Session I (March 3, 8:00 – 10:00 AM)
Networks are ubiquitous today, linking organizations, group members, and computers around the world. Analysis of these networks has become an important tool in many organizations. Decision-making in fields such as criminal investigation, expert assessment, community understanding, and crisis management, relies increasingly on analysis of network information. Practitioners in these and other fields are demanding support for effective network analysis, and various computerized network visualization tools have been developed in response. At the same time, many formal quantitative concepts such as centrality, betweenness, and structural equivalence have been invented in the field of social network analysis to measure and to represent different network topological features, including sub-groups, key players, and structurally equivalent members. Although those key concepts could be very helpful and have been commonly applied for the understanding of a network, they are usually displayed in the tabular format consisting of only rows and columns of data that may not be easily comprehended by a user. Various structural patterns embedded in the quantitative results in this format thus can be hardly visible to users. Therefore, supporting the visual comprehension of these key network concepts becomes more and more significant. Without such visual support, the comprehension of those network concepts can be a difficult task to decision-makers. On the other hand, most network visualization research has focused on the aesthetic aspect of network visualization, assuming that a clearly drawn network representation will deliver topological features automatically. This assumption, however, does not always hold true, especially when a user needs to take a cognitive inference to grasp such quantitative concepts as structural equivalence (or betweeness) from a point and line representation. The cognitive effort required could easily overload him/her and thus hinders the comprehension of those important concepts. This paper proposes a novel network representation that harnesses the results from both previous visualization and social network analysis researches to identify and deliver structural patterns of a network. The paper also describes the implementation of the proposed algorithm on the criminal network data from Arizona Tucson Police Department. The effectiveness of the algorithm is further validated by a lab experiment that compares the algorithm proposed with the conventional representation approach. The results indicate that the proposed representation facilitates not only a better comprehension of concepts (such as betweenness, centrality, and gatekeepers of groups), but also a faster identification of sub-groups and structural equivalent network actors, in comparison with its conventional counterpart.
In today's extremely dynamic business world, understanding relationships between firms and technologies represents an important, but challenging task for industry analysts, experts, and senior management. In this paper we discuss the complexity of these relationships, review some of the prior work on this issue, and present a novel graph-theoretic approach for modeling some aspects of technology/firm relationships in technology ecosystems.
In an effort to reveal the fine-grained relationships between IT use, the structure of information flows, and individual information-worker productivity, we study task level practices of information workers at a midsize executive recruiting firm. We conducted both project level and individual level analyses using: (1) detailed accounting data on revenues, completion rates, team participation and compensation for over 1300 projects over 5 years, (2) data on a matched set of individual workers self-reported information technology skills, use and information sharing, and (3) direct observation of over 125,000 e-mail messages over a period of 10 months by these same workers. These data make it possible to develop and econometrically test a multistage model of production and interaction activities at the firm, and to analyze the correlations among key technologies, work practices, and output. We find that (a) an inverted-U shaped relationship exists between multitasking and productivity such that after a certain threshold, there are diminishing and negative productivity returns to more multitasking, (b) information technology use and skills are positively correlated with increased revenues and project completion; (c) asynchronous information seeking such as email and database use promote multitasking while synchronous information seeking over the phone shows a negative correlation and (d) the structure and size of workers’ communication networks, including such social network metrics as betweeness and structural holes are highly correlated with performance. We also find evidence of specialization in information seeking practices and a division of labor across these practices at the team level. Overall, these data show a statistically significant and positive relationship among technology use, social network characteristics, completed projects and revenues for project-based information workers. The results are consistent with simple models of queuing and multitasking and the methods can be replicated in other settings, suggesting a new frontier for IT value research.
Session II (March 3, 7:00 – 9:00 PM)
Companies use business
intelligence technologies in order to document, analyze, predict, and respond
to changes in their competitive environment and internal operations. To achieve
this goal, business intelligence tools require large amounts of data that comes
in many forms from many internal and external information systems (IS). What
impact does the existing information technology (IT) infrastructure of a
company and of its clients and partners have on the firm’s ability to
successfully exploit business intelligence tools? How can companies design
innovative IT and business process solutions that enhance their business
intelligence efforts? In this paper, I investigate these questions in the
context of the corporate travel industry. The paper presents a detailed case
study of one of the top 3 corporate travel agencies in the
Consumers increasingly obtain direct
consultations from experts, thanks to the prevalent use of Information
Technology and the Internet. Motivated
by the online consultation practices of reputed institutions such as
There are currently many active movements towards computerizing patient healthcare information, which has been widely acknowledged as being both crucial and long overdue. As Electronic Medical Record (EMR) systems become widely adopted and deployed in the foreseeable future, the next big challenge will be the development of methods to effectively utilize this massive information source. Based on the observation that simple text-word based information retrieval fails to yield satisfactory performance—due to the complex semantic relationships among symptoms and diagnostic/therapeutic processes—we propose a framework for developing an intelligent information retrieval system which can serve as an adjunct to a unified EMR system. The proposed framework integrates various technologies, including information retrieval, domain ontologies (e.g., the Unified Medical Language System), automatic semantic relationship learning (e.g., term co-occurrence), as well as a body of domain knowledge elicited from healthcare experts. This will allow effective retrieval of healthcare records related to a patient's presenting signs and symptoms. Knowledge of semantic relationships among medical concepts, such as symptoms, exams and tests, diagnoses, and treatments, as well as knowledge of synonyms, hypernym/hyponyms, is used to expand and enhance initial queries posed by a user. Such a semantic retrieval system can liberate doctors from the daunting task of digging into voluminous medical records for a few relevant documents, with a consequent improvement in healthcare efficiency and quality. We will discuss the major challenges anticipated and research issues, and will outline our research plan in the paper. We expect to elicit comments and will welcome suggestions from the discussants at the conference.
Session III (March 4, 7:30 – 9:00 AM)
Shawndra Hill, NYU
Network-based marketing refers to a collection of marketing techniques which take advantage of links between consumers to increase sales. We are most interested in the consumer networks formed using direct interactions (e.g., communications) between consumers. We survey the diverse literature on such marketing with an emphasis on the statistical methods used and the data to which these methods are applied. We also provide a discussion of challenges and opportunities for this burgeoning research topic. Our survey highlights a gap in the literature. Because of inadequate data, prior studies have not been able to provide direct, statistical support for the hypothesis that network linkage can directly affect product/service adoption. We present a new data set regarding a new telecommunications service and show very strong support for the hypothesis. Specifically, we show three main results. 1) “Network neighbors”—those consumers linked to a prior customer—adopt the service at a rate 3-5 times greater than baseline groups selected by the best practices of the firm’s marketing team. In addition, analyzing the network allows the firm to acquire new customers that otherwise would have fallen through the cracks, because they would not have been identified based on traditional attributes. 2) Statistical models, built with a very large amount of geographic, demographic, and prior purchase data, are significantly and substantially improved by including network information. 3) Fine-grained network information allows the ranking of the network-neighbors so as to allow the selection small sets of individuals with very high probabilities of adoption.
Up to 80 percent of a company’s information is found in textual databases. Mining this textual data can help firms manage their knowledge assets. Finding and naming entities can help firms retrieve useful information and track relevant things. We present a lexical representation for finding and naming entities that contains both semantic and syntactic information. We evaluated our extraction tool first on the corpus of Message Understanding Conference-7 (MUC-7) documents and then on a corpus of financial news documents retrieved from the Internet. On the MUC-7 corpus, our algorithm achieved a 90 percent f-score, near the top performers for that year. On the finance corpus, the algorithm achieved a 93 percent f-score. In addition to the seven MUC entities, we identified three additional entity types in the finance documents. The minimal additional training to prepare the algorithm for the finance documents and the good performance are encouraging as to the portability of the algorithm between content domains.
Web search engines typically provide search results without considering user interests or context. We propose a personalized search approach that can easily extend a conventional search engine on the client side. Our mapping framework automatically maps a set of known user interests onto a group of categories in the Open Directory Project (ODP) and takes advantage of manually edited data available in ODP for training text classifiers that correspond to, and therefore categorize and personalize search results according to, user interests. In two sets of controlled experiments, we compare our personalized categorization system (PCS) with a list interface system (LIS) that mimics a typical search engine and with a non-personalized categorization system (NPCS). In both experiments, we analyze system performances on the basis of the type of task and query length. We find that PCS is preferable to LIS for Information Gathering type of task and for searches with short queries, and PCS outperforms NPCS in both Information Gathering and Finding type of task, and for searches associated with free-form queries. From the subjects’ answers to a questionnaire, we find that PCS enables a user to find relevant Web pages quicker and easier than LIS and NPCS.