Mark Sifer

User Interfaces for the Exploration of Hierarchical Multi-dimensional Data

A variety of user interfaces have been developed to support the querying of hierarchical multi-dimensional data in an OLAP setting such as pivot tables and Polaris. They are used to regularly check portions of a dataset and to explore a new dataset for the first time. In this paper, we establish criteria for OLAP user interface capabilities to facilitate comparison. Two criteria are the number of displayed dimensions along which comparisons can be made and the number of dimensions that are viewable at once - visual comparison depth and width. We argue that interfaces with greater visual comparison depth support regular checking of known data by users that know roughly where to look, while interfaces with greater comparison width support exploration of new data by users that have no a priori starting point and need to scan all dimensions. Pivot tables and Polaris are examples of the former. The main contribution of this paper is to introduce a new scalable interface that uses parallel dimension axis which supports the latter, greater visual comparison width. We compare our approach to both table based and parallel coordinate based interfaces. We present an implementation of our interface SGViewer, user scenarios and provide an evaluation that supports the usability of our interface

Filter co-ordinations for exploring multi-dimensional data

Many interface designs have been developed for the exploration of multi-dimensional data sets which are based on finding subsets by filtering attribute values. Systems such as dynamic queries use a collection of independent filters to interactively query by restricting attribute values. However, for large data sets there is a need for an alternative style of filtering that better supports stepwise query refinement. This article introduces a new filter coordination which supports both stepwise query refinement and independent filters. Our filter visualization also supports the visualization of attribute value hierarchies enabling multi-level data distribution overviews to be given. Our coordination design is implemented in our SGViewer query tool which we demonstrate with a multi-dimensional web log data set. An evaluation of SGViewer showed that after a short learning period users were able to use it to read trends and proportions and make drill-down queries.

Refining search results with facet landscapes

1. OVERVIEW Text search systems typically return lists of results while the corpus remains invisible in the background. How results relate to each other and to the corpus as a whole is not shown. Faceted search improves upon this by presenting categorised results in predefined facets. The distribution of results across categories is shown in order to guide query refinement through category selection. Faceted search systems may be text based [1] or use an interactive visualisation [4]. They show the relationship between results but not their relationship to the corpus as a whole. In this demonstration we present a new approach that combines faceted search with corpus landscapes, to reveal search result clusters and their distribution across a whole corpus. We demonstrate with a portion of a local file system. Our interface shows multiple landscapes, one for each facet, and a keyword search panel. A search starts by entering a term such as OLAP. A list of found files and facet landscapes decorated with file locations are shown. The facet landscapes are proportion tree views of the corpus where category width is proportional to the number of contained files and leaves represent buckets of files. The result locations are the leaf positions that have been highlighted with a magenta border. Location visibility can be enhanced by a distortion that increases width. A query is refined by selecting facet categories. Figure 1 shows our interface after 2006 and 2007 in the date facet were selected, restricting the other facets and the results list to this subset. Additional category selections in the date and other facets would further refine this. Our prototype uses the open source search engine lucene to create a word index and an OLAP data preparation tool MakeSGF to create a facet index. MakeSGF loads a scan of a file system and combines it with derived and predefined facet hierarchies to generate a sparse data-cube. It also generates the cell to file mapping needed for coordination of the SGViewer tool and result list. SGViewer started as a sitemap [2], but was extended to support querying over hierarchical multi-dimensional data [3]. Files with the same facet values are aggregated into one datacube cell. This improves scalability but can decrease the resolution of the facet landscapes. If our system was split into a client and server, such decreased resolution can be an advantage for applications that need to limit disclosure of detail.

Integrating keyword search with multiple dimension tree views over a summary corpus data cube

We demonstrate a system that integrates a novel OLAP component with a keyword search engine, to support querying over sparse and ragged corpus data. The key contribution of our system is the integration of dynamically selected point sets such as search results with OLAP querying over aggregated data. During the demonstration, participants will be able to enter a keyword search; observe the returned list of result files; observe distributional features such as outliers and clusters of results in the corpus in multiple dimension views; and select and partition corpus slices in the OLAP component to narrow search results. Participants will be able to experience not just the individual querying features of our system, but the way that they work together to facilitate smooth interaction sequences that combine OLAP and keyword search querying.

Efficient Visualisation of the Relative Distribution of Keyword Search Results in a Corpus Data Cube

Most keyword searches target precision for finding the most relevant document. However some target recall, finding all relevant documents. Our system supports high recall searches that return hundreds or thousands of relevant results. In particular, it provides a visualization that shows the distribution of search results relative to the distribution of items for the entire corpus. Such relative distributional features include over and under representation, clusters and outliers. The contribution of this paper is efficient visualisation, that is, how to provide the best relative distribution view for a given data cube size. This requirement is translated to: for which limited size meta-data summary cube are search results disambiguated the most in our relative distribution view. We identify metrics and several algorithms for such a summary cube selection.

Exploring hierarchical multidimensional data with unified views of distribution and correlation

Data analysts explore data by inspecting features such as clustering, distribution and correlation. Much existing research has focused on different visualisations for different data exploration tasks. For example, a data analyst might inspect clustering and correlation with scatterplots, but use histograms to inspect a distribution. Such visualisations allow an analyst to confirm prior expectations. For example, a scatterplot may confirm an expected correlation or may show deviations from the expected correlation. In order to better facilitate discovery of unexpected features in data, however, a combination of different perspectives may be needed. In this paper, we combine distributional and correlational views of hierarchical multidimensional data. Our unified view supports the simultaneous exploration of data distribution and correlation. By presenting a unified view, we aim to increase the chances of discovery of unexpected data features, and to provide the means to explore such features in detail. Further, our unified view is equipped with a small number of primitive interaction operators which a user composes to facilitate smooth and flexible exploration. Highlights? A visualisation that shows both distribution and correlation is presented. ? Interactive features such as filtering, colouring, hashing, rotating and zooming are described. ? Builds on a parallel tree visualisation for exploring multi-dimensional data. ? Includes support for dimension selection via dimension ranking.

Exploring website log data with coordinated dimension hierarchies and network pivots

Many interfaces exist for exploring multi-dimensional data, but most are not suitable for OLAP as they do not support aggregation and selection via dimension hierarchies. We introduce an interface for exploring OLAP data via coordinated dimension hierarchies and network views. Three dimension coordinations are identified; progressive, global and result only, for filtering and presenting data. We integrate these with a network pivot that supports joins through an arbitrary relation. We demonstrate with an example web log dataset of site visits organised into time, downloads, visitor address and referrer address dimensions and a network of prior and later visits.

SCALABILITY FOR GRAPH-BASED CASE TOOLS

Software engineering is particularly concerned with the construction of large systems. Existing software engineering tools tend to be adequate for medium-sized systems, but not as useful for large and very large systems. This paper presents a model viewing system as the basis for graph-based CASE tools which overcomes this lack of scalability. With existing commercial tools the number of user steps to browse or edit a model increases with the size of the model. With the approach of this paper, the number of steps remains constant regardless of the model size. In fact, only one step is required for operations such as adding a flow between two processes anywhere in the model, or moving a submodel to a new parent. This paper outlines the approach with a structured analysis example, provides a formal description of the model viewing system, and discusses some limitations.

Bottom-up Faceted Search: Creating Search Neighbourhoods with Datacube Cells

Browsing a collection can start with a keyword search. A user visits a library, performs a keyword search to find a few books of interest; finding their location in the library. Then they go to these locations; the corresponding bookshelves, where they do not just retrieve the found books, but rather they start browsing the nearby books; the books which have a similar Dewey classification. This paper extends this approach to curated corpora that contain items or documents that have been classified in multiple dimensions (facets), where each dimension classification may be a hierarchy. In particular (i) a technique for determining near items based on OLAP datacube cells and (ii) user interfaces that support browsing of near items are presented.

Exploring OLAP data with parallel dimension views

Existing OLAP user interfaces typically explore hierarchical multi-dimensional data through tabular data cube views. Aggregation is supported by dimension hierarchy level selection and filtering by slice and dice operations. Aggregation determines the size of data cube cells while filtering determines the cells in the view. Table based interfaces provide views that typically include two or three dimensions at a chosen level of aggregation. This paper describes an interface that is based on an alternative paradigm, parallel coordinates. However, instead of parallel axis, we use parallel dimension trees. The interface supports data aggregation and filtering operations. It supports both proportional and fixed value dimension scales. It supports a range of exploration tasks including viewing data distribution, comparing data distributions and viewing correlation. The main benefit of our interface is its support for rapid and flexible overviews across many dimensions and multiple hierarchy levels at the cost of less detailed views.