Date

2014

Document Type

Dissertation

Degree

Doctor of Philosophy

Department

Computer Science

First Adviser

Heflin, Jeff

Other advisers/committee members

Hillman, Donald; Davison, Brian; Huang, Wei-Min

Abstract

The amount of Semantic Web data is huge and still keeps growing rapidly today. However most users are still not able to use a Semantic Web Knowledge Base (KB) effectively as desired to due to the lack of various background knowledge. Furthermore, the data is usually heterogeneous, incomplete, and even contains errors, which further impairs understanding the dataset. How to quickly familiarize users with the ontology and data in a KB is an important research challenge to the Semantic Web community.The core part of our proposed resolution to the problem is the contextual tag cloud system: a novel application that helps users explore a large scale RDF(Resource Description Framework) dataset. The tags in our system are ontological terms (classes and properties), and a user can construct a context with a set of tags that defines a subset of instances. Then in the contextual tag cloud, the font size of each tag depends on the number of instances that are associated with that tag and all tags in the context. Each contextual tag cloud serves as a summary of the distribution of relevant data, and by changing the context, the user can quickly gain an understanding of patterns in the data. Furthermore, the user can choose to include different RDFS entailment regimes in the calculations of tag sizes, thereby understanding the impact of semantics on the data. To resolve the key challenge of scalability, we combine a scalable preprocessing approach with a specially-constructed inverted index and co-occurrence matrix, use three approaches to prune unnecessary counts for faster online computations, and design a paging and streaming interface. Via experimentation, we show how much our design choices benefit the responsiveness of our system. We conducted a preliminary user study on this system, and find novice participants felt the system provided a good means to investigate the data and were able to complete assigned tasks more easily than using a baseline interface.We then extend the definition of tags to more general categories, particularly including property values, chaining property values, or functions on these values. With a totally different scenario and more general tags, we find the system can be used to discover interesting value space patterns. To adapt the different dataset, we modify the infrastructure with new indexing data structure, and propose two strategies for online queries, which will be chosen based on different requests, in order to maintain responsiveness of the system.In addition, we consider other approaches to help users locate classes by natural language inputs. Using an external lexicon, Word Sense Disambiguation (WSD) on the label words of classes is one way to understand these classes. We propose our novel WSD approach with our probability model, derive the problem formula into small computable pieces, and propose ways to estimate the values of these pieces. For the other approach, instead of relying on external sources, we investigate how to retrieve query-relevant classes by using the annotations of instances associated with classes in the knowledge base. We propose a general framework of this approach, which consists of two phases: the keyword query is first used to locate relevant instances; then we induce the classes given this list of weighted matched instances.Following the description of the accomplished work, I propose some important future work for extending the current system, and finally conclude the dissertation.

Share

COinS