"""
This block searches for Spanish ditransitive verbs with two clitics (e.g. 'dárselo').
The hypothesis is that in most such cases, 'se' just substitutes for 'le' or 'les'
and is not reflexive.
"""
from udapi.core.block import Block
[docs]
class MarkSeLo(Block):
def __init__(self, print_it=False, **kwargs):
"""
Default: Print the annotation patterns but do not fix anything.
fix=1: Do not print the patterns but fix them.
"""
super().__init__(**kwargs)
self.print_it = print_it
[docs]
def process_node(self, node):
if node.upos == 'VERB':
se_children = [x for x in node.children if x.form.lower() == 'se']
if len(se_children) >= 1:
lo_children = [x for x in node.children if x.form.lower() in ['lo', 'la', 'los', 'las']]
if len(lo_children) >= 1:
# Annotation errors lead to patterns like "los que se esperan".
# Therefore, require that "se" and "lo" are adjacent.
if abs(se_children[0].ord-lo_children[0].ord) == 1:
# We expect 1 se and 1 lo, not more.
# Nevertheless, for the moment, mark them all.
for x in se_children:
x.misc['Mark'] = 'se'
for x in lo_children:
x.misc['Mark'] = 'lo'
node.misc['Mark'] = 'VERB'
# We may either use the marks above with udapy -TAM or -HAM,
# or optionally print the observed examples to STDOUT and
# then aggregate them with | sort | uniq -c | sort -rn.
if self.print_it:
reflex = 'Yes' if se_children[0].feats['Reflex'] == 'Yes' else 'No'
print(f'{node.lemma}\tReflex={reflex}\tse={se_children[0].deprel}\tlo={lo_children[0].deprel}')