|
1
|
|
|
"""Clases for supporting document corpora""" |
|
2
|
|
|
|
|
3
|
|
|
import glob |
|
4
|
|
|
import os.path |
|
5
|
|
|
import re |
|
6
|
|
|
import gzip |
|
7
|
|
|
import annif.util |
|
8
|
|
|
from itertools import islice |
|
9
|
|
|
from .types import DocumentCorpus |
|
10
|
|
|
from .subject import SubjectSet |
|
11
|
|
|
|
|
12
|
|
|
logger = annif.logger |
|
13
|
|
|
|
|
14
|
|
|
|
|
15
|
|
|
class DocumentDirectory(DocumentCorpus): |
|
16
|
|
|
"""A directory of files as a full text document corpus""" |
|
17
|
|
|
|
|
18
|
|
|
def __init__(self, path, require_subjects=False): |
|
19
|
|
|
self.path = path |
|
20
|
|
|
self.require_subjects = require_subjects |
|
21
|
|
|
|
|
22
|
|
|
def __iter__(self): |
|
23
|
|
|
"""Iterate through the directory, yielding tuples of (docfile, |
|
24
|
|
|
subjectfile) containing file paths. If there is no key file and |
|
25
|
|
|
require_subjects is False, the subjectfile will be returned as None.""" |
|
26
|
|
|
|
|
27
|
|
|
for filename in sorted(glob.glob(os.path.join(self.path, '*.txt'))): |
|
28
|
|
|
tsvfilename = re.sub(r'\.txt$', '.tsv', filename) |
|
29
|
|
|
if os.path.exists(tsvfilename): |
|
30
|
|
|
yield (filename, tsvfilename) |
|
31
|
|
|
continue |
|
32
|
|
|
keyfilename = re.sub(r'\.txt$', '.key', filename) |
|
33
|
|
|
if os.path.exists(keyfilename): |
|
34
|
|
|
yield (filename, keyfilename) |
|
35
|
|
|
continue |
|
36
|
|
|
if not self.require_subjects: |
|
37
|
|
|
yield (filename, None) |
|
38
|
|
|
|
|
39
|
|
|
@property |
|
40
|
|
|
def documents(self): |
|
41
|
|
|
for docfilename, keyfilename in self: |
|
42
|
|
|
with open(docfilename, errors='replace', |
|
43
|
|
|
encoding='utf-8-sig') as docfile: |
|
44
|
|
|
text = docfile.read() |
|
45
|
|
|
with open(keyfilename, encoding='utf-8-sig') as keyfile: |
|
46
|
|
|
subjects = SubjectSet.from_string(keyfile.read()) |
|
47
|
|
|
yield self._create_document(text=text, |
|
48
|
|
|
uris=subjects.subject_uris, |
|
49
|
|
|
labels=subjects.subject_labels) |
|
50
|
|
|
|
|
51
|
|
|
|
|
52
|
|
|
class DocumentFile(DocumentCorpus): |
|
53
|
|
|
"""A TSV file as a corpus of documents with subjects""" |
|
54
|
|
|
|
|
55
|
|
|
def __init__(self, path): |
|
56
|
|
|
self.path = path |
|
57
|
|
|
|
|
58
|
|
|
@property |
|
59
|
|
|
def documents(self): |
|
60
|
|
|
if self.path.endswith('.gz'): |
|
61
|
|
|
opener = gzip.open |
|
62
|
|
|
else: |
|
63
|
|
|
opener = open |
|
64
|
|
|
with opener(self.path, mode='rt', encoding='utf-8-sig') as tsvfile: |
|
65
|
|
|
for line in tsvfile: |
|
66
|
|
|
yield from self._parse_tsv_line(line) |
|
67
|
|
|
|
|
68
|
|
|
def _parse_tsv_line(self, line): |
|
69
|
|
|
if '\t' in line: |
|
70
|
|
|
text, uris = line.split('\t', maxsplit=1) |
|
71
|
|
|
subjects = [annif.util.cleanup_uri(uri) |
|
72
|
|
|
for uri in uris.split()] |
|
73
|
|
|
yield self._create_document(text=text, |
|
74
|
|
|
uris=subjects, |
|
75
|
|
|
labels=[]) |
|
76
|
|
|
else: |
|
77
|
|
|
logger.warning('Skipping invalid line (missing tab): "%s"', |
|
78
|
|
|
line.rstrip()) |
|
79
|
|
|
|
|
80
|
|
|
|
|
81
|
|
|
class DocumentList(DocumentCorpus): |
|
82
|
|
|
"""A document corpus based on a list of other iterable of Document |
|
83
|
|
|
objects""" |
|
84
|
|
|
|
|
85
|
|
|
def __init__(self, documents): |
|
86
|
|
|
self._documents = documents |
|
87
|
|
|
|
|
88
|
|
|
@property |
|
89
|
|
|
def documents(self): |
|
90
|
|
|
yield from self._documents |
|
91
|
|
|
|
|
92
|
|
|
|
|
93
|
|
|
class TruncatingDocumentCorpus(DocumentCorpus): |
|
94
|
|
|
"""A document corpus that wraps another document corpus but truncates the |
|
95
|
|
|
documents to a given length""" |
|
96
|
|
|
|
|
97
|
|
|
def __init__(self, corpus, limit): |
|
98
|
|
|
self._orig_corpus = corpus |
|
99
|
|
|
self._limit = limit |
|
100
|
|
|
|
|
101
|
|
|
@property |
|
102
|
|
|
def documents(self): |
|
103
|
|
|
for doc in self._orig_corpus.documents: |
|
104
|
|
|
yield self._create_document(text=doc.text[:self._limit], |
|
105
|
|
|
uris=doc.uris, |
|
106
|
|
|
labels=doc.labels) |
|
107
|
|
|
|
|
108
|
|
|
|
|
109
|
|
|
class LimitingDocumentCorpus(DocumentCorpus): |
|
110
|
|
|
"""A document corpus that wraps another document corpus but limits the |
|
111
|
|
|
number of documents to a given limit""" |
|
112
|
|
|
|
|
113
|
|
|
def __init__(self, corpus, docs_limit): |
|
114
|
|
|
self._orig_corpus = corpus |
|
115
|
|
|
self.docs_limit = docs_limit |
|
116
|
|
|
|
|
117
|
|
|
@property |
|
118
|
|
|
def documents(self): |
|
119
|
|
|
for doc in islice(self._orig_corpus.documents, self.docs_limit): |
|
120
|
|
|
yield self._create_document(text=doc.text, |
|
121
|
|
|
uris=doc.uris, |
|
122
|
|
|
labels=doc.labels) |
|
123
|
|
|
|
NatLibFi /
Annif
