def parseDocFile(self, sFilename, iVerbose=0):
"""
Load that document as a CRF Graph.
Also set the self.doc variable!
Return a CRF Graph object
"""
self.doc = etree.parse(sFilename)
self.lNode, self.lEdge = list(), list()
self.lNodeBlock = [] # text node
self.lNodeCutLine = [] # cut line node
root = self.doc.getroot()
doer = BaselineCutAnnotator()
doer.setLabelScheme_SIO() #use SIO instead of SO labels!
#doer.setModulo(self.iModulo) # this is optional
#load the groundtruth table separators, if any, per page (1 in tABP)
ltlYlX = doer.get_separator_YX_from_DOM(root, self.fMinPageCoverage)
for (lHi, lVi) in ltlYlX:
traceln(" - found %d horizontal, %d vertical GT separators" %
(len(lHi), len(lVi)))
#create DOM node reflecting the cuts
#first clean (just in case!)
n = doer.remove_cuts_from_dom(root)
if n > 0:
traceln(" - removed %d pre-existing cut lines" % n)
# if GT, then we have labelled cut lines in DOM
_ltlYCutXCut = doer.add_cut_to_DOM(root, ltlYlX=ltlYlX)
lClassicType = [
nt for nt in self.getNodeTypeList() if nt in self._lClassicNodeType
]
lSpecialType = [
nt for nt in self.getNodeTypeList()
if nt not in self._lClassicNodeType
]
for (pnum, page, domNdPage) in self._iter_Page_DocNode(self.doc):
#now that we have the page, let's create the node for each type!
lClassicPageNode = [
nd for nodeType in lClassicType
for nd in nodeType._iter_GraphNode(self.doc, domNdPage, page)
]
lSpecialPageNode = [
nd for nodeType in lSpecialType
for nd in nodeType._iter_GraphNode(self.doc, domNdPage, page)
]
self.lNode.extend(lClassicPageNode) # e.g. the TextLine objects
self.lNodeBlock.extend(lClassicPageNode)
self.lNode.extend(lSpecialPageNode) # e.g. the cut lines!
self.lNodeCutLine.extend(lSpecialPageNode)
#no previous page to consider (for cross-page links...) => None
lClassicPageEdge = Edge.computeEdges(None, lClassicPageNode)
self.lEdge.extend(lClassicPageEdge)
# Now, compute edges between special and classic objects...
lSpecialPageEdge = self.computeSpecialEdges(
lClassicPageNode, lSpecialPageNode, doer.bCutIsBeforeText)
self.lEdge.extend(lSpecialPageEdge)
#if iVerbose>=2: traceln("\tPage %5d %6d nodes %7d edges"%(pnum, len(lPageNode), len(lPageEdge)))
if iVerbose >= 2:
traceln("\tPage %5d" % (pnum))
traceln("\t block: %6d nodes %7d edges (to block)" %
(pnum, len(lClassicPageNode), len(lClassicPageEdge)))
traceln("\t line: %6d nodes %7d edges (from block)" %
(pnum, len(lSpecialPageNode), len(lSpecialPageEdge)))
if iVerbose:
traceln("\t\t (%d nodes, %d edges)" %
(len(self.lNode), len(self.lEdge)))
return self
def parseDocFile(self, sFilename, iVerbose=0):
"""
Load that document as a CRF Graph.
Also set the self.doc variable!
Return a CRF Graph object
"""
self.doc = etree.parse(sFilename)
self.lNode, self.lEdge = list(), list()
self.lNodeBlock = [] # text node
self.lNodeGridLine = [] # grid line node
root = self.doc.getroot()
doer = GridAnnotator(self.iGridStep_H, self.iGridStep_V)
#map the groundtruth table separators, if any, to our grid
ltlHlV = doer.get_grid_GT_index_from_DOM(root, self.fMinPageCoverage)
for (lHi, lVi) in ltlHlV:
traceln(" - found %d horizontal, %d vertical GT separators" %
(len(lHi), len(lVi)))
#create DOM node reflecting the grid
#first clean (just in case!)
n = doer.remove_grid_from_dom(root)
if n > 0: traceln(" - removed %d existing grid lines" % n)
# we add GridSeparator elements. Groundtruth ones have type="1"
n = doer.add_grid_to_DOM(root, ltlHlV)
traceln(" - added %d grid lines %s" %
(n, (self.iGridStep_H, self.iGridStep_V)))
lClassicType = [
nt for nt in self.getNodeTypeList() if nt in self._lClassicNodeType
]
lSpecialType = [
nt for nt in self.getNodeTypeList()
if nt not in self._lClassicNodeType
]
for pnum, page, domNdPage in self._iter_Page_DocNode(self.doc):
#now that we have the page, let's create the node for each type!
lClassicPageNode = [
nd for nodeType in lClassicType
for nd in nodeType._iter_GraphNode(self.doc, domNdPage, page)
]
lSpecialPageNode = [
nd for nodeType in lSpecialType
for nd in nodeType._iter_GraphNode(self.doc, domNdPage, page)
]
self.lNode.extend(lClassicPageNode) # e.g. the TextLine objects
self.lNodeBlock.extend(lClassicPageNode)
self.lNode.extend(lSpecialPageNode) # e.g. the grid lines!
self.lNodeGridLine.extend(lSpecialPageNode)
#no previous page to consider (for cross-page links...) => None
lClassicPageEdge = Edge.computeEdges(None, lClassicPageNode)
self.lEdge.extend(lClassicPageEdge)
# Now, compute edges between special and classic objects...
lSpecialPageEdge = self.computeSpecialEdges(
lClassicPageNode, lSpecialPageNode)
self.lEdge.extend(lSpecialPageEdge)
#if iVerbose>=2: traceln("\tPage %5d %6d nodes %7d edges"%(pnum, len(lPageNode), len(lPageEdge)))
if iVerbose >= 2:
traceln("\tPage %5d" % (pnum))
traceln("\t block: %6d nodes %7d edges (to block)" %
(pnum, len(lClassicPageNode), len(lClassicPageEdge)))
traceln("\t line: %6d nodes %7d edges (from block)" %
(pnum, len(lSpecialPageNode), len(lSpecialPageEdge)))
if iVerbose:
traceln("\t\t (%d nodes, %d edges)" %
(len(self.lNode), len(self.lEdge)))
return self