AI & Data

How to handle XML?

XML = eXtensible Markup Language

• Data are structured in a hierarchy of node
• Nodes can be elements, text nodes or attributes
• It can be used to represent structures with complex details

XML: an example

<breakfast_menu>
	<food>
		<name>Belgian Waffles</name>
        <price>$5.95</price>
		<description>Two of our famous Belgian Waffles with...</description>
		<calories>650</calories>
    </food>
</breakfast_menu>

• <breakfast_menu>: the root element
• <food>: an element
• Belgian Waffles: a text content

XML: targeting element thanks to XPath

• One can query the document thanks to XPath expressions
• XPath takes the path from the root element to the targeted node
• Each level is represented by a /

XML: Read a document using python (native option)

import xml.etree.ElementTree as xmlReader
# read from wml
tree = xmlReader.parse('menu.xml')

XML: Read a document using python (with lxml)

from lxml import etree
# read from xml
tree = etree.parse('menu.xml')
root = tree.getroot()
print(root)

lxml has a more extensive support of XPath, and it is really convenient

XML: get the elements in a list using xpath

elems = root.findall('./food')
data = [[elem.find("./name").text,
         elem.find("./price").text
         ] for elem in elems]

print(data)

XPATH 101:

• When a query starts with :/: the query looks data from the root
• query starts with ./: path is relatively taken from the current path
• one can filter data thanks to predicates, example

  ./food[starts-with(./name/text(), 'Be')]

Exercise : load this xml file from your preferred python environment, then do the same in Orange (using Python Script widget)

Bind xml results to an Orange data table

• It is possible to bind "raw data" in Orange tables
• this is available thanks to this code snippet (from Orange docs):

  from Orange.data import *
  data = [
      ['green', 4, 1.2, 'apple'],
      ['orange', 5, 1.1, 'orange'],
      ['yellow', 4, 1.0, 'peach']
  ]
  color = DiscreteVariable('color', values=set([row[0] for row in data]))
  calories = ContinuousVariable('calories')
  fiber = ContinuousVariable('fiber')
  fruit = DiscreteVariable('fruit', values=set([row[3] for row in data]))
  
  domain = Domain([color, calories, fiber], class_vars=fruit)
  
  table = Table.from_list(domain, data)

XML exercises with Orange

• In Orange, import the previous file through a widget named "Python Script" and transform it to a data table.
• Store this data table in a variable named "output_data".
• Then put a table widget after the python script widget

XML exercises with Orange (2)

• Consider the zip: file, it contains data about article published on ML over time in XML format
• try yourself on extracting data on one file
• extract the year, month, title, and add a column "topic" that contains always ML
• iterate over the files present in this folder to extract all the information, and then integrate it in orange thanks to the Python script widget
• Analyse data thanks to orange (visualization)