Junki Ohmura jojonki

Sony Researcher. Previously, visiting researcher at Language Technologies Institute, Carnegie Mellon University. Email: junki dot ohmura at gmail.

jojonki / simple_viterbi.py

Created January 3, 2019 06:45

Simple viterbi algorithm example

	# Use Graham's example.
	# http://www.phontron.com/slides/nlp-programming-ja-03-ws.pdf

	INF = 1e6
	edge_list = [
	None, # e0
	{ # e1
	'id': 1,
	'score': 2.5,
	'begin_node_id': 0,

jojonki / global_variables.py

Created February 4, 2019 07:29

Hack for read-only class variables.

	class MetaGlobalVariables(type):
	@property
	def HOGE(cls):
	return cls._GlobalVariables__hoge # mangling


	class GlobalVariables(object, metaclass=MetaGlobalVariables):
	__hoge = 'xxxx'

	# read OK

jojonki / timeshift_sony_actioncam.py

Created October 27, 2019 03:00

Timeshift modification for Sony's actioncam on Mac

	from datetime import datetime
	from pytz import timezone
	import subprocess
	import glob
	import xml.etree.ElementTree as ET

	# replaced time zones
	SRC_TZ_ZONES = ['UTC+09:00', 'UTC+01:00']
	# target time zone
	TGT_TZ_ZONE = ['Europe/Berlin']

jojonki / plotly_orca_example.py

Last active December 12, 2019 06:13

Call quay.io/plotly/orca API

	"""Use plotly orca (Docker) to save png/html files.

	Run plotly-orca server before running this code.
	% docker run -d -p 9091:9091 quay.io/plotly/orca

	I refer the following forum to write this code and slightly modify the code for proper import.
	https://community.plot.ly/t/plotly-io-with-external-orca/25600/2
	"""
	import io
	import requests

jojonki / launch.json

Last active January 18, 2020 00:47

python vscode sandbox

	{
	// Use IntelliSense to learn about possible attributes.
	// Hover to view descriptions of existing attributes.
	// For more information, visit: https://go.microsoft.com/fwlink/?linkid=830387
	"version": "0.2.0",
	"configurations": [
	{
	"name": "Python: Current File",
	"type": "python",
	"pythonPath": "${config:python.pythonPath}",

jojonki / convert.py

Created January 17, 2020 12:10

Convert latex to text with pandocfilters

	"""Remove latex blocks and replace math functiosn with (MATH)

	- How to debug:
	pandoc -t json aaa.tex \| python ./this-file.py \| pandoc -f json -t plain

	- How to use:
	pandoc -s aaa.tex --filter=./this-file.py -o out.text
	"""
	from pandocfilters import toJSONFilter, Str

jojonki / lower_bound_and_upper_bound.py

Last active March 5, 2020 13:53

	# けんちょんさんのめぐるん式二部探索解説
	# https://qiita.com/drken/items/97e37dd6143e33a64c8c

	def lowerBound(nums, key):
	"""numsの中でkey以上の要素のうちの一番左のインデックス"""
	ng, ok = -1, len(nums)
	while abs(ok - ng) > 1:
	mid = (ok + ng) // 2
	if key <= nums[mid]:
	ok = mid

jojonki / .direnvrc

Last active February 20, 2021 17:58

direnv for anaconda witch automatically switches anaconda environments.

jojonki / bpe.py

Created June 17, 2020 08:53

BPE (Neural Machine Translation of Rare Words with Subword Units, Rico Sennrich.)

	import collections
	import re


	def get_stats(vocab):
	pairs = collections.defaultdict(int)
	for word, freq in vocab.items():
	symbols = word.split()
	for i in range(len(symbols)-1):
	pairs[symbols[i],symbols[i+1]] += freq

jojonki / gist:eb158ba2ed7a886b4f555fb2e952188f

Created October 7, 2020 14:25

test