89 lines
3.0 KiB
Python
89 lines
3.0 KiB
Python
from nltk.tokenize import word_tokenize
|
|
from nltk import trigrams
|
|
from collections import defaultdict, Counter
|
|
import pandas as pd
|
|
import csv
|
|
import regex as re
|
|
|
|
|
|
trigrams_list = []
|
|
model = defaultdict(lambda: defaultdict(lambda: 0))
|
|
|
|
|
|
def preprocess(text):
|
|
text = str(text).lower().replace("-\\n", "").replace("\\n", " ")
|
|
text = re.sub(r'\p{P}', '', text)
|
|
words = word_tokenize(text)
|
|
if len(words):
|
|
return words
|
|
return ['']
|
|
|
|
|
|
def predict(word_before, word_after):
|
|
prob_list = dict(Counter(model[(word_before, word_after)]).most_common(5)).items()
|
|
predictions = []
|
|
prob_sum = 0.0
|
|
for key, value in prob_list:
|
|
prob_sum += value
|
|
predictions.append(f'{key}:{value}')
|
|
if prob_sum == 0.0:
|
|
return 'the:0.2 be:0.2 to:0.2 of:0.1 and:0.1 a:0.1 :0.1'
|
|
elif prob_sum < 1.0:
|
|
predictions.append(f':{max(1 - prob_sum, 0.01)}')
|
|
return ' '.join(predictions)
|
|
|
|
|
|
file_in = pd.read_csv('train/in.tsv.xz', sep='\t', on_bad_lines='skip', header=None, quoting=csv.QUOTE_NONE, nrows=200000)
|
|
file_expected = pd.read_csv('train/expected.tsv', sep='\t', on_bad_lines='skip', header=None, quoting=csv.QUOTE_NONE, nrows=200000)
|
|
|
|
for index, (line_in, expected) in enumerate(zip(file_in.iterrows(), file_expected.iterrows())):
|
|
if index % 1000 == 0:
|
|
print('zbieranie trigramów', index)
|
|
before = line_in[1][6]
|
|
after = line_in[1][7]
|
|
expected = expected[1][0]
|
|
before, expected, after = preprocess(before), preprocess(expected), preprocess(after)
|
|
words = before + expected + after
|
|
trigrams_list += trigrams(words, pad_right=True, pad_left=True)
|
|
length = len(trigrams_list)
|
|
|
|
trigrams_len = len(trigrams_list)
|
|
|
|
for index, trigram in enumerate(trigrams_list):
|
|
if index % 100000 == 0:
|
|
print(f'uczenie modelu: {index / trigrams_len}')
|
|
if trigram[0] and trigram[1] and trigram[2]:
|
|
model[(trigram[0], trigram[2])][trigram[1]] += 1
|
|
|
|
|
|
model_len = len(model)
|
|
|
|
for index, words_1_3 in enumerate(model):
|
|
if index % 100000 == 0:
|
|
print(f'normalizacja: {index / model_len}')
|
|
count = sum(model[words_1_3].values())
|
|
for word_2 in model[words_1_3]:
|
|
model[words_1_3][word_2] += 1
|
|
model[words_1_3][word_2] /= float(count + len(word_2)))
|
|
|
|
|
|
def make_prediction(file):
|
|
file_in = pd.read_csv(f'{file}/in.tsv.xz', sep='\t', on_bad_lines='skip', header=None, quoting=csv.QUOTE_NONE)
|
|
with open(f'{file}/out.tsv', 'w', encoding='utf-8') as file_out:
|
|
print(f'zapisywanie {file}')
|
|
for line_in in file_in.iterrows():
|
|
before = line_in[1][6]
|
|
after = line_in[1][7]
|
|
if len(before) < 3 or len(after) < 3:
|
|
prediction = 'the:0.2 be:0.2 to:0.2 of:0.1 and:0.1 a:0.1 :0.1'
|
|
else:
|
|
word_before_in, word_after_in = preprocess(before)[-1], preprocess(after)[0]
|
|
prediction = predict(word_before_in, word_after_in)
|
|
file_out.write(prediction + '\n')
|
|
|
|
|
|
make_prediction('test-A')
|
|
make_prediction('dev-0')
|
|
|
|
print('koniec')
|