def process(ptb_file, ccg_file, deps_file, ccg_auto_out, ccg_parg_out, higher, quotes, quoter):
'''Reinstates quotes given a PTB file and its corresponding CCGbank file and deps file.'''
with file(ccg_auto_out, 'w') as ccg_out:
with file(ccg_parg_out, 'w') as parg_out:
penn_trees = list(PTBReader(ptb_file))
ccg_trees = list(CCGbankReader(ccg_file))
deps = list(CCGbankDepsReader(deps_file))
matched_penn_trees = match_trees(penn_trees, ccg_trees)
for (ptb_bundle, ccg_bundle, dep) in zip(matched_penn_trees, ccg_trees, deps):
ptb_tree, ccg_tree = ptb_bundle.derivation, ccg_bundle.derivation
quote_spans = spans(ptb_tree)
while quote_spans:
value = quote_spans.pop(0)
span_start, span_end, quote_type = value
if span_start is None and span_end is None: continue
info("Reinstating quotes to %s (%s, %s)", ccg_bundle.label(), span_start, span_end)
ccg_tree, quote_indices = quoter.attach_quotes(ccg_tree, span_start, span_end, quote_type, higher, quotes)
# In case a new root has been installed, re-assign the new root to the CCGbank bundle
ccg_bundle.derivation = ccg_tree
# Shift remaining quote span indices by the number of quotes that have been inserted
quote_spans = fix_quote_spans(quote_spans, quote_indices)
dep = fix_dependencies(dep, quote_indices)
print >> parg_out, dep
print >> ccg_out, ccg_bundle
def __iter__(self):
for left, right in izip(self.reader(self.leftdir), self.reader(self.rightdir)):
info("Processing %s/%s", left.label(), right.label())
deriv = Derivation(left, right)
yield deriv
del deriv
del left
del right
def run_filters(self, filters, files):
# If all given filters were not found or had wrong argument count, do nothing
if not filters: return
reader_args = {}
if self.reader_class_name:
try:
reader_class = globals()[self.reader_class_name]
info("Using reader class %s.", self.reader_class_name)
reader_args['reader_class'] = reader_class
except KeyError:
raise RuntimeError("Reader class %s not found." % self.reader_class_name)
for file in self.transform(files):
if self.is_pair_spec(file):
meta_reader = PairedReader
else:
meta_reader = DirFileGuessReader
try:
self.last_exceptions = []
for derivation_bundle in meta_reader(file, verbose=self.verbose, **reader_args):
if self.verbose: info("Processing %s...", derivation_bundle.label())
try:
for filter in filters:
filter.context = derivation_bundle
if filter.accept_leaf is not None:
for leaf in leaves(derivation_bundle.derivation):
for filter in filters:
filter.accept_leaf(leaf)
if filter.accept_comb_and_slash_index is not None:
try:
for slash_index, comb in enumerate(applications_per_slash(leaf)):
filter.accept_comb_and_slash_index(leaf, comb, slash_index)
except AttributeError: # TODO: hacky and inefficient, need this to work for PTB too
pass
for filter in filters:
filter.accept_derivation(derivation_bundle)
filter.context = None
except IOError, e:
# If output is going to a pager, and the user requests an interrupt (^C)
# the filter fails with IOError: Broken pipe
# In that case, running filters on further derivations will continue to
# lead to 'Broken pipe', so just bail out
if e.errno == errno.EPIPE: return
except Exception, e:
self.last_exceptions.append( (derivation_bundle, sys.exc_info()) )
if self._break_on_exception:
raise FilterException(e, None)
def main(argv):
parser = OptionParser()
register_builtin_switches(parser)
opts, args = parser.parse_args(argv)
if not all_required_args_present(opts):
parser.print_help()
sys.exit(1)
quoter_class = {
'span': SpanQuoter,
'lca' : LCAQuoter
}[opts.quote_method]
punct_class = {
'swap' : SwapComma,
'shift': ShiftComma
}.get(opts.punct_method, None)
quoter = quoter_class(punct_class)
remaining_args = args[1:]
if not remaining_args:
# If no sec/doc specifiers are given, assume 'all sections all documents'
remaining_args.append(':')
ptb_files_spec = parse_requested_derivs(remaining_args)
for sec_glob, doc_glob in ptb_files_spec:
for ptb_file in glob(os.path.join(opts.penn_in, sec_glob, "wsj_%s%s.mrg" % (sec_glob, doc_glob))):
info("Processing %s", ptb_file)
matches = PTBFileRegex.search(ptb_file)
if matches and len(matches.groups()) == 2:
sec, doc = matches.groups()
ccg_file = os.path.join(opts.ccg_in, 'AUTO', sec, "wsj_%s%s.auto" % (sec, doc))
deps_file = os.path.join(opts.ccg_in, 'PARG', sec, "wsj_%s%s.parg" % (sec, doc))
if not opts.quiet:
if not os.path.exists(ccg_file):
warn("No corresponding CCGbank file %s for Penn file %s", ccg_file, ptb_file)
if not os.path.exists(deps_file):
warn("No corresponding CCGbank dependency file %s for CCG file %s", deps_file, ccg_file)
ccg_auto_dir, ccg_parg_dir = [os.path.join(opts.outdir, part, sec) for part in ('AUTO', 'PARG')]
if not os.path.exists(ccg_auto_dir): os.makedirs(ccg_auto_dir)
if not os.path.exists(ccg_parg_dir): os.makedirs(ccg_parg_dir)
ccg_auto_out, ccg_parg_out = (os.path.join(ccg_auto_dir, 'wsj_%s%s.auto' % (sec, doc)),
os.path.join(ccg_parg_dir, 'wsj_%s%s.parg' % (sec, doc)))
process(ptb_file, ccg_file, deps_file, ccg_auto_out, ccg_parg_out,
opts.higher, opts.quotes, quoter)
else:
warn("Could not find, so ignoring %s", ptb_file)
def __iter__(self):
for left, right in izip(self.reader(self.leftdir),
self.reader(self.rightdir)):
info("Processing %s/%s", left.label(), right.label())
deriv = Derivation(left, right)
yield deriv
del deriv
del left
del right
def __iter__(self):
# TODO: duplication
for section_path in self.sections:
# If _topdir_ has directories under, expand to use the files it contains
if os.path.isdir(section_path):
docs = glob(os.path.join(section_path, '*'))
for doc_path in docs:
if self.verbose: info("Processing %s...", doc_path)
reader = self.reader(doc_path)
for deriv_bundle in reader:
yield deriv_bundle
del reader
# Otherwise _topdir_ is flat: read the files it contains
else:
reader = self.reader(section_path)
for deriv_bundle in reader:
yield deriv_bundle
del reader
def process(ptb_file, ccg_file, deps_file, ccg_auto_out, ccg_parg_out, higher,
quotes, quoter):
'''Reinstates quotes given a PTB file and its corresponding CCGbank file and deps file.'''
with file(ccg_auto_out, 'w') as ccg_out:
with file(ccg_parg_out, 'w') as parg_out:
penn_trees = list(PTBReader(ptb_file))
ccg_trees = list(CCGbankReader(ccg_file))
deps = list(CCGbankDepsReader(deps_file))
matched_penn_trees = match_trees(penn_trees, ccg_trees)
for (ptb_bundle, ccg_bundle, dep) in zip(matched_penn_trees,
ccg_trees, deps):
ptb_tree, ccg_tree = ptb_bundle.derivation, ccg_bundle.derivation
quote_spans = spans(ptb_tree)
while quote_spans:
value = quote_spans.pop(0)
span_start, span_end, quote_type = value
if span_start is None and span_end is None: continue
info("Reinstating quotes to %s (%s, %s)",
ccg_bundle.label(), span_start, span_end)
ccg_tree, quote_indices = quoter.attach_quotes(
ccg_tree, span_start, span_end, quote_type, higher,
quotes)
# In case a new root has been installed, re-assign the new root to the CCGbank bundle
ccg_bundle.derivation = ccg_tree
# Shift remaining quote span indices by the number of quotes that have been inserted
quote_spans = fix_quote_spans(quote_spans, quote_indices)
dep = fix_dependencies(dep, quote_indices)
print >> parg_out, dep
print >> ccg_out, ccg_bundle
def main(argv):
parser = OptionParser()
register_builtin_switches(parser)
opts, args = parser.parse_args(argv)
if not all_required_args_present(opts):
parser.print_help()
sys.exit(1)
quoter_class = {'span': SpanQuoter, 'lca': LCAQuoter}[opts.quote_method]
punct_class = {
'swap': SwapComma,
'shift': ShiftComma
}.get(opts.punct_method, None)
quoter = quoter_class(punct_class)
remaining_args = args[1:]
if not remaining_args:
# If no sec/doc specifiers are given, assume 'all sections all documents'
remaining_args.append(':')
ptb_files_spec = parse_requested_derivs(remaining_args)
for sec_glob, doc_glob in ptb_files_spec:
for ptb_file in glob(
os.path.join(opts.penn_in, sec_glob,
"wsj_%s%s.mrg" % (sec_glob, doc_glob))):
info("Processing %s", ptb_file)
matches = PTBFileRegex.search(ptb_file)
if matches and len(matches.groups()) == 2:
sec, doc = matches.groups()
ccg_file = os.path.join(opts.ccg_in, 'AUTO', sec,
"wsj_%s%s.auto" % (sec, doc))
deps_file = os.path.join(opts.ccg_in, 'PARG', sec,
"wsj_%s%s.parg" % (sec, doc))
if not opts.quiet:
if not os.path.exists(ccg_file):
warn(
"No corresponding CCGbank file %s for Penn file %s",
ccg_file, ptb_file)
if not os.path.exists(deps_file):
warn(
"No corresponding CCGbank dependency file %s for CCG file %s",
deps_file, ccg_file)
ccg_auto_dir, ccg_parg_dir = [
os.path.join(opts.outdir, part, sec)
for part in ('AUTO', 'PARG')
]
if not os.path.exists(ccg_auto_dir): os.makedirs(ccg_auto_dir)
if not os.path.exists(ccg_parg_dir): os.makedirs(ccg_parg_dir)
ccg_auto_out, ccg_parg_out = (os.path.join(
ccg_auto_dir, 'wsj_%s%s.auto' % (sec, doc)),
os.path.join(
ccg_parg_dir,
'wsj_%s%s.parg' %
(sec, doc)))
process(ptb_file, ccg_file, deps_file, ccg_auto_out,
ccg_parg_out, opts.higher, opts.quotes, quoter)
else:
warn("Could not find, so ignoring %s", ptb_file)
def run_filters(self, filters, files):
# If all given filters were not found or had wrong argument count, do nothing
if not filters: return
reader_args = {}
if self.reader_class_name:
try:
reader_class = globals()[self.reader_class_name]
info("Using reader class %s.", self.reader_class_name)
reader_args['reader_class'] = reader_class
except KeyError:
raise RuntimeError("Reader class %s not found." %
self.reader_class_name)
for file in self.transform(files):
if self.is_pair_spec(file):
meta_reader = PairedReader
else:
meta_reader = DirFileGuessReader
try:
self.last_exceptions = []
for derivation_bundle in meta_reader(file,
verbose=self.verbose,
**reader_args):
if self.verbose:
info("Processing %s...", derivation_bundle.label())
try:
for filter in filters:
filter.context = derivation_bundle
if filter.accept_leaf is not None:
for leaf in leaves(derivation_bundle.derivation):
for filter in filters:
filter.accept_leaf(leaf)
if filter.accept_comb_and_slash_index is not None:
try:
for slash_index, comb in enumerate(
applications_per_slash(
leaf)):
filter.accept_comb_and_slash_index(
leaf, comb, slash_index)
except AttributeError: # TODO: hacky and inefficient, need this to work for PTB too
pass
for filter in filters:
filter.accept_derivation(derivation_bundle)
filter.context = None
except IOError, e:
# If output is going to a pager, and the user requests an interrupt (^C)
# the filter fails with IOError: Broken pipe
# In that case, running filters on further derivations will continue to
# lead to 'Broken pipe', so just bail out
if e.errno == errno.EPIPE: return
except Exception, e:
self.last_exceptions.append(
(derivation_bundle, sys.exc_info()))
if self._break_on_exception:
raise FilterException(e, None)
