eXVisXML , uma ferramenta emblemática na análise documental

EXVISXML, UMA FERRAMENTA EMBLEMÁTICA NA ANÁLISE DOCUMENTAL

Daniela da Cruz, Pedro Rangel Henriques

Departamento de InformáticaUniversidade do Minho

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro

CONTEXT 6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro

MOTIVATION 6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro

MOTIVATION 6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro

MOTIVATION 6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro

MOTIVATION 6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro

MOTIVATION 6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro

XML DOCUMENT VISUALIZATION

The role of the visualization technology (in PC and SE) is recognized as very fruitful.

The use of SV features allows us to capture a great amount of information in a faster way

Graphical representations cause a positive impact in learning process

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro


Retrieve information from plain documents efficiently IS NOT AN EASY TASK

Machine manipulation: XSL and other production-systems can easily

extract information and transform them

Human manipulation: It is not as easy as desirable The annotation is complex / Document is too big

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro


Many tools appear to aid in the visualization of XML documents: XML Schema Designer (Microsoft) Xpath Analyzer (Altova) …

Although these tools offer highlighted syntax, and easy manipulation (collapse/expand), their view is a hierarchical and textual.

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro

TRADITIONAL XML DOCUMENT VISUALIZATION 6

Março

, 20

09

Univ

ersid

ade d

e A

veiro

OUR PROPOSAL FOR XML DOCUMENT VISUALIZATION

In this context, we want to get a visualization that makes easier the comprehension process.

However, we should take care with the graphical or iconic representations hence it depends on problem domain.

Inspired in Alma, the eXVisXML interface for the visual inspection of XML documents is divided into 3 main parts:

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro

OUR PROPOSAL FOR XML DOCUMENT VISUALIZATION

One window that displays the source document;

One window exhibiting the textual hierarchy

One window to show the tree associated with the source document (graphical);

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro

OUR PROPOSAL FOR XML DOCUMENT VISUALIZATION 6

Março

, 20

09

Univ

ersid

ade d

e A

veiro

XML DOCUMENT SLICING

Slicing concept appears in 1979, by Weiser.

Its applied to a program considering a slicing criterion (a pair composed by a line number and a set of variables).

The objective is to find the statements that possibly affect those variables.

This technique can be also applied to XML documents. How?

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro


XML document + slicing criterion (a Xpath expression can be regarded as a slicing criterion, but simplified)

A document slice is a new XML document composed by those elements that are strictly necessary to maintain the tree structure.

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro


It is proved, by Josep Silva, in Slicing XML documents, that slicing techniques applied to XML and DTD documents produce valid XML and DTD slices with the respect to the slicing criterion.

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro


Given the whole XML document of Romeo and Juliet screenplay

and

The slicing criterion Greg

the result is:

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro

XML DOCUMENTS SLICING 6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro

XML DOCUMENT METRICS

Effective management of any process requires quantification, measurement, and modeling.

Software metrics provide a quantitative basis for the development and validation of models of the software development process

Metrics can be used to improve software productivity and quality

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro


In the field of XML, quality assessment is also relevant because the approach followed by engineers or end-users, to design the annotation schema or even to markup existent tests, is many times improvised and naïf.

Concepts like well-formedness or validity are not sufficient to appraise XML documents.

So, a set of metrics were defined to form the basis of the quality measurement of a XML document.

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro


Size Structure Complexity Structure Depth Fan-in / Fan-out Instability Tree impurity Attributes per Element Non-used components Text length

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro


Sucessor Graph

Given a DTD, we say that a new component (element/attribute) is an immediate successor of the element under definition.

Then, we introduce an arrow (oriented edge) from the element to the component.

Example:< !ELEMENT Item (FileName, Artist?) >

<!ELEMENT FileName (#PCDATA)>

<!ELEMENT Artist (#PCDATA)>

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro

SUCESSOR GRAPH (ROMEO AND JULIET

SCREENPLAY) 6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro


Size

Given a DTD, its size (i.e. the value for this metric) is the total number of nodes in the SG (number of DTD components).

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro


Structure complexity

Where e is the number of edges in the SG, n is the number of nodes in the SG and n_idref is the number of IDREF attributes.

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro


Structure Depth

According to Meike Klettke, in Metrics for XML document collections, a SG with a depth much higher than 7 is complex and reveals a bad DTD design.

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro


Fan-in / Fan-out

For the graph as a whole, the average and the maximum values for those parameters can be useful to spot unusual nodes, which can be inspected to detect the anomaly and fix the problem.

Elements with a high Fan-in/Fan-out value are more complex than other elements with a lower value.

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro


Instability

A node with a low instability allows us to conclude that it is less dependent of other nodes, while many nodes are depend on it.

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro


Tree Impurity

A tree impurity of 0% means that a graph is a tree and a tree impurity of 100% means that it is a fully connected graph.

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro


Attributes per Element

The AttrsEle(DTD) metric allows us to figure out the average number of attributes defined per element in the DTD.

The AttrsEle(XML) metric, applied directly to the XML document, allows us to figure out the average number of attributes actually used per effective elements present in the XML document.

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro


Non-used Components

if Attr(DTD) represents the set of attributes defined in the DTD, and Attr(XML) represents the set of actual attributes (the attributes used in the XML document instance), then NonAttr(XML) is the set of non-used attributes.

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro


Text Length

where, length(PCDATA) computes the total length of the document's text (the sum of the length of all text fragments, i.e., text associated with element tags, or untagged text), and nPCDATA is the number of text fragments (the number of PCDATA leaves that appear in the XML document tree).

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro

METRIC RESULTS (ROMEO AND JULIET SCREENPLAY)

Metric Value

Size 27

Structure Complexity 13

Structure Depth 7

Fan-in (node scene) 3

Fan-out (node scene) 6

Instability 3,3%

Tree impurity 58,9%

Attributes per Elem (DTD) 0,08

Attributes per Elem (XML) 0,027

Non Used Components (Elem)

1 (stagedir element)

Text Length (Elem) 37,46

Text Length (Attr) 1

6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro

CONCLUSION 6 M

arço

, 20

09

Univ

ersid

ade d

e A

veiro

Documents

eXVisXML , uma ferramenta emblemática na análise documental