Marcello Leida

Case Study in Process Mining in a Multinational Enterprise

Process mining has become an active area of research and while there are numerous papers on approaches to process mining there are fewer detailing its application to real industrial scenarios and its applicability in these spaces. In this paper we introduce the approach to process mining used in a number of multinational enterprises and then reflect upon the issues that have been encountered during our ongoing work. In our opinion these issues are a clear example of the challenges that need to be addressed during business process discovery from heterogeneous data.

A Lightweight RDF Data Model for Business Process Analysis

This article presents a lightweight data representation model designed to support real time monitoring of business processes. The model is based on a shared vocabulary defined using open standard representations (RDF) allowing independence and extremely flexible interoperability between applications. The main benefit of this representation is that it is transparent to the data creation and analysis processes; furthermore it can be extended progressively when new information is available. Business Process data represented with this data model can be easily published on-line and shared between applications. After the definition of the data model, in this article, we demonstrate that with the use of this representation it is possible to retrieve and make use of domain specific information without any previous knowledge of the process. This model is a novel approach to real-time process data representation and paves the road to a complete new breed of applications for business process analysis.

Distributed SPARQL Query Answering over RDF Data Streams

The RDF framework is the underpinning element of Semantic Web stack, its widespread adoption requires efficient tools to store and query RDF data. A number of efficient local RDF stores already exist, while distributed indexing and distributed query processing are only starting to develop, furthermore dynamically growing and fail-safe solutions are not yet available. To remedy this situation, we propose an approach for efficient and scalable query processing over RDF graphs, distributed over a local data grid. Our system is based on a distributed architecture, where neither single point of failure nor specialised nodes exist. The query processing framework, presented in the paper, includes a sophisticated query planning and query execution algorithm, which is designed expressively for storage and query of a stream of incoming RDF triples, allowing the users to register queries that will be notified in real time of new relevant data. We finally evaluate our approach through performance measurement of a real deployment in the areas of business process monitoring.

Semantics-aware data integration for heterogeneous data sources

This article presents a novel definition of a declarative mapping language, which is able to map precisely and unambiguously the semantics of a domain conceptualization (defined as an ontology) into queries to a set of data sources, where the data is residing. In this way, a system making use of this mapping language is able to access the data actually stored in the data sources thought a semantically rich representation. The mapping model proposed in this paper is also an ontology and therefore is machine understandable: it can be shared with other users or systems, processed by external tools for consistency checking, or collaboratively created and so on. Besides the contributions of the mapping model itself, this paper introduces the concepts of Semantic Join and Semantic Identifiers: a declarative approach to semantic data fusion and entity resolution over multiple unrelated databases, which allow to define extremely expressive mapping.

Extendible data model for real-time business process analysis

This paper presents a promising data representation model for real time monitoring of business processes. The main benefit of this representation is that is transparent to the data creation and analysis processes and it is extendible at realtime. The model is based on a shared vocabulary defined using RDF standard representation allowing independence between applications. This model is a novel approach to real-time process data representation and paves the road to a complete new breed of applications for business process analysis.

Outlines for dynamic visualization of semantic web data

Ontologies, the data model underpinning the Semantic Web vision, are nowadays widely used to represent data from different origins and of diverse nature, supported by a plethora of tools for the storage and querying of ontologies. However, in the area of information visualization there is still much to be achieved. Several strategies and implementations have appeared over the past years as a result of the increase in publicly available data modelled as Resource Description Framework (RDF). The problem of representing this data using charts, dashboards, maps and so on has become pressing, in particular to prove the value of the Semantic Web to enhance the analysis of business data. In this paper we describe the problem of ontology visualization by presenting the major approaches, focusing on the problems that prevent these approaches from providing an automatic, dynamic, generic and flexible ontology visualization tool for analytical purposes.

Semantics-aware matching strategy (SAMS) for the Ontology meDiated Data Integration (ODDI)

Data integration systems are used to integrate heterogeneous data sources in a single view. Recent work on business intelligence highlights the need of on-time, reliable and sound data access systems relying on methods based on semi-automatic procedures. A crucial factor for any semi-automatic algorithm is that of the matching strategy. Different categories of matching operators carry different semantics. For this reason, combining them into a single strategy is a non-trivial process that has to take into account a variety of options. This paper presents SAMS, a matching strategy based on a semantics-aware categorisation of matching operators that allows to group similar attributes on a semantically-rich form.

A Knowledge Base Visual Analytics Technique for Semantic Web

One of the technologies underpinning the future vision of the Web as huge database is Linked Data (LD). LD provides structured data over the Web that is understandable by machines. Therefore, it enables smart queries between different datasets over the Web. Consequently, effective Visual Analytics (VAs) techniques become necessary to efficiently extract and visualize the desired information from these data graphs. In this paper, we propose a theoretical VAs framework for LD as an approach for the automatic suggestion of information visualization graphs. The key objective of the framework is to amplify user perception by suggesting the best visual representation such as bar chart, map, and timeline of the selected data. The core of the framework is a well-defined artificial knowledge base information visualization ontology for automating the visualization process. The framework helps in the analysis process of data by providing the best visual representation to produce accurate decisions and to pave the road for next generation of VAs tools for the semantic web data.

Zeus: a distributed triple store for proactive SPARQL queries

The RDF framework is the underpinning element of semantic web stack; its widespread adoption requires efficient tools to store and query RDF data. A number of efficient local RDF stores already exist, while distributed indexing and distributed query processing are only starting to appear; furthermore dynamically growing and fail-safe solutions are not yet available. To remedy this situation, in this paper we present Zeus: an approach for efficient and scalable query processing over RDF graphs based on grid computing. Zeus is based on a distributed architecture, where neither single point of failure nor specialised nodes exist. The query processing framework, presented in the paper, includes a sophisticated query planning and query execution algorithm, which is designed expressively for storage and query of a stream of RDF triples. This mechanism allows Zeus users to register queries that will be notified in real time of new relevant results.

Visual Analytics in the web of data

Nowadays the web is composed of huge, complex, and heterogeneous datasets. Making sense of this huge amount of data is challenging. Two emerging technologies developed to solve this issue namely, Linked Data, and Visual Analytics. Linked Data provides structured data over the web. Structured data are understandable by machines therefore smart queries, automated visualization, and linked between different datasets over the web might be possible. On the other hand, Visual Analytics amplify analyst cognitive ability by providing data analysis process enhanced with interactive visualization. In our project, we are aiming to integrate these two technologies to provide a powerful Visual Analytics System.