ozbolt/luscenje_struktur
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

lemma/msd representations now global! Need to also use for agreements

Browse Source
This commit is contained in:
Ozbolt Menegatti 2019-05-22 11:55:51 +02:00
parent d14efff709
commit e99ba59908
1 changed files with 21 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

35
wani.py
View File

@ -217,7 +217,7 @@ class ComponentRendition:
return self.rendition is rendition return self.rendition is rendition
@staticmethod @staticmethod
def set_representations(matches, structure): def set_representations(matches, structure, word_renderer):
representations = { representations = {
c.idx: [[], None] if c.representation.isit(Rendition.WordForm) else [True, ""] c.idx: [[], None] if c.representation.isit(Rendition.WordForm) else [True, ""]
for c in structure.components for c in structure.components
@ -236,11 +236,8 @@ class ComponentRendition:
for word in sorted_lst: for word in sorted_lst:
othw = are_agreements_ok(word, representations_to_check) othw = are_agreements_ok(word, representations_to_check)
if othw is not None: if othw is not None:
if doprint: matches.representations[word_component_id[othw.id]] = othw.most_frequent_text(word_renderer)
print("AOK", othw.text, othw) matches.representations[word_component_id[word.id]] = word.most_frequent_text(word_renderer)
matches.representations[word_component_id[othw.id]] = othw.text
matches.representations[word_component_id[word.id]] = word.text
return return
def are_agreements_ok(word, words_to_try): def are_agreements_ok(word, words_to_try):
@ -859,16 +856,19 @@ class Word:
return self.links[link] return self.links[link]
def most_frequent_text(self, word_renderer):
return word_renderer.render(self.lemma, self.msd)
class WordMsdRenderer: class WordMsdRenderer:
def __init__(self): def __init__(self):
self.all_words = [] self.all_words = []
self.rendered_words = {} self.rendered_words = {}
def add_word(self, word): def add_words(self, words):
self.all_words.append(word) self.all_words.extend(words)
def generate_renders(self): def generate_renders(self):
data = defaultdict(lambda: defaultdict([])) data = defaultdict(lambda: defaultdict(list))
for w in self.all_words: for w in self.all_words:
data[w.lemma][w.msd].append(w.text) data[w.lemma][w.msd].append(w.text)
@ -1114,10 +1114,10 @@ class ColocationIds:
if group: if group:
break break
def set_representations(self, structures): def set_representations(self, structures, word_renderer):
components_dict = {structure.id: structure for structure in structures} components_dict = {structure.id: structure for structure in structures}
for _1, sm in self.data.items(): for _1, sm in self.data.items():
ComponentRendition.set_representations(sm, components_dict[sm.structure_id]) ComponentRendition.set_representations(sm, components_dict[sm.structure_id], word_renderer)
def match_file(words, structures): def match_file(words, structures):
@ -1144,6 +1144,7 @@ def main(input_file, structures_file, args):
logging.debug(str(s)) logging.debug(str(s))
colocation_ids = ColocationIds() colocation_ids = ColocationIds()
word_renderer = WordMsdRenderer()
if args.parallel: if args.parallel:
num_parallel = int(args.parallel) num_parallel = int(args.parallel)
@ -1170,22 +1171,28 @@ def main(input_file, structures_file, args):
# fancy interface to wait for threads to finish # fancy interface to wait for threads to finish
for id_input in executor.map(func, [i for i, _ in enumerate(args.input)]): for id_input in executor.map(func, [i for i, _ in enumerate(args.input)]):
with open("{}/{}.p".format(tmpdirname, id_input), "rb") as fp: with open("{}/{}.p".format(tmpdirname, id_input), "rb") as fp:
matches = pickle.load(fp) words, matches = pickle.load(fp)
colocation_ids.add_matches(matches) colocation_ids.add_matches(matches)
word_renderer.add_words(words)
else: else:
for words in load_files(args): for words in load_files(args):
matches = match_file(words, structures) matches = match_file(words, structures)
# just save to temporary file, used for children of a parallel process # just save to temporary file, used for children of a parallel process
# MUST NOT have more than one file
if args.match_to_file is not None: if args.match_to_file is not None:
with open(args.match_to_file, "wb") as fp: with open(args.match_to_file, "wb") as fp:
pickle.dump(matches, fp) pickle.dump((words, matches), fp)
return return
else: else:
colocation_ids.add_matches(matches) colocation_ids.add_matches(matches)
word_renderer.add_words(words)
# get word renders for lemma/msd
word_renderer.generate_renders()
# figure out representations! # figure out representations!
colocation_ids.set_representations(structures) colocation_ids.set_representations(structures, word_renderer)
if args.all: if args.all:
Writer.make_all_writer(args).write_out(structures, colocation_ids) Writer.make_all_writer(args).write_out(structures, colocation_ids)