su_task3-nlp/da_detector.py

#!/usr/bin/env python3
"""
this file holds a class, which implements a "detector" - language classifier -
that learns from predefined, frequency-analysed sets of ngrams
"""


class da_detector:
	def __init__(self, langs_to_check: list = ["sk", "en"]):
		# langs to check
		# to be picked from ["cz", "sk", "de", "en", "fr"]
		self.langs_to_check = langs_to_check

	def rm_interpunction(data):
		from string import punctuation
		for ngram in data:
			try:
				ngram = ngram.translate(str.maketrans('', '', punctuation))
			except Exception as e:
				raise e
		return data

	def rm_digits(data):
		from string import digits
		try:
			for ngram in data:
				ngram = ngram.translate(None, digits)
		except Exception as e:
			raise e
		return data

	def parse_freqs(self, path: str):
		import json

		fullpath = freqs_folder + path
		try:
			with open(fullpath, 'r') as f:
				j_data = f.read()
		except Exception as e:
			raise e

		try:
			obj = json.loads(j_data)
		except Exception as e:
			raise e
		return obj

	def pick_ngrams(self, what_grams: int, how_many: int, text: str):
		from random import randint
		if not isinstance(what_grams, int):
			raise TypeError("what_grams has to be an int")
		if not isinstance(how_many, int):
			raise TypeError("how_many has to be an int")
		if not isinstance(text, str):
			raise TypeError("text has to be a str")

		if (what_grams <= 0):
			raise ValueError("this is bogus, give me a number from ℕ")
		elif (what_grams > 5):
			raise ValueError("not doing larger-grams than 5")
		if (how_many <= 0):
			raise ValueError("how_many ought to be larger than 0")
		if (len(text) <= 10):
			raise ValueError("not doing anything with text shorter than 10 characters")

		t_len = len(text)
		# list of random n-grams
		r_ngrams = []
		# how many times to attempt to skin the cat
		insanity_threshold = 1000
		sanity_ctr = 0

		while (len(r_ngrams) < how_many and sanity_ctr < insanity_threshold):
			# not truly random, but hey..
			r_position = randint(0, t_len - 1)
			if (r_position + what_grams >= (t_len - 1)):
				continue

			# this is the block where we start counting how many times we've
			# been there
			++sanity_ctr
			candidate_ngram = text[r_position:r_position + what_grams]
			if (candidate_ngram not in r_ngrams):
				r_ngrams.append(candidate_ngram)
		print(r_ngrams)


freqs_folder = "./freqs/"
test_str = "what freaking ever, nobody cares one bit of a heck"

detector = da_detector()
detector.pick_ngrams(
	what_grams=3,
	how_many=10,
	text=test_str)
detector.pick_ngrams(
	what_grams=2,
	how_many=10,
	text=test_str)

sk_json = detector.parse_freqs("sk.json")
cz_json = detector.parse_freqs("cz.json")
de_json = detector.parse_freqs("de.json")
en_json = detector.parse_freqs("en.json")
fr_json = detector.parse_freqs("fr.json")

# vim: ff=unix noexpandtab