Structured data access annotations for massively parallel computations

We describe an approach aimed at addressing the issue of joint exploitation of control (stream) and data parallelism in a skeleton based parallel programming environment, based on annotations and refactoring. Annotations drive efficient implementation of a parallel computation. Refactoring is used to transform the associated skeleton tree into a more efficient, functionally equivalent skeleton tree. In most cases, cost models are used to drive the refactoring process. We show how sample use case applications/kernels may be optimized and discuss preliminary experiments with FastFlow assessing the theoretical results. © 2013 Springer-Verlag.

Kernel Methods for Tree Structured Data

Machine learning comprises a series of techniques for automatic extraction of meaningful information from large collections of noisy data. In many real world applications, data is naturally represented in structured form. Since traditional methods in machine learning deal with vectorial information, they require an a priori form of preprocessing. Among all the learning techniques for dealing with structured data, kernel methods are recognized to have a strong theoretical background and to be effective approaches. They do not require an explicit vectorial representation of the data in terms of features, but rely on a measure of similarity between any pair of objects of a domain, the kernel function. Designing fast and good kernel functions is a challenging problem. In the case of tree structured data two issues become relevant: kernel for trees should not be sparse and should be fast to compute. The sparsity problem arises when, given a dataset and a kernel function, most structures of the dataset are completely dissimilar to one another. In those cases the classifier has too few information for making correct predictions on unseen data. In fact, it tends to produce a discriminating function behaving as the nearest neighbour rule. Sparsity is likely to arise for some standard tree kernel functions, such as the subtree and subset tree kernel, when they are applied to datasets with node labels belonging to a large domain. A second drawback of using tree kernels is the time complexity required both in learning and classification phases. Such a complexity can sometimes prevents the kernel application in scenarios involving large amount of data. This thesis proposes three contributions for resolving the above issues of kernel for trees. A first contribution aims at creating kernel functions which adapt to the statistical properties of the dataset, thus reducing its sparsity with respect to traditional tree kernel functions. Specifically, we propose to encode the input trees by an algorithm able to project the data onto a lower dimensional space with the property that similar structures are mapped similarly. By building kernel functions on the lower dimensional representation, we are able to perform inexact matchings between different inputs in the original space. A second contribution is the proposal of a novel kernel function based on the convolution kernel framework. Convolution kernel measures the similarity of two objects in terms of the similarities of their subparts. Most convolution kernels are based on counting the number of shared substructures, partially discarding information about their position in the original structure. The kernel function we propose is, instead, especially focused on this aspect. A third contribution is devoted at reducing the computational burden related to the calculation of a kernel function between a tree and a forest of trees, which is a typical operation in the classification phase and, for some algorithms, also in the learning phase. We propose a general methodology applicable to convolution kernels. Moreover, we show an instantiation of our technique when kernels such as the subtree and subset tree kernels are employed. In those cases, Direct Acyclic Graphs can be used to compactly represent shared substructures in different trees, thus reducing the computational burden and storage requirements.

Integrating structured data using property precedence

Data integration systems offer uniform access to a set of autonomous and heterogeneous data sources. One of the main challenges in data integration is reconciling semantic differences among data sources. Approaches that been used to solve this problem can be categorized as schema-based and attribute-based. Schema-based approaches use schema information to identify the semantic similarity in data; furthermore, they focus on reconciling types before reconciling attributes. In contrast, attribute-based approaches use statistical and structural information of attributes to identify the semantic similarity of data in different sources. This research examines an approach to semantic reconciliation based on integrating properties expressed at different levels of abstraction or granularity using the concept of property precedence. Property precedence reconciles the meaning of attributes by identifying similarities between attributes based on what these attributes represent in the real world. In order to use property precedence for semantic integration, we need to identify the precedence of attributes within and across data sources. The goal of this research is to develop and evaluate a method and algorithms that will identify precedence relations among attributes and build property precedence graph (PPG) that can be used to support integration.

The MUSE Machine -- an Architecture for Structured Data Flow Computation

Computers employing some degree of data flow organisation are now well established as providing a possible vehicle for concurrent computation. Although data-driven computation frees the architecture from the constraints of the single program counter, processor and global memory, inherent in the classic von Neumann computer, there can still be problems with the unconstrained generation of fresh result tokens if a pure data flow approach is adopted. The advantages of allowing serial processing for those parts of a program which are inherently serial, and of permitting a demand-driven, as well as data-driven, mode of operation are identified and described. The MUSE machine described here is a structured architecture supporting both serial and parallel processing which allows the abstract structure of a program to be mapped onto the machine in a logical way.

Exploiting Structured Linked Data in Enterprise Knowledge Management Systems: An Idea Management Case Study

In parallel to the effort of creating Open Linked Data for the World Wide Web there is a number of projects aimed for developing the same technologies but in the context of their usage in closed environments such as private enterprises. In the paper, we present results of research on interlinking structured data for use in Idea Management Systems - a still rare breed of knowledge management systems dedicated to innovation management. In our study, we show the process of extending an ontology that initially covers only the Idea Management System structure towards the concept of linking with distributed enterprise data and public data using Semantic Web technologies. Furthermore we point out how the established links can help to solve the key problems of contemporary Idea Management Systems

Discovery driven analysis on semi-structured text data

Discovery Driven Analysis (DDA) is a common feature of OLAP technology to analyze structured data. In essence, DDA helps analysts to discover anomalous data by highlighting 'unexpected' values in the OLAP cube. By giving indications to the analyst on what dimensions to explore, DDA speeds up the process of discovering anomalies and their causes. However, Discovery Driven Analysis (and OLAP in general) is only applicable on structured data, such as records in databases. We propose a system to extend DDA technology to semi-structured text documents, that is, text documents with a few structured data. Our system pipeline consists of two stages: first, the text part of each document is structured around user specified dimensions, using semi-PLSA algorithm; then, we adapt DDA to these fully structured documents, thus enabling DDA on text documents. We present some applications of this system in OLAP analysis and show how scalability issues are solved. Results show that our system can handle reasonable datasets of documents, in real time, without any need for pre-computation.

A novel method for finding similarities between unordered trees using matrix data model

Trees are capable of portraying the semi-structured data which is common in web domain. Finding similarities between trees is mandatory for several applications that deal with semi-structured data. Existing similarity methods examine a pair of trees by comparing through nodes and paths of two trees, and find the similarity between them. However, these methods provide unfavorable results for unordered tree data and result in yielding NP-hard or MAX-SNP hard complexity. In this paper, we present a novel method that encodes a tree with an optimal traversing approach first, and then, utilizes it to model the tree with its equivalent matrix representation for finding similarity between unordered trees efficiently. Empirical analysis shows that the proposed method is able to achieve high accuracy even on the large data sets.

Management and Analysis of Unstructured Construction Data Types

Compared with structured data sources that are usually stored and analyzed in spreadsheets, relational databases, and single data tables, unstructured construction data sources such as text documents, site images, web pages, and project schedules have been less intensively studied due to additional challenges in data preparation, representation, and analysis. In this paper, our vision for data management and mining addressing such challenges are presented, together with related research results from previous work, as well as our recent developments of data mining on text-based, web-based, image-based, and network-based construction databases.

Visual Segmentation-Based Data Record Extraction From Web Documents

Li, Longzhuang, Liu, Yonghuai, Obregon, A., Weatherston, M. Visual Segmentation-Based Data Record Extraction From Web Documents. Proceedings of IEEE International Conference on Information Reuse and Integration, 2007, pp. 502-507. Sponsorship: IEEE

Quantifying data quality for clinical trials using electronic data capture.

BACKGROUND: Historically, only partial assessments of data quality have been performed in clinical trials, for which the most common method of measuring database error rates has been to compare the case report form (CRF) to database entries and count discrepancies. Importantly, errors arising from medical record abstraction and transcription are rarely evaluated as part of such quality assessments. Electronic Data Capture (EDC) technology has had a further impact, as paper CRFs typically leveraged for quality measurement are not used in EDC processes. METHODS AND PRINCIPAL FINDINGS: The National Institute on Drug Abuse Treatment Clinical Trials Network has developed, implemented, and evaluated methodology for holistically assessing data quality on EDC trials. We characterize the average source-to-database error rate (14.3 errors per 10,000 fields) for the first year of use of the new evaluation method. This error rate was significantly lower than the average of published error rates for source-to-database audits, and was similar to CRF-to-database error rates reported in the published literature. We attribute this largely to an absence of medical record abstraction on the trials we examined, and to an outpatient setting characterized by less acute patient conditions. CONCLUSIONS: Historically, medical record abstraction is the most significant source of error by an order of magnitude, and should be measured and managed during the course of clinical trials. Source-to-database error rates are highly dependent on the amount of structured data collection in the clinical setting and on the complexity of the medical record, dependencies that should be considered when developing data quality benchmarks.

Web Data Extraction from Query Result Pages Based on Visual and Content Features

A rapidly increasing number of Web databases are now become accessible via
their HTML form-based query interfaces. Query result pages are dynamically generated
in response to user queries, which encode structured data and are displayed for human
use. Query result pages usually contain other types of information in addition to query
results, e.g., advertisements, navigation bar etc. The problem of extracting structured data
from query result pages is critical for web data integration applications, such as comparison
shopping, meta-search engines etc, and has been intensively studied. A number of approaches
have been proposed. As the structures of Web pages become more and more complex, the
existing approaches start to fail, and most of them do not remove irrelevant contents which
may a®ect the accuracy of data record extraction. We propose an automated approach for
Web data extraction. First, it makes use of visual features and query terms to identify data
sections and extracts data records in these sections. We also represent several content and
visual features of visual blocks in a data section, and use them to ¯lter out noisy blocks.
Second, it measures similarity between data items in di®erent data records based on their
visual and content features, and aligns them into di®erent groups so that the data in the
same group have the same semantics. The results of our experiments with a large set of
Web query result pages in di®erent domains show that our proposed approaches are highly
e®ective.

Visually extracting data records from the deep web

Web sites that rely on databases for their content are now ubiquitous. Query result pages are dynamically generated from these databases in response to user-submitted queries. Automatically extracting structured data from query result pages is a challenging problem, as the structure of the data is not explicitly represented. While humans have shown good intuition in visually understanding data records on a query result page as displayed by a web browser, no existing approach to data record extraction has made full use of this intuition. We propose a novel approach, in which we make use of the common sources of evidence that humans use to understand data records on a displayed query result page. These include structural regularity, and visual and content similarity between data records displayed on a query result page. Based on these observations we propose new techniques that can identify each data record individually, while ignoring noise items, such as navigation bars and adverts. We have implemented these techniques in a software prototype, rExtractor, and tested it using two datasets. Our experimental results show that our approach achieves significantly higher accuracy than previous approaches. Furthermore, it establishes the case for use of vision-based algorithms in the context of data extraction from web sites.

Autonomic scheduling of tasks from data parallel patterns to CPU/GPU core mixes

We propose a methodology for optimizing the execution of data parallel (sub-)tasks on CPU and GPU cores of the same heterogeneous architecture. The methodology is based on two main components: i) an analytical performance model for scheduling tasks among CPU and GPU cores, such that the global execution time of the overall data parallel pattern is optimized; and ii) an autonomic module which uses the analytical performance model to implement the data parallel computations in a completely autonomic way, requiring no programmer intervention to optimize the computation across CPU and GPU cores. The analytical performance model uses a small set of simple parameters to devise a partitioning-between CPU and GPU cores-of the tasks derived from structured data parallel patterns/algorithmic skeletons. The model takes into account both hardware related and application dependent parameters. It computes the percentage of tasks to be executed on CPU and GPU cores such that both kinds of cores are exploited and performance figures are optimized. The autonomic module, implemented in FastFlow, executes a generic map (reduce) data parallel pattern scheduling part of the tasks to the GPU and part to CPU cores so as to achieve optimal execution time. Experimental results on state-of-the-art CPU/GPU architectures are shown that assess both performance model properties and autonomic module effectiveness. © 2013 IEEE.

Exploiting evidence from unstructured data to enhance master data management

Master data management (MDM) integrates data from multiple
structured data sources and builds a consolidated 360-
degree view of business entities such as customers and products.
Today’s MDM systems are not prepared to integrate
information from unstructured data sources, such as news
reports, emails, call-center transcripts, and chat logs. However,
those unstructured data sources may contain valuable
information about the same entities known to MDM from
the structured data sources. Integrating information from
unstructured data into MDM is challenging as textual references
to existing MDM entities are often incomplete and
imprecise and the additional entity information extracted
from text should not impact the trustworthiness of MDM
data.
In this paper, we present an architecture for making MDM
text-aware and showcase its implementation as IBM InfoSphere
MDM Extension for Unstructured Text Correlation,
an add-on to IBM InfoSphere Master Data Management
Standard Edition. We highlight how MDM benefits from
additional evidence found in documents when doing entity
resolution and relationship discovery. We experimentally
demonstrate the feasibility of integrating information from
unstructured data sources into MDM.