Igor O. Popov

Trust me, i'm partially right: incremental visualization lets analysts explore large datasets faster

Queries over large scale (petabyte) data bases often mean waiting overnight for a result to come back. Scale costs time. Such time also means that potential avenues of exploration are ignored because the costs are perceived to be too high to run or even propose them. With sampleAction we have explored whether interaction techniques to present query results running over only incremental samples can be presented as sufficiently trustworthy for analysts both to make closer to real time decisions about their queries and to be more exploratory in their questions of the data. Our work with three teams of analysts suggests that we can indeed accelerate and open up the query process with such incremental visualizations.

Connecting the dots: a multi-pivot approach to data exploration

The purpose of data browsers is to help users identify and query data effectively without being overwhelmed by large complex graphs of data. A proposed solution to identify and query data in graph-based datasets is Pivoting (or set-oriented browsing), a many-to-many graph browsing technique that allows users to navigate the graph by starting from a set of instances followed by navigation through common links. Relying solely on navigation, however, makes it difficult for users to find paths or even see if the element of interest is in the graph when the points of interest may be many vertices apart. Further challenges include finding paths which require combinations of forward and backward links in order to make the necessary connections which further adds to the complexity of pivoting. In order to mitigate the effects of these problems and enhance the strengths of pivoting we present a multi-pivot approach which we embodied in tool called Visor. Visor allows users to explore from multiple points in the graph, helping users connect key points of interest in the graph on the conceptual level, visually occluding the remainder parts of the graph, thus helping create a road-map for navigation. We carried out an user study to demonstrate the viability of our approach.

Put in your postcode, out comes the data: a case study

Article describes the UK Open Government Data project which the two authors have been leading and the planned launch of data.gov.uk a single point of access for all public non-personal government datasets. It outlines the benefits that will flow from more accessible and open data. The article first appeared in The Times 18th Nov 2009 http://www.thetimes.co.uk/tto/law/columnists/article2049543.ece

ONTOCOM Revisited: Towards Accurate Cost Predictions for Ontology Development Projects

Reliable methods to assess the costs and benefits of ontologies are an important instrument to demonstrate the tangible business value of semantic technologies within enterprises, as an argument to encourage their wide-scale adoption. The economic aspects of ontologies have been investigated in previous work of ours. With ONTOCOM we proposed a cost estimation model for ontologies and ontology development projects. This paper revisits this model and presents its latest achievements. We report on a comprehensive calibration of ONTOCOM based on a considerably larger data set of 148 ontology development projects. The calibration used a combination of statistical methods, ranging from preliminary data analysis to regression and Bayes analysis, and resulted a significant improvement of the prediction quality of up to 50%. In addition, the availability of a representative data set allowed us to identify meaningful directions for customizing the generic cost model along particular types of ontologies, and ontology-like structures as those specific to the emerging Web 3.0. Last but not least, we developed a software tool that allows ontology development project managers to easily use and adapt and to systematically calibrate the model, thus facilitating its adoption in real-world projects.

ONTOCOM: A reliable cost estimation method for ontology development projects

We present ONTOCOM, a method to estimate the costs of ontology engineering, as well as project management tools that support the application of the method. ONTOCOM is part of a broader framework we have developed over the five years, whose aim is to assess the business value of semantic technologies through a suite of methods, estimation models and project management tools, by which the costs and benefits of the corresponding projects are defined, measured and analyzed. The framework supports the engineering of different types of knowledge structures, including ontologies, taxonomies and folksonomies, and of information management systems leveraging such knowledge structures. It also includes benefit analysis models whose results can be used in conjunction with cost-related information in order to identify potential cost savings and to assess the feasibility of specific engineering strategies, in particular ontology reuse. The application of the methods proposed in the framework is supported by project management tools which can be used to customize these methods to a given project environment, to evaluate and validate the underlying estimations using empirical data, and to take into account their results for planning and controlling purposes.

Achieving Maturity: The State of Practice in Ontology Engineering in 2009

In this paper we give an account of the current state of practice in ontology engineering (OE) based on the findings of a 6 months empirical survey that analyzed 148 OE projects. The survey focused on process-related issues and looked into the impact of research achievements on real-world OE projects, the complexity of particular ontology development tasks, the level of tool support, and the usage scenarios for ontologies. The main contributions of this survey are twofold: 1) the size of the data set is larger than every other similar endeavor; 2) the findings of the survey confirm that OE is an established engineering discipline w.r.t the maturity and level of acceptance of its main components, methodologies, etc. whereas further research should target economic aspects of OE and the customization of existing technology to the specifics of vertical domains.

Framework for Value Prediction of Knowledge-Based Applications

Knowledge-based applications are characterized by their use of machine-understandable formalizations of expert knowledge. Complex knowledge structures, and the features which exploit them, can have a significant effect on the effort needed to develop such applications. Means to estimate this effort are, however, lacking. Furthermore, precise benefits of such applications, which are directly attributed to specific functionalities, remain unknown.

mashpoint: supporting data-centric navigation on the web

Large numbers of Web sites support rich data-centric features to explore and interact with data en masse. For example, online shopping or travel sites routinely offer users ways to filter, explore and interact with their data. The ability to find related information on other Web sites about selected subsets of the data, however, is currently compromised by the Webs inherent, one document at a time, navigation. In this paper we present work-in-progress on mashpoint, a framework that allows distributed data-powered Web applications to exchange subsets of their data, in effect enabling many-to-many navigation on the Web, at a more granular, data level. We hypothesise that allowing such navigation unlocks novel possibilities for information exploration and interaction on the Web. We present an initial prototype and discuss the opportunities and challenges of facilitating this kind of interaction.

mashpoint: Surfing the web in a data-oriented way

Simple information lookup tasks (e.g. “What the weather like in London?” or “What is the population of the UK?”), are currently well supported with traditional search engines, and more recently with intelligent personal assistants. Intensive knowledge tasks, (e.g. “How do countries with low GDP per capita rank in emigration?”), however, require combining and cross referencing data from multiple sources to get to an answer have typically not been well supported. Our ability to support these types of information tasks on the Web is currently compromised by the inherent document/application nature of the Web itself. End-user mashup tools traditionally approach this problem by assisting users in structuring unstructured content form web pages and then support information-oriented tasks over the structured content. Motivated by the fact that more and more structured data is available on Web pages we investigate another possible solution: how to extend traditional Web navigation, which the majority of end users find intuitive, to include more data-centric behaviour. With mashpoint we propose a simple architecture, which would support an interaction that allows web pages to be linked based on similarities of the entities in their data. Linked in this way, queries that traditionally require the tedious work of joining information form several pages can be performed with simple web-like navigation. The paper focuses on evaluating if the proposed interaction is one that users would be able to understand to execute intensive knowledge tasks. We ran two separate studies: first to explore if the interaction concepts introduced are easily learnable and to gather initial feedback on our prototype, and second to explore design options which can inform how to address discovery challenges when large amount of pages are linked in this way, therefore assessing the feasibility of this model to work on a Web-scale.

mashpoint: browsing the web along structured lines

Large numbers of Web sites support rich data-centric features to explore and interact with data. In this paper we present mashpoint, a framework that allows distributed data-powered Web applications to linked based on similarities of the entities in their data. By linking applications in this way we allow browsing with selections of data from one application to another application. This sort of browsing allows complex queries and exploration of data to be done by average Web users using multiple applications. We additionally use this concept to surface structured information to users in Web pages. In this paper we present this concept and our initial prototype.