I use WEKA for data mining, which uses ARFF files. 

WEKA is a powerful tool, but sometimes I want to look at the datasets in a more universal format where I can manipulate them as I please.  This Python script will convert ARFF files to XML or JSON.

Usage:

./arff_parser.py [filename1.arff filename2.arff ... ] [-json | -xml] [--debug] 

# Debug mode will print a variable dump of the schema read in. Default value is False.

Expects:
	
	- a WEKA file ('some-filename.arff') formatted as follows:
	
		@relation <dataset-name>
		
		% Comments (ignored)
		
		@attribute1 [{value1,value2,value3} | datatype] 
		@attribute2 [{value1,value2,value3} | datatype] 
		...
		
		@data
		[attr1,attr2,...]
		[attr1,attr2,...]
		[attr1,attr2,...]
		[attr1,attr2,...]
		
		
Outputs:
	

	-json:
		
		'some-filename.json'
		
		{
			"relation":dataset-name,
			"attributes":[
				{'name':attribute1,'values': [value1,value2...]},
				{'name':attribute2,'values': [value1,value2...]}
			],
			"data":[
				{attr1:val1,attr2:val2...},
				{attr1:val1,attr2:val2...},
				{attr1:val1,attr2:val2...},
				...
			]
		}
		
	-xml:
		
		'some-filename.xml'
		
		<dataset>
			<relation>dataset-name</relation>
			<attributes>
				<attribute>
					<name>attribute1</name>
					<values>real</values>
				</attribute>
				<attribute>
					<name>attribute2</name>
					<values>
						<value>val1</value>
						<value>val2</value>
						<value>val3</value>
						<value>...</value>
					</values>
				</attribute>
			</attributes>
			<data>
				<entry><attr1>val</attr1><attr2>val</attr2>...</entry>
				<entry><attr1>val</attr1><attr2>val</attr2>...</entry>
				...
			</data>
		</dataset>