related issues are the most difficult, explore platform specific issues, and investigate the types (e.g., what, how, or why) of questions mobile developers ask. Our findings help highlight the challenges facing mobile developers that require more attention from the software engineering research and development communities in the future and establish a novel approach for analyzing questions asked on Q&A forums.