Michelle Cheatham

String Similarity Metrics for Ontology Alignment

Ontology alignment is an important part of enabling the semantic web to reach its full potential. The vast majority of ontology alignment systems use one or more string similarity metrics, but often the choice of which metrics to use is not given much attention. In this work we evaluate a wide range of such metrics, along with string pre-processing strategies such as removing stop words and considering synonyms, on different types of ontologies. We also present a set of guidelines on when to use which metric. We furthermore show that if optimal string similarity metrics are chosen, those alone can produce alignments that are competitive with the state of the art in ontology alignment systems. Finally, we examine the improvements possible to an existing ontology alignment system using an automated string metric selection strategy based upon the characteristics of the ontologies to be aligned.

Application of Social Network Analysis to Collaborative Team Formation

Team formation is a challenging problem in many large organizations in which it is entirely possible for two individuals to work on similar projects without realizing it. By applying social network analysis to mappings of co-authors and to mappings of related research paper keywords, we are able to help generate teams of diverse individuals with similar interests and aptitudes.

The GeoLink Modular Oceanography Ontology

GeoLink is one of the building block projects within EarthCube, a major effort of the National Science Foundation to establish a next-generation knowledge infrastructure for geosciences. As part of this effort, GeoLink aims to improve data retrieval, reuse, and integration of seven geoscience data repositories through the use of ontologies. In this paper, we report on the GeoLink modular ontology, which consists of an interlinked collection of ontology design patterns engineered as the result of a collaborative modeling effort. We explain our design choices, present selected modeling details, and discuss how data integration can be achieved using the patterns while respecting the existing heterogeneity within the participating repositories.

Conference v2.0: An Uncertain Version of the OAEI Conference Benchmark

The Ontology Alignment Evaluation Initiative is a set of benchmarks for evaluating the performance of ontology alignment systems. In this paper we re-examine the Conference track of the OAEI, with a focus on the degree of agreement between the reference alignments within this track and the opinion of experts. We propose a new version of this benchmark that more closely corresponds to expert opinion and confidence on the matches. The performance of top alignment systems is compared on both versions of the benchmark. Additionally, a general method for crowdsourcing the development of more benchmarks of this type using Amazons Mechanical Turk is introduced and shown to be scalable, cost-effective and to agree well with expert opinion.

Privacy in the age of big data

Lately it seems that every month brings with it a new privacy breach. This can happen due to attackers breaking into a supposedly secure database, such as in the recent cases of Target, Home Depot, and the United States Postal Service, or because data has been released publicly after it has (purportedly) been anonymized. One of the most famous examples of this latter type of privacy compromise is the case of the Netflix Challenge, in which Netflix offered a million dollars to any individual or team who could use an anonymized dataset containing movies and ratings to improve the companys recommendation system by ten percent or more. Just days after the contest began, the anonymization algorithm was broken and the viewing habits of many individuals within the dataset were revealed [1].

AI planning in portal-based workflow management systems

Workflow management systems (WfMS) allow multiple agents to work towards achieving a common goal by facilitating communication between them. This paper discusses the distinctive characteristics of portal-based WfMS and considers the utility of using techniques employed in other WfMS environments in this domain. Specifically, the idea of constructing workflows by applying artificial intelligence planning techniques to a user-specified goal is explored.

The OceanLink project

Todays scientific investigations are producing large numbers of scholarly products. These products continue to increase in diversity and complexity as researchers recognize that scholarly achievements are not only published articles but also datasets, software, and associated supporting materials. OceanLink is an online platform that addresses scholarly discovery and collaboration in the ocean sciences. The OceanLink project leverages Semantic Web technologies, web mining, and crowdsourcing to identify links between data centers, digital repositories, and professional societies to enhance discovery, enable collaboration, and begin to assess research contribution.

Feature and Prototype Evolution for Nearest Neighbor Classification of Web Documents

A nearest neighbor classifier (NNC) approaches the problem of text classification by computing a similarity metric between feature vector representations of an unknown document and a set of known prototype documents. The accuracy and speed of the NNC are dependent upon the choices of features and prototypes. In this paper, we consider the use of a genetic algorithm to optimize the feature and prototype sets for an NNC. We also examine whether simultaneously evolving the feature and prototype sets produces better results than sequential optimization

Special issue on ontology and linked data matching

Semantic web technologies break down many of the barriers to leveraging the large amount of data and information that has been collected or created. The use of unique identifiers, transport protocols like HTTP, and uniform data description languages like RDF go a considerable way towards providing seamless access to this data. Consequently, the semantic web has grown with the continual creation of new ontologies and linked data covering a wide variety of domains, and applications and analytical techniques using this data have been created. However, while physical data silos have waned, the lack of semantic links between ontologies and linked datasets, supports, in effect, invisible virtual silos preventing these resources from being queried, browsed, or leveraged in a truly uniform way. If such links could be generated in a reliable and scalable way, the network effect would greatly increase the utility of these resources. It is for this reason that the topic of ontology and linked data matching is both important and timely. Ontology and linked data matching has been an active area of research for over a decade now [4],

Metrics for trust in layered sensing

In this paper we discuss metrics for trust in layered sensing from the perspective of a military commander or intelligence analyst. Specifically, the views here are based on the operational perspective and experiences of an intelligence analyst who has worked in NATO, Joint Forces, and Army-specific combat environments. Our goal is to create a shared understanding among the researchers working to develop the theoretical framework and technologies needed for trusted sensing and the military personnel who will ultimately leverage the results of those efforts. Our hypothesis is that trust in the layered sensing environment can be assigned a level of confidence based on a combination of two different metrics: 1) quantifiable aspects of the physical organization and communications rates and patterns of a sensor network and 2) qualitative aspects of the environment of the source.