def main(goldFile, predFile, outCSVFile, description, fieldName1, fieldValue1):
(goldRoot, goldNodes) = evaluate.getCoarseGrainedTreeLabelsFile(goldFile)
(predRoot, predNodes) = evaluate.getCoarseGrainedTreeLabelsFile(predFile)
rootData = evaluate.printStatsCoarseInt(goldRoot, predRoot, prefix='root')
nodeData = evaluate.printStatsCoarseInt(goldNodes, predNodes)
nodeData.update(rootData)
if description:
nodeData = evaluate.ins(['Description'],
[description], nodeData)
if fieldName1 and fieldValue1:
nodeData = evaluate.ins([fieldName1],
[fieldValue1], nodeData)
evaluate.statsToFile(nodeData, outCSVFile, delim=';')
def rootAccStatistic(gold, predicted):
# This looks somewhat convoluted - that's some legacy
# from the CSV code.
# combinedRootLabelAccMacro has always been used for root
# accuracy. This method can be used to compute any
# macro accuracy as 'root' is only symbolic,
# but I only ever use it for root accuracy
data = evaluate.printStatsCoarseInt(gold, predicted, prefix='root')
return data['combinedRootLabelAccMacro']
def nodeF1Statistic(gold, predicted):
data = evaluate.printStatsCoarseInt(gold, predicted)
return data['combinedLabelFMacro']
