Seungwon bowbowbow

Hello

bowbowbow / download_youtube_srt.py

Last active January 28, 2020 11:03

	import os
	import pysrt


	def download_youtube_srt(link):
	video_id = link[link.find('?v=') + 3:]
	import subprocess
	subprocess.call(['youtube-dl',
	# for sub
	'--sub-lang', 'en',

bowbowbow / extract_morpheme_from_sejong_corpus.py

Last active December 16, 2019 19:51

세종 코퍼스에서 형태소 분석 데이터 추출

	import glob
	from bs4 import BeautifulSoup
	from tqdm import tqdm
	import json

	if __name__ == '__main__':
	paths = glob.glob('./corpus-utf8/*.txt')
	print('len(paths):', len(paths))

	pair_count = 0

bowbowbow / load_glove.py

Created November 14, 2019 02:05

	def load_glove(glove_path):
	word2vec = {}
	with open(glove_path, 'r') as f:
	for line in f:
	split_line = line.split()
	word = split_line[0]
	embedding = np.array([float(val) for val in split_line[1:]])
	word2vec[word] = embedding
	print("Done.",len(word2vec)," words loaded!")
	return word2vec

bowbowbow / bert_sound_model_server.py

Last active August 23, 2019 03:51

	import os
	import time
	import sys
	import datetime
	import random
	import json
	import pickle
	import numpy as np
	from flask import Flask, session, g, request, render_template, redirect
	from flask_mongoengine import MongoEngine

bowbowbow / elasticsearch_old_index.py

Created August 21, 2019 18:43

	"settings": {
	"index": {
	"analysis": {
	"analyzer": {
	"korean": {
	"type": "custom",
	"tokenizer": "seunjeon_default_tokenizer",
	"min_token_length": 2,
	"decompound_mode": "mixed",
	"filter": ['cobak_synonym_filter', 'token_length_filter'],

bowbowbow / sound_text_model.py

Created August 21, 2019 16:05

	# Convolutional neural network (two convolutional layers)
	class ConvNet(nn.Module):
	def __init__(self, hidden_size=200):
	super(ConvNet, self).__init__()
	self.hidden_size = hidden_size

	self.layer1 = nn.Sequential(
	nn.Conv1d(34, 68, kernel_size=5, stride=1, padding=2),
	nn.BatchNorm1d(68),
	nn.CELU(),

bowbowbow / bert_net.py

Created August 17, 2019 19:26

	class BertNet(nn.Module):
	def __init__(self, finetuning=False, num_classes=3, hidden_size=50):
	super().__init__()
	self.bert = BertModel.from_pretrained('bert-base-uncased')
	self.bert_output_size = 768
	self.hidden_size = hidden_size

	self.rnn = nn.LSTM(input_size=self.bert_output_size, hidden_size=self.hidden_size, batch_first=True, bidirectional=True)
	self.fc = nn.Linear(self.hidden_size * 2, num_classes)
	self.finetuning = finetuning

bowbowbow / draw.py

Last active January 3, 2019 09:34

	def draw(epoch, input_x, input_y, predicts, input_c_pos, id2label, id2word):
	sents_visual_file = './visualization/{}.html'.format(epoch)

	batch_size = len(input_y)
	with open(sents_visual_file, "w") as html_file:
	html_file.write('<!DOCTYPE html><html lang="ko"><head><meta charset="UTF-8"/></head>')

	for i in range(batch_size):
	if input_y[i] == predicts[i]: continue

bowbowbow / pymongo_example.py

Created December 4, 2018 04:02

	import pymongo

	_connection = None


	def get_db():
	global _connection
	if not _connection:
	_connection = pymongo.MongoClient('mongodb://127.0.0.1')
	return _connection['dbname']

bowbowbow / find_candidates_by_keyword.py

Created December 3, 2018 16:51