Contributions in Merlin

Extracts of this paper were presented at the conference "Work Beyond 60: Preparing for the Demographic Shock", 6–7 March 2003 in Vienna organized by The Geneva Association, The Club of Rome, and The Risk Institute. Parts were also presented at the Bertelsmann Foundation conference "Strategien gegen den Fachkräftemangel" in Berlin, 2 July 2002 and at the Bertelsmann Foundation conference "Reformen zur Steigerung der Beschäftigungsfähigkeit älterer Arbeitskräfte" in Berlin, 26 October 2001. The authors would like to thank the participants as well as Jaap van Dam, Thomas Liebig, Fred Henneberger, and Geneviève Reday-Mulvey for their valuable comments and discussions. Alfonso Sousa-Poza would like to thank the Swiss National Science Foundation for financial assistance. The usual disclaimer applies.

Large amounts of technical documentation are available in machine readable format, however there is a lack of effective ways to access them. In this paper we propose an approach based on linguistic techniques, geared towards the creation of a domain-specific Knowledge Base, starting from the available technical documentation. We then discuss an effective way to access the information encoded in the Knowledge Base. Given a user question phrased in natural language the system is capable of retrieving the encoded semantic information that most closely matches the user input, and present it by highlighting the textual elements that were used to deduct it.

This article develops a theory of multiunit auctions where short squeezes can occur in the secondary market. Both uniform and discriminatory auctions are studied and bidders can submit multiple bids. We show that bidders with short and long preauction positions have different valuations in an otherwise common value setting. Discriminatory auctions lead to more short squeezing and higher revenue than uniform auctions, ceteris paribus. Asymptotically, as the auction size approaches infinity, the two formats lead to equivalent outcomes. Shorts employ more aggressive equilibrium bidding strategies. Most longs strategically choose to be passive. Free riding on a squeeze by small, long players has no impact on these results, but affects revenue in discriminatory auctions.

This paper addresses the classification of linked entities. We introduce a relational vector-space (VS) model (in analogy to the VS model used in information retrieval) that abstracts the linked structure, representing entities by vectors of weights. Given labeled data as background knowledge/training data, classification procedures can be defined for this model, including a straightforward, “direct” model using weighted adjacency vectors. Using a large set of tasks from the domain of company affiliation identification, we demonstrate that such classification procedures can be effective. We then examine the method in more detail, showing that as expected the classification performance correlates with the relational autocorrelation of the data set. We then turn the tables and use the relational VS scores as a way to analyze/visualize the relational autocorrelation present in a complex linked structure. The main contribution of the paper is to introduce the relational VS model as a potentially useful addition to the toolkit for relational data mining. It could provide useful constructed features for domains with low to moderate relational autocorrelation; it may be effective by itself for domains with high levels of relational autocorrelation, and it provides a useful abstraction for analyzing the properties of linked data.

This paper describes a novel theoretical and empirical approach to tasks such as business process redesign and knowledge management. The project involves collecting examples of how different organizations perform similar processes, and organizing these examples in an on-line ìprocess handbook"". The handbook is intended to help people: (1) redesign existing organizational processes, (2) invent new organizational processes (especially ones that take advantage of information technology), and (3) share ideas about organizational practices. A key element of the work is an approach to analyzing processes at various levels of abstraction, thus capturing both the details of specific processes as well as the ""deep structure"" of their similarities. This approach uses ideas from computer science about inheritance and from coordination theory about managing dependencies. A primary advantage of the approach is that it allows people to explicitly represent the similarities (and differences) among related processes and to easily find or generate sensible alternatives for how a given process could be performed. In addition to describing this new approach, the work reported here demonstrates the basic technical feasibility of these ideas and gives one example of their use in a field study.

In the past, most collaboration support systems have focused on either automating fixed work processes or simply supporting communication in ad-hoc processes. This results in systems that are usually inflexible and difficult to change or that provide no specific support to help users decide what to do next. This paper describes a new kind of tool that bridges the gap between these two approaches by flexibly supporting processes at many points along the spectrum: from highly specified to highly unspecified. The development of this approach was strongly based on social science theory about collaborative work.

null

We present a Question Answering system for technical domains which makes an intelligent use of paraphrases to increase the likelihood of finding the answer to the user's question. The system implements a simple and efficient logic representation of questions and answers that maps paraphrases to the same underlying semantic representation. Further, paraphrases of technical terminology are dealt with by a separate process that detects surface variants.

Question Answering provides a method of locating precise answers to specific questions but in technical domains the amount of Multi-Word Terms complicates this task. This paper outlines the Question Answering task in such a domain and explores two ways of detecting relations between Multi-Word Terms. The first targets specific semantic relations, the second uses a clustering algorithm, but they are both based on the idea of syntactic variation. The paper demonstrates how the combination of these two methodologies provide sophisticated access to technical domains.