Knowledge sharing through online communities of practice: the case of linux ports to embedded processors

IADIS MULTI CONFERENCE ON COMPUTER SCIENCE AND INFORMATION SYSTEMS call for papers 21 - 23 June Proceedings Edited by: Gunilla Bradley Piet Kommers international association for development of the information society Algarve, Portugal of ICT, SOCIETY AND HUMAN BEINGS 2009 and WEB BASED COMMUNITIES 2009 IADIS INTERNATIONAL CONFERENCE ICT, SOCIETY AND HUMAN BEINGS 2009 and IADIS INTERNATIONAL CONFERENCE WEB BASED COMMUNITIES 2009 part of the IADIS MULTI CONFERENCE ON COMPUTER SCIENCE AND INFORMATION SYSTEMS 2009 iii PROCEEDINGS OF THE SECTION I IADIS INTERNATIONAL CONFERENCE ICT, SOCIETY AND HUMAN BEINGS 2009 SECTION II IADIS INTERNATIONAL CONFERENCE WEB BASED COMMUNITIES 2009 part of the IADIS MULTI CONFERENCE ON COMPUTER SCIENCE AND INFORMATION SYSTEMS 2009 Algarve, Portugal JUNE 21 - 23, 2009 Organised by IADIS International Association for Development of the Information Society iv Copyright 2009 IADIS Press All rights reserved This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, re-use of illustrations, recitation, broadcasting, reproduction on microfilms or in any other way, and storage in data banks. Permission for use must always be obtained from IADIS Press. Please contact [email protected] Volume Editors: Gunilla Bradley and Piet Kommers Computer Science and Information Systems Series Editors: Piet Kommers, Pedro Isaías and Nian-Shing Chen Associate Editors: Luís Rodrigues and Patrícia Barbosa ISBN: 978-972-8924-82-9 SUPPORTED BY SECTION II IADIS INTERNATIONAL CONFERENCE WEB BASED COMMUNITIES 2009 part of the IADIS MULTI CONFERENCE ON COMPUTER SCIENCE AND INFORMATION SYSTEMS 2009 iv SHORT PAPERS A MEASUREMENT MODEL ON BLOGGERS’ SELF-DISCLOSURE Jih-Hsin Tang and Cheng-Chung Chen 87 EXPERIENCING CHANGE THROUGH THE ADOPTION OF NEW DIGITAL TECHNOLOGY Mitul Shukla, Nik Bessis and Alfredo Gaitan 91 ANALYSIS OF ONLINE PARTICIPATION IN OPEN SOURCE COMMUNITIES M. R. Martínez-Torres, S. L. Toral and F. Barrero 96 SEARCHING FOR MARKET TRENDS AND BUSINESS OPPORTUNITIES IN WEB COMMUNITIES Graziella Martins Caputo and Nelson Francisco Favilla Ebecken KNOWLEDGE SHARING THROUGH ONLINE COMMUNITIES OF PRACTICE: THE CASE OF LINUX PORTS TO EMBEDDED PROCESSORS S. L. Toral, F. Barrero and M. R. Martínez-Torres 102 107 CYCLIC ENTROPY OPTIMIZATION OF A SCALE-FREE SOCIAL NETWORK USING EVOLUTIONARY ALGORITHM Maytham Safar, Nosayba El-Sayed and Khaled Mahdi 113 DEVELOPMENT OF A CUSTOMER SATISFACTION MODEL FOR ENABLING E-COLLABORATION IN SECOND LIFE Mitul Shukla, Nik Bessis, Marc Conrad and Gordon Clapworthy 119 CHARACTERIZING COLLABORATIVE SOCIAL NETWORKS USING CYCLIC ENTROPY, CASE STUDY: WIKIPEDIATM Khaled Mahdi, Lama Jammal and Maytham Safar 125 HOW DOES SERVICE QUALITY AFFECT CUSTOMER SATISFACTION IN PERSONALIZED COMMUNITY SERVICE USAGE? Su Jin Kim and Bong Gyou Lee 131 USING A SOCIAL NETWORK TO SUPPORT PROJECT MANAGEMENT Paulyne Jucá, Ricardo Costa and Silvio Meira 136 A MULTIDIMENSIONAL MODEL OF TRUST AND ITS ANTECEDENTS FOR KNOWLEDGE SHARING IN VIRTUAL COMMUNITIES Abel Usoro and Grzegorz Majewski 140 REFLECTION PAPERS RIMI ONLINE COLLABORATIVE COMMUNITY Amy S. C. Leh 147 CLASSIFYING KNOWLEDGE ON SOCIAL NETWORKS: A REAL CHALLENGE Ricardo Costa, Paulyne Jucá and Silvio Meira, Isabel Mendonça and Haidee Lima AUTHOR INDEX 151 KNOWLEDGE SHARING THROUGH ONLINE COMMUNITIES OF PRACTICE: THE CASE OF LINUX PORTS TO EMBEDDED PROCESSORS S. L. Toral, F. Barrero E. S. Ingenieros, University of Seville * Avda. Camino de los Descubrimientos s/n, 41092, Seville, Spain M. R. Martínez-Torres E. U. Estudios Empresariales, University of Seville Avda. San Francisco Javier, s/n 41018 Sevilla (Spain) ABSTRACT The paper explores knowledge sharing processes in online communities analyzing their explicit and stored messages. Their content is processed using latent semantic analysis techniques, obtaining several indicators related to knowledge sharing activities. A factor analysis is then applied to obtain the main dimensions affecting knowledge sharing. The obtained results provide new insights in the underlying knowledge processes in online communities of practice. KEYWORDS Knowledge sharing, online communities, communities of practice, factor analysis. 1. INTRODUCTION The notion of community has been at the heart of the Internet since its inception (Lesser et al., 2000). Initially, Internet was used by scientist to share knowledge, collaborate on research and exchange messages, and today, million of Internet users worldwide communicate themselves using electronic tools. The advent of Web 2.0 and social software have propitiated the organization of users around communities of interest. The distinctive feature of these online communities is the intensive use of electronic media for people getting in contact. The theoretical background behind online communities has been treated by numerous authors. Some of them (Preece, 2001) highlight the connection of online communities with the social learning theory and communities of practice developed by Wenger (1998), while others are focused on their relation with knowledge sharing, knowledge creation, and innovation models (Lee and Cole, 2003; Kuk, 2006). Of particular importance are the online communities supporting OSS (Open Source Software) projects, as they are changing the way in which software is produced. Several case studies can be found on the literature. Mockus et al. (2002) raise some questions about OSS development and analyze two case examples based on Apache and Mozilla projects. Lee and Cole (2003) focused on the most well known OSS project: Linux. In this context, knowledge sharing means transforming individual knowledge into collective, organizational knowledge (Liebowitz, 2001). That is precisely one of the main objectives of a community of practice. Knowledge sharing is not stimulated by imposing structures and tools but by rich social interaction and its immersion in practice (van den Hooff & Huysman, in press). Consequently, knowledge sharing can be considered one of the most important processes involved in the development of an online community. The purpose of this paper consists of analyzing the main dimensions affecting knowledge sharing activities in online communities, using their stored information as a staring point. This information will be processed using latent semantic analysis techniques, extracting several indicators related to knowledge sharing, and these indicators will be statistically treated to identify the critical dimensions related to knowledge sharing. IADIS International Conference on Web Based Communities 2009 107 2. KNOWLEDGE SHARING In the early days of knowledge management, knowledge was often considered an object, which could be captured, codified and stored with the aid of information technologies. Knowledge management was mainly focused on optimizing these three processes. The problem of this approach is that the stored knowledge did not reflect real practices (Wenger et al., 2002). The insight that knowledge is not simply an aggregate of information that could be de-coupled from its context was then introduced and attention shifted to the idea that knowledge is socially embedded in the context where it takes shape (van Hooff & Huysman, in press). The online community can be defined as a social relationship aggregation, facilitated by Internet-based technology, in which users communicate and build personal relationships (Rheingold 1993). Individuals engage in knowledge sharing, problem solving, and learning through posting and responding to questions on professional advice, storytelling of personal experiences, and debate on issues relevant to the network. Examples of online communities can be found on fields like education, software development or consumer behaviour (Toral et al., 2005; Shang et al., 2006). Online communities have been frequently connected with communities of practice (Lin and Lee, 2006), in the sense that communities develop their own routines, formal and informal ‘‘rules’’, and practices evolve as a result of learning. The concept of Communities of Practice (CoP) was developed by Lave & Wenger (1991). This concept refers to the process of social learning that occurs when people who have a common interest in some subject or problem collaborate over an extended period to share ideas, find solutions, and build innovations. CoPs are not formal structures, such as departments or project teams. Instead, they are informal entities, which exist in the mind of their members, and are glued together by the connections the members have with each other, and by their specific shared problems or areas of interest (Wenger and Snyder, 2000). Intangible, tacit knowledge embedded in an organization's members is an asset that is not easy to capture. CoPs, however, offer a practical mechanism to help their members share and internalize tacit knowledge (Wang et al., 2008). Several researchers have noted that CoPs appear to be a more effective tool for dealing with unstructured problems and knowledge sharing/ creation than traditional and formal ways of structuring interaction in organizations. Understanding the processes and mechanisms that enable members to share knowledge in CoPs is very important for knowledge sharing within and between such communities (Pan and Leidner, 2003). A methodology based on semantic analysis technoques will be used for the identification of the main dimension with an influence in knowledge sharing. Modern semantic analysis techniques are based vector space model, in which documents are summarized and represented by vectors of words (term vectors). However, the main problem of such representation is the high dimensionality of the feature space (one dimension for each unique word). Therefore, it is desirable to first project the documents into a lower-dimensional subspace in which the semantic structure of the document space becomes clear (Cai et al., 2005). In the low-dimensional semantic space, the traditional clustering algorithms can be then applied. For instance, clustering using Latent Semantic Indexing (Zha et al., 2001) is one of the most well-known techniques. Different to LSI, in this paper we have applied the topic model, which is a statistical language model that relates words and documents through topics. It is based upon the idea that documents are mixtures of topics, where a topic is a probability distribution over words (Blei et al., 2003; Hofmann, 2001). Representing the content of words and documents with probabilistic topics has one distinct advantage over the purely spatial representation of LSI. Each topic is individually interpretable, providing a probability distribution over a word that picks out a coherent cluster of correlated terms. Hofmann (2001) introduced the probabilistic topic approach to document modelling in his Probabilistic Latent Semantic Indexing method (pLSI). Blei et al. (2003) extended this model by introducing a Dirichlet prior, calling the resulting generative model Latent Dirichlet Allocation (LDA). In both cases, the basic idea is that documents are represented as random mixtures over latent topics, where each topic is characterized by a distribution over words. Given T topics, the probability of the ith word in a given document can be written as ∑ = === T j iiii jzPjzwPwP 1 )()|()( (1) where zi is a latent variable indicating the topic from which the ith word was drawn and P(wi|zi=j) is the probability of the word wi under the jth topic. P(zi=j) gives the probability of choosing a word from topics j in the current document, which will vary across different documents. Intuitively, P(w|z) indicates which ISBN: 978-972-8924-82-9 © 2009 IADIS 108 words are important to a topic, whereas P(z) is the prevalence of those topics within a document. Latent Dirichlet Allocation combines Eq. (1) with a prior probability distribution to provide a complete generative model for documents. The Latent Dirichlet Allocation algorithm involves the selection of several parameters to achieve a topic description of the domain under study. Table 1. Dimensions involved in LDA algorithm. Parameter Description D Number of documents in corpus W Number of words in vocabulary N Total number of words in corpus L Average length of document in words (L = N/D) T Number of topics ITER Number of iterations Table 2. Selected indicators Indicato r Description I1 Number of topics I2 Polisemy I3 Rated Polisemy I4 Average messages size (characters) I5 Average number of messages per topic I6 Messages per topic distribution I7 Average messages size (words) I8 Size of threads I9 Average number of threads per topic I10 Threads per topic distribution Table 1 details the key dimensions or size parameters describing a corpus (D, W, N and L) and a topic model run (T and ITER). In our study, we will analyze several online communities, extracting several indicators related to knowledge sharing by applying the topic model for each year. Selected indicators are detailed in Table 2. The first indicator I1 is the number of topics. This is a parameter of the topic model. Consequently, it has been chosen attending to the perplexity criterion. Perplexity is a standard measure of performance for statistical models of natural language (Manning & Schutze, 1999), and it is defined as: ⎟ ⎠ ⎞ ⎜ ⎝ ⎛−= ∑=)|(log 1 exp 1 W nnn dwP W pplex (2) Perplexity varies from 1 to W; lower perplexity is better, and the maximum perplexity of W is reached when all words in the vocabulary are equally likely. In our case, LDA algorithm has been run for a number of topics varying between 1 and 50, selecting the number of topics leading to a minimum perplexity value. Indicators I2 and I3 are related to polysemy. Generative models have the ability of capturing polysemy, where the same word has multiple meanings. This is because they do not impose restrictions about mutual exclusivity that restricts words to be part of one topic. I2 measures the polysemy as the number of times a word wi appears more than once in different topics while I3 consider this previous value rated with the probability P(wi|zi=j) of the word wi under the jth topic. The next four indicators are related to messages. They consider the average size of messages in characters and words, the average number of messages per topic and their distribution over topics. Finally, the last three indicators are related to threads of discussion. It is usual that online communities are organized by threads of discussion. Threads are groups of messages sharing the same subject. A thread is initiated by someone who posts a message asking for help, suggesting some improvements, or just considering some new idea. Then people start answering this initial message, posting possible solutions, sources of information or just extending posted considerations. The indicator I8 relative to the size of threads considers those threads with al least one answer. Indicators I9 and I10 are the same than I5 and I6 but using threads instead of topics as the unit of analysis. 3. CASE STUDY The case study is based on Linux ports to embedded processors. Linux is a PC-based operating system that has been developed as Open Source Software along the structure of the UNIX operating system, and it is one of the most prominent examples of OSS projects. The same than Windows is the most prominent operating system released under a proprietary software license, Linux is the most prominent operating system released under a free license like GPL. Although Linux started as a hobby in 1991, it represents today a serious threat to Microsoft Windows’s market dominance in operating systems. Nevertheless, the proposed case study will be focused on Linux ports to other processor architectures not intended for desktop or personal computer market. There are several reasons for this choice. First, Linux is firmly in first place as the operating system IADIS International Conference on Web Based Communities 2009 109