Understanding Mobile Q&A Usag e

Master’s ThesisUnderstanding Mobile Q&A Service Usage: An Ex-ploratory Study and New InterpretationHyanghong KangKnowledge Service EngineeringKAIST2012Understanding Mobile Q&A Service Usage: An Ex-ploratory Study and New InterpretationMKSE 20103022 강향홍. Kang, Hyanghong. Understanding Mobile Q&A Service: Exploratory Study and New Intertation모바일Q&A 서비스이해: 탐험적연구및새로운해석. Depart-ment of Knowledge Service Engineering. 2012. XX p. Advisor Prof. Jussi Kantola. Text in EnglishABSTRACTRecently questioning and answering (Q&A) communities that facilitate knowledge sharing among the people were introduced to the mobile environments such as Naver Mobile Q&A and ChaCha. Mobile Q&A is very different from conventional Q&A sites in that questions and answers are short in length and are ex-changed via wireless communications (e.g., SMS and mobile Internet). While conventional Q&A serviced have been investigated, so far little is known about the mobile Q&A usage. To understand mobile Q&A usage, I ana-lyzed over 2.4 million questions and answers pairs spanning a 13 month period from Naver Mobile Q&A and performed a complementary survey study of 274 active mobile Q&A users. I find that mobile Q&A is deeply wired into users’ everyday life activities - its usage is largely dependent on users’ spatial, temporal, social con-texts - and the key factors of mobile Q&A usage are accessibility/convenience of mobile Q&A, promptness of receiving answers, promptness of posting questions and users’satisficing behavior information seeking. Fur-thermore, I found a unique usage pattern that is asking same questions repeatedly in different from conventional Q&A service, and it have many meaning (e.g., estimating users’ expected time to get answers). This findings reported in this thesis have significant implications on the design of mobile Q&A system.Keywords: Q&A Systems, Online Community, Human FactorsTable of ContentsAbstract (i)Table of Contents (ii)List of Figures (iv)List of Tables (v)List of Terms (vi)Chapter 1. Introduction1.1 Research Background (1)1.2 Motivation and Contribution (2)1.3 Research Purpose (2)1.4 Dissertation Structure (4)Chapter 2. Related Works2.1 Introduction to Q&A Service (5)2.2 Analyses of Popular Q&A Service (6)2.2.1 Conventional Q&A Service (6)2.2.2 Social Q&A Service (7)2.2.3 Mobile Q&A Service (8)2.2 Mobile Web and Web Search Usage (9)2.2 Question Classification (9)Chapter 3. Methodology3.1 Data Collection (10)3.2 Interaction in Naver Mobile Q&A (11)3.3 Analysis for Context (12)3.4 Question Classification (12)3.5 Unique Usage Pattern (14)3.6 Overall Process of Analysis (14)Chapter 4. Result4.1 Why People Use Mobile Q&A (16)4.1.1 Everyday Life Information Seeking (16)4.1.2 Factors of Mobile Q&A Usage (16)4.1.2.1 Accessibility of Mobile Q&A (18)4.1.2.2 Promptness of Posting Questions (19)4.1.2.3 Promptness of Receiving Answers (23)4.1.2.4 Users’ Satisficing Behavior of Information Seeking (24)4.2 Questions in Mobile Q&A (25)4.3 Interactive Behavior: Repeating and Refining Questions (28)4.3.1 Repeating Questions (31)4.3.2 Refining Questions (35)Chapter 5. Conclusion5.1 Summary (38)5.2 Design Implication (39)5.1 Conclusion (40)References (41)List of FiguresFigure 1 Wilson T.D (1999) (3)Figure 2 Extracted data from web page of Naver Mobile Q&A (10)Figure 4 The steps of this thesis (15)Figure 5 The result of survey (17)Figure 6 The result of survey (18)Figure 7 Hourly usage percentage (19)Figure 8 Hourly usage percentage of sports questions in Mobile Q&A (21)Figure 10 Hourly usage percentage of location questions in Mobile Q&A (22)Figure 11 Hourly usage percentage of location questions in Conventional Q&A (22)Figure 12 Response time of Mobile Q&A (23)Figure 13 Topic distribution of CQA and MQA (Total sum of genral and under 14 age) (26)Figure 14 Topic distribution in CQA and MQA (under 14 years) (26)Figure 15 Sceniors of Interactive behavior (28)Figure 16 Users' expected waiting time distribution (30)Figure 17 Topic distribution of repeating and non-repeating question (32)Figure 19 Active Period (Top 20% of users) (34)Figure 20 User distribution of repeating and non-repeating (35)Figure 21 Fraction of repeated/refined questions per user (37)List of TablesTable 1 Top 10 Q&A services (5)Table 2 Classification of mobile Q&A service (8)Table 3 List of categories (genral and under 14) (25)Table 4 Quality analysis of topic (27)Table 5 Independent sample t-test (34)Table 6 Example of situation1 (29)List of TermsTerm MeaningMobile Q&A serviceKind of Q&A service that users can post questions via their phone (e.g.,ChaCha, Naver Mobile Q&A).Conventional Q&A serviceKind of Q&A service that users can post questions via their PC or laptop(e.g., Yahoo! Answers, Ask, Naver Kanowled-In).Repeating questions Behavior that a user posting the same questions repeatedly.Non-repeating questionsBehavior that a user posting questions only once.Response rate Proportion of questions having answers.Response timeGap of time between posting question and answering that question.Chapter 1.Introduction1.1Research BackgroundOver last several years, the number of internet users and the average time spent online have increased rapidly. The high amount of using internet makes it possible that people can communicate with diversity of people even they haven’t met before. These changes enabled the emergence of Q&A services such as Naver Kin, Yahoo Answer---“the place to ask questions and get real answers from real people.” In Korea, it became an integral part of Koreans.In recent years, a mobile version of such Q&A services has been introduced to the market with rise of population of mobile and smart phones such as Naver Mobile Q&A, Jisiklog and ChaCha. Users can post questions by using their mobile phone (sent via SMS or mobile App) and receive answers via their phone. The number of Mobile Q&A services has been increasing over the past several years; e.g., Naver Mobile Q&A was introduced in April 2010, and more than 500,000 people have posted more than 3 million of ques-tions so far. The following are some examples of typical questions:∙What is the result of a baseball game today?∙How much does it cost to take a cab from Seoul station to the city hall?∙Is Homeplus open today?Mobile Q&A services have the same purpose of asking/answering questions as in conventional Q&A services, but there are some different characteristics when they arare compared to conventional Q&A ser-vices:1.Questions are posted more frequently.2.The length of questions is shorter.3.Askers can not search similar questions they want.(1) Using a conventional Q&A service, users can post questions only when they have access to a computer. Mobile Q&A service users can post questions whenever and wherever they want because of the portability and accessibility of mobile phones. Therefore, questions are more frequently posted than on internet Q&A service. (2) Since there are limitations regarding sending messages by using a phone, users can post relative-ly short questions. For example, the maximum length of a question in Naver Mobile Q&A is 150 letters against 1000 letters in Naver Kin; conventional Q&A service. (3) In Mobile Q&A services, users can not search questions that have already been posted before sending their own question. Those characteristic dif-ferences definitely can give implications to serious design issues.1.2Motivation and ContributionConventional Q&A services have received a lot of attention from research community, and signifi-cant amount of research has been conducted about users’ usage pattern, motivation and the design issues of a Q&A service. However, little is known about Mobile Q&A services. With increasing mobile or smart phone usage, the number of Mobile Q&A service users has rapidly increased. So far little is known about who typi-cally uses a Mobile Q&A service, and what kinds of questions are asked. There are several successful Mo-bile Q&A services, and those services have enough data (questions and answers) and users to analyze. In terms of these contexts, I need to understand Mobile Q&A services better, and hopefully this thesis can con-tribute to first suggestion of guidelines for improving the design of Mobile Q&A services. Also this work is the first large scale study on Mobile Q&A service.If the usage pattern of a Mobile Q&A service is different or unique when it is compared to conven-tional Q&A service, I can extract new meaning from the Mobile Q&A dataset. Moreover, it will shed light on improving the existing service.In last, this thesis is not proving specific phenomenon, which have already happened in other do-main. This thesis tries to explore what is going on Mobile Q&A service. It is a first step to understand gen-eral phenomenon, and it should be derive many issues future research in terms of Mobile Q&A service.1.3Research PurposeIn this thesis, I attempt to clarify the differences between conventional Q&A service and Mobile Q&A service, the largest Q&A service in Korea. Naver Mobile Q&A service was recently launched in April2010. In this work, I analyze almost the Q&A data set from June 2010 to July 2011. The data set doesn’t in-clude from April 2010 to May 2010, because there was a special promotion and many people used Naver Mo-bile Q&A to get some reward. I also perform a survey study with 240 active Mobile Q&A users to answer the research questions.Figure 1 is Wilson’s model of information behavior [1]. In briefly, people have a context of infor-mation need, and people have and consider an intervening variable which means constraints. An activating mechanism invokes the explicit risk reward theory and the social learning theory. These theories depend on people’s characteristics: what is their priority motivation? So, an activating mechanism can be related with users. Lastly, people select how to search information under the system that they selected. An information seeking process in a Q&A service can be represented by information of questions, namely topic, posting time, length of questions, and so on. Let’s suppose that a man needs to know how to get to the city hall and there is nothing to consult with. In this context, he considers ‘how urgent this situation is’, ‘where he is’ and any con-straints he has. Once he recognizes his environment around him, then the next process is that he takes an ac-tion to find an answer; e.g., asking people around him, calling the city hall or asking a Q&A service. This pro-cess is largely influenced by his characteristics or previous experiences. If he is too timid to directly ask a question to someone, he may considerer using a mobile Q&A service because a mobile phone is one of the most convenient ways for communications. An activating mechanism process is not the end of an information seeking process because the answer mostly depends on how he asks. He could ask with specific situation or just ask how to get to the city hall. After this process he can get answers and if it is not enough he can repeat a similar process with a different information source.Figure 1 Wilson T.D (1999)This process can be also divided into three concepts. i.e., contexts, users, and questions/answers. In this thesis, I seek answers to the following research questions based on this high level model:Research question1. Why people use mobile Q&A services?Users have their contexts when they have questions, and those contexts are connected to users’ characteristics. Sometimes, those contexts directly connect to questions asked in Mobile Q&A service.Research question 2. What kinds of questions are frequently asked in Mobile Q&A?Since answerers never know what is the context of the askers, there are no direct connection points between askers and answerers. However, questions are connection points between them. Therefore, I study any rela-tions between questions and answers. For example, what types of questions are answered well, and what is the usual latency for receiving answers.Research question3. How do users interact with mobile Q&A to satiate their information needs Interact with Mobile Q&A is should be different because of different characteristics of mobile phones; display size, access to internet and maximum length of questions.1.4Dissertation StructureThis thesis tries to solve conceptual questions to prove future implementation opportunities in the aim of developing practical and effective solutions based on existing theories.The rest of this thesis is organized as follows. In the chapter 2, I first review the related work on conventional Q&A services, social Q&A services (real-time Q&A service) and Mobile Q&A services. In chapter 2, I explain features of Naver Mobile Q&A service that is the data set for analysis. The methodology of data collection and survey study are presented in chapter 3.In chapter 4, the results of analysis are shown according to contexts, users and questions, and I ex-plain the meaning of results in terms of research questions. In chapter 5, I present the design implications based on this work to Mobile Q&A services; which domains will be good for applying Mobile Q&A service and what are the concerns when Mobile Q&A services are launched. And also chapter 5 includes conclude this thesis.Chapter 2.Related works2.1Introduction to Q&A servicesAdamic et al. [2] mentioned that “Everyone knows something” when analyzing Yahoo! Answers, one of the largest Q&A services. A Q&A service is a web service that people use to ask questions, and to seek answers through the Web. The topics in a Q&A service are very diverse. More than hundreds of Q&A sites have been launched so far. Some of the largest Q&A sites are listed in Table 1[3].Table 1 Top 10 Q&A servicesName Website URL Launched in Unique Visitors*M 1996 54 2005 48MM 1999 41.9M 2006 9.9 2003 6.6MM 2003 5.5M 2006 5.2M 2007 4.4M 1998 4.3M 2003 1.4* Monthly unique visitor estimates reported by in 2011.Many researchers have tried to understand these Q&A services; to name a few, what the motivations of users are, what do users ask, what is the usage behavior of users? Especially, Yahoo Answers is analyzed by many researchers [2, 4-7]. The most sites are conventional Q&A services except ChaCha. ChaCha is one of the largest Mobile Q&A services where people can ask questions by SMS. ChaCha focuses on that people’s curiosity occur anytime and anywhere. As a result, more than 6 million people used the service so far. This trend of using a Q&A service also came up (e.g., Naver Knowledge-In, Naver Mobile Q&A, Daum Live and Jisjiklog) in Korea. In particular, Naver Knowledge-In has 16 million users, and this service has been investi-gated by Nam et al. [8].2.2Analyses of Popular Q&A serviceThere are three types of Q&A services, namely conventional Q&A services, social Q&A services, and Mobile Q&A services. Some of the large-scale Q&A services analyzed in the research community include Yahoo! Answer, Naver KiN, Microsoft Live Q&A, Google Answer, and Mahalo. In the following, I review the related work that characterizes the Q&A services.2.2.1Conventional Q&A serviceAdamic et al. [2] showed that in Yahoo! Q&A, interaction among users varies depending on the topic and is highly skewed (via the social network analysis), knowledge sharing patterns across different topic cate-gories (i.e., experts in different domains help one another) exist, and best answers can be predicted based on reply/thread length. In their analysis of Naver KiN, Nam et al. [8] studied the user behavior of answerers and found that the level of participation is highly skewed, answerers’ participation tends to be intermittent, and the expected level of their expertise is lower than that in specialized online forums.Questions in Q&A services are generally classified as soliciting facts, opinions, and suggestions [9]. Harper et al. [10] showed that the types of questions in Q&A services are broadly classified into two catego-ries, namely informational questions (e.g., fact and advice) and conventional questions (e.g., opinion and self-expression). Rodrigues et al. [11] found that Q&A services can be also used for socializing such as chatting and entertainment (puzzles, riddles), and unlike Yahoo Answer, a significant fraction of answers in Microsoft Live Q&A (16.4%) belongs to such a category.In Q&A services, the response time tends to be fairly long. Hsieh and Counts [12] reported that in Mi-crosoft Live Q&A, the average time to receive an answer took 2 hours and 53 minutes, and 20% of the ques-tions never received an answer. Zhang et al. [13] showed that in a specialized Q&A service for Java program language, the average time to receive an answer was around 9 hours.The motivation behind answering questions is largely dependent on a mixture of intrinsic factors (e.g., perceived ownership of knowledge, enjoyment and feelings of gratitude and respect) and extrinsic factors (e.g., monetary rewards, reputation systems) [14]. Online discussion forums such as Usenet groups do not offer any explicit incentive mechanisms and mostly rely on intrinsic factors. In contrast, Q&A services typi-cally employ incentive mechanisms to solicit users’ participation. Nam et al. [8] showed that the motivation of answering in Naver Knowledge-In is due to both intrinsic and extrinsic factors; i.e., altruism is the leading factor, followed by business motives, learning, hobby, and earning points (in its reputation system). When monetary rewards are involved, it is reported that fee-based services elicit better answers than none-fee based services (e.g., Google Answer vs. Yahoo! Answer), and higher quality answers can be acquired by paying more money [10]. Hsieh et al. [12] analyzed Mahalo, a fee-based Q&A service and showed that askers wish to pay when requesting facts and will pay more when asking difficult questions.2.2.2Social Q&A serviceOne of the ways to ask question through the Web is using SNS (Social Networking Service). Teevan and Panovich [15] analyzed what kinds of questions (topic, type) are asked through the Web and the motiva-tion (e.g., trust, answer speed, failed search) of using SNS to solve their curiousity. They surveyed 624 users who are actually using Twitter and Facebook, and organized users’ motivations and type of questions. Thom et al. [16] showed a large scale analysis of emergent question and answer (Q&A) behavior in an enterprise SNS and then described user motivations for social information seeking.2.2.3Mobile Q&A serviceThere was no large-scale analyses of Mobile Q&A service so far, even though there have been already successful Mobile Q&A services. Popular mobile Q&A services include ChaCha and Ask People in the US, AQA in the UK, Jisiklog, Naver Mobile Q&A and Daum Live Q&A in Korea. Mobile Q&A services in the market can be classified based on the following dimensions: (1) whether question posting is limited to mobile devices (i.e., SMS/MMS and mobile App/Web), (2) whether there is a monetary incentive mechanism, and if any, who pays the fees, and (3) whether a mobile Q&A service is mixed with a conventional Q&A service (i.e., presented together). Table 2 summarizes the classification of various mobile Q&A services. Note that those services that only support mobile devices (ChaCha, AQA, Naver Mobile Q&A) have the length limit on questions/answers as SMS/MMS is also used, but a mixed service like Ask People does not have such a con-straint. ChaCha originally accepted questions via SMS only, but this restriction was recently relaxed, and questions can now be posted via the Web. In this thesis, I analyze the usage patterns of Naver Mobile Q&A, a representative mobile Q&A service as it only supports question posting via mobile devices. I believe that this work sheds lights on understanding the overall usage patterns of mobile Q&A systems in general.Table 2 Classification of mobile Q&A serviceInteraction Method Incentive Mixed Naver Mobile Q&A SMS, Mobile App/Web Reputation No ChaCha SMS, Mobile App/Web, Web ChaCha pays -Jisiklog SMS, Mobile App Asker pays -AQA SMS, Mobile App Asker pays - Ask People Mobile App/Web, Web Reputation YesDaum Live Mobile App/Web, Web Reputation No2.3Mobile Web and Web Search UsageThe taxonomy of web tasks is generally classified as information seeking, casual browsing, communications (e.g., chatting), transaction (e.g., bank services) [17]. In their recent study, Cui and Roto [18] reported that in mobile environments, tasks are rather limited to information seeking, communications, and transaction with an additional task of maintaining personal content as a part of personal space extension. The user behavior study about mobile Internet shows that such web tasks are largely dependent on users’ contexts; e.g., personal context (e.g., goals) and environmental context (e.g., distractions, social interactions) [17, 18]. Kamvar and Baluja [19] performed quantitative analysis of mobile web search usage from Google Mobile Search data set in 2004 (connected via mobile Internet). The overall topics are quite consistent with the desktop search statis-tics. Popular topics include adults (e.g., “free porn”), computer and communications (e.g., “free ringtones”), and entertainments (e.g.,“ imdb”). The average number of words per query is 2.3, and that of letters is 15.5. Users tend to perform a series of related queries by refining/repeating previous queries, including automatic spell correction. In their recent study on mobile search pattern evolution, Yi and Maghoul [20] reported that the query topics experienced significant change (consistent with desktop search) when comparing the data set from year 2007 and 2010 (e.g., decrease of entertainment queries, increase of local and commerce queries). In this thesis, I discuss mobile Q&A communities that are very different from traditional web search. This work compliments the prior research on the user behavior of mobile web search in that we use a large-scale mobile Q&A data set supplemented with a survey study of actual users to characterize information seeking behavior with mobile Q&A.2.4Question ClassificationQuestion classification is one of the key points in analyzing Q&A service. Usually, questions in con-ventional Q&A service have their categories classified by askers. Therefore, most of papers [11, 15] directly used those categories. If there were not any benchmark categories, separate question classification is needed. Teevan and Panovich [15] used manual classification in their research. Differently, Zhang and Lee [21] compared among automatic classification method; Nearest Neighbors, Naïve Bayes, Decision Tree, Sparse Network of Windows, and Support Vector Machines.Chapter 3.Methodology3.1Data CollectionTo understand Mobile Q&A service, I gathered data of Naver Mobile Q&A service [22], which is one of the most famous Mobile Q&A services in Korea. More than 5000 questions are usually posted every day. I could extracted total seven fields from the answer pages7 kinds of information as shown in Figure 2.Figure 2 Extracted data from web page of Naver Mobile Q&AThe explanation of data shown in this format:1. Question2. Asker’s ID3. Time that question is posted4. Number of answers5. Answers6. Time that answer is posted7. Answerer’s IDThe crawled data is from 1st June 2010 to 31th July 21, 2011. Naver Mobile Q&A was actually launched in April 2010. However, there was a special promotion from April to May 2010. Since promotion may influence some part of users’ motivation, I decided to omit the data from the promotion period.The total number of questions and answers in the dataset is 2,483,624 and 3,086,365, respectively. The average number of answers per question in the dataset is 1.27. In contrast to conventional Q&A services, the categories of questions are not specified because it allows users to post questions via SMS (short message service). That means GUI can’t be provided to users, so only textual information is exchanged.To compare between Mobile Q&A service and conventional Q&A service, I crawled Naver Knowledge-in (from 2011/10/03 to 2011/10/15) [23]. Naver Knowledge-In is provided by same company of Naver Mobile Q&A.3.2Interaction in Naver Mobile Q&AIn Naver Mobile Q&A, users can post questions only via mobile devices; a list of supported methods of com-munications includes SMS/MMS, Mobile Web, and Naver App. Since it basically uses SMS, there is a strict limit on the length of a question/answer, i.e., 150 letters. For a given question, users can also post the related image using MMS. Registered users of Naver can post the questions on the Web. As shown in Figure 2, a requester’s user ID is revealed along with the delivery method (e.g., SMS, MobileWeb, and Naver App). Nav-er currently restricts that a user can post at most 30 questions a day.The registered users can post answers through the Naver Mobile Q&A Web site as in the conventional Q&A sites (see Figure 2). The front page of the Web site presents a list of 20 most recent questions (in the form of the upper box in Figure 2 stacked). The old lists can be browsed by clicking the next button located at the bot-tom of the page. The questions are ranked on the basis of recency; whenever a new question arrives, the entire rank trickles down by one, and the 20th ranked question on the front page will be placed on the second page. Answerers can pause/resume real-time shuffling of questions using the pause/play button located at the top of the page. Requesters can choose to receive answers via SMS (regardless of whether they posted questions via Mobile Web or Naver App). Due to the financial burden involved, Naver Mobile Q&A only delivers at most three answers via SMS. If more than three answers are made, the rest can be accessed via mobile Web and Naver App. Naver has the point based reputation system of answerers. A user can earn 10 points for each an-swer. The following are the notable differences from the traditional Q&A: (1) a question is expired after 24 hours, and expired questions cannot be answered; (2) users cannot edit the submitted questions/answers; (3) there is no option for selecting the best answer, and (4) there is no predefined question category.3.3Analysis for Understanding ContextsFor the context that people use Mobile Q&A site, I analyze general phenomena such as number of questions per hour (or day), inter-arrival time of posting questions and response time to get answers. The re-sults from this analysis will be compared with people’s common life pattern, as it can explain users’ general context. To better explain the context of question asking, I show the topics of questions in conventional Q&A service and Mobile Q&A service.I surveyed people who are actually using the mobile Q&A service in order to identify the context of Mobile Q&A service users. Total 30,000 Naver Mobile Q&A users are randomly selected from the entire user set, and 274 users answered our survey questionnaire. The first part of the survey asked how users are famil-iar to use mobile phone and their general motivations to a Mobile Q&A service. In the second part of the sur-vey, I showed recent questions which user actually asked, and asked users to describe their specific situations when they post questions. The result of second survey can explain more detail contexts.3.4Question ClassificationFigure 3 The process of question classificationIn contrast to a conventional Q&A service, the categories of questions are not present in Mobile Q&A. To know what kinds of questions are posted in Mobile Q&A service, topic classification is necessary. Teevan and。