|
1
|
|
|
"""Common functionality for analyzers.""" |
|
2
|
|
|
|
|
3
|
|
|
import abc |
|
4
|
|
|
import functools |
|
5
|
|
|
import unicodedata |
|
6
|
|
|
|
|
7
|
|
|
_KEY_TOKEN_MIN_LENGTH = 'token_min_length' |
|
8
|
|
|
|
|
9
|
|
|
|
|
10
|
|
View Code Duplication |
class Analyzer(metaclass=abc.ABCMeta): |
|
|
|
|
|
|
11
|
|
|
"""Base class for language-specific analyzers. Either tokenize_words or |
|
12
|
|
|
_normalize_word must be overridden in subclasses. Other methods may be |
|
13
|
|
|
overridden when necessary.""" |
|
14
|
|
|
|
|
15
|
|
|
name = None |
|
16
|
|
|
token_min_length = 3 # default value, can be overridden in instances |
|
17
|
|
|
|
|
18
|
|
|
def __init__(self, **kwargs): |
|
19
|
|
|
if _KEY_TOKEN_MIN_LENGTH in kwargs: |
|
20
|
|
|
self.token_min_length = int(kwargs[_KEY_TOKEN_MIN_LENGTH]) |
|
21
|
|
|
|
|
22
|
|
|
def tokenize_sentences(self, text): |
|
23
|
|
|
"""Tokenize a piece of text (e.g. a document) into sentences.""" |
|
24
|
|
|
import nltk.tokenize |
|
25
|
|
|
return nltk.tokenize.sent_tokenize(text) |
|
26
|
|
|
|
|
27
|
|
|
@functools.lru_cache(maxsize=50000) |
|
28
|
|
|
def is_valid_token(self, word): |
|
29
|
|
|
"""Return True if the word is an acceptable token.""" |
|
30
|
|
|
if len(word) < self.token_min_length: |
|
31
|
|
|
return False |
|
32
|
|
|
for char in word: |
|
33
|
|
|
category = unicodedata.category(char) |
|
34
|
|
|
if category[0] == 'L': # letter |
|
35
|
|
|
return True |
|
36
|
|
|
return False |
|
37
|
|
|
|
|
38
|
|
|
def tokenize_words(self, text, filter=True): |
|
39
|
|
|
"""Tokenize a piece of text (e.g. a sentence) into words. If |
|
40
|
|
|
filter=True (default), only return valid tokens (e.g. not |
|
41
|
|
|
punctuation, numbers or very short words)""" |
|
42
|
|
|
|
|
43
|
|
|
import nltk.tokenize |
|
44
|
|
|
return [self._normalize_word(word) |
|
45
|
|
|
for word in nltk.tokenize.word_tokenize(text) |
|
46
|
|
|
if (not filter or self.is_valid_token(word))] |
|
47
|
|
|
|
|
48
|
|
|
def _normalize_word(self, word): |
|
49
|
|
|
"""Normalize (stem or lemmatize) a word form into a normal form.""" |
|
50
|
|
|
pass # pragma: no cover |
|
51
|
|
|
|
NatLibFi /
Annif
Loading history...