|
1
|
|
|
"""Classes for supporting vocabulary files in SKOS/RDF format""" |
|
2
|
|
|
|
|
3
|
|
|
from __future__ import annotations |
|
4
|
|
|
|
|
5
|
|
|
import collections |
|
6
|
|
|
import os.path |
|
7
|
|
|
import shutil |
|
8
|
|
|
from typing import TYPE_CHECKING |
|
9
|
|
|
|
|
10
|
|
|
import rdflib |
|
11
|
|
|
import rdflib.util |
|
12
|
|
|
from rdflib.namespace import OWL, RDF, RDFS, SKOS |
|
13
|
|
|
|
|
14
|
|
|
import annif.util |
|
15
|
|
|
|
|
16
|
|
|
from .types import Subject, VocabSource |
|
17
|
|
|
|
|
18
|
|
|
if TYPE_CHECKING: |
|
19
|
|
|
from collections.abc import Iterator, Sequence |
|
20
|
|
|
|
|
21
|
|
|
from rdflib.term import URIRef |
|
22
|
|
|
|
|
23
|
|
|
|
|
24
|
|
|
def serialize_subjects_to_skos(subjects: Iterator, path: str) -> None: |
|
25
|
|
|
"""Create a SKOS representation of the given subjects and serialize it |
|
26
|
|
|
into a SKOS/Turtle file with the given path name.""" |
|
27
|
|
|
import joblib |
|
28
|
|
|
|
|
29
|
|
|
graph = rdflib.Graph() |
|
30
|
|
|
graph.namespace_manager.bind("skos", SKOS) |
|
31
|
|
|
for subject in subjects: |
|
32
|
|
|
graph.add((rdflib.URIRef(subject.uri), RDF.type, SKOS.Concept)) |
|
33
|
|
|
for lang, label in subject.labels.items(): |
|
34
|
|
|
graph.add( |
|
35
|
|
|
( |
|
36
|
|
|
rdflib.URIRef(subject.uri), |
|
37
|
|
|
SKOS.prefLabel, |
|
38
|
|
|
rdflib.Literal(label, lang), |
|
39
|
|
|
) |
|
40
|
|
|
) |
|
41
|
|
|
graph.add( |
|
42
|
|
|
( |
|
43
|
|
|
rdflib.URIRef(subject.uri), |
|
44
|
|
|
SKOS.notation, |
|
45
|
|
|
rdflib.Literal(subject.notation), |
|
46
|
|
|
) |
|
47
|
|
|
) |
|
48
|
|
|
graph.serialize(destination=path, format="turtle") |
|
49
|
|
|
# also dump the graph in joblib format which is faster to load |
|
50
|
|
|
annif.util.atomic_save( |
|
51
|
|
|
graph, *os.path.split(path.replace(".ttl", ".dump.gz")), method=joblib.dump |
|
52
|
|
|
) |
|
53
|
|
|
|
|
54
|
|
|
|
|
55
|
|
|
class VocabFileSKOS(VocabSource): |
|
56
|
|
|
"""A subject corpus that uses SKOS files""" |
|
57
|
|
|
|
|
58
|
|
|
PREF_LABEL_PROPERTIES = (SKOS.prefLabel, RDFS.label) |
|
59
|
|
|
|
|
60
|
|
|
_languages = None |
|
61
|
|
|
|
|
62
|
|
|
def __init__(self, path: str) -> None: |
|
63
|
|
|
self.path = path |
|
64
|
|
|
if path.endswith(".dump.gz"): |
|
65
|
|
|
import joblib |
|
66
|
|
|
|
|
67
|
|
|
self.graph = joblib.load(path) |
|
68
|
|
|
else: |
|
69
|
|
|
self.graph = rdflib.Graph() |
|
70
|
|
|
self.graph.parse(self.path, format=rdflib.util.guess_format(self.path)) |
|
71
|
|
|
|
|
72
|
|
|
@property |
|
73
|
|
|
def languages(self) -> set[str]: |
|
74
|
|
|
if self._languages is None: |
|
75
|
|
|
self._languages = { |
|
76
|
|
|
label.language |
|
77
|
|
|
for concept in self.concepts |
|
78
|
|
|
for label_type in self.PREF_LABEL_PROPERTIES |
|
79
|
|
|
for label in self.graph.objects(concept, label_type) |
|
80
|
|
|
if label.language is not None |
|
81
|
|
|
} |
|
82
|
|
|
return self._languages |
|
83
|
|
|
|
|
84
|
|
|
def _concept_labels(self, concept: URIRef) -> dict[str, str]: |
|
85
|
|
|
by_lang = self.get_concept_labels(concept, self.PREF_LABEL_PROPERTIES) |
|
86
|
|
|
return { |
|
87
|
|
|
lang: ( |
|
88
|
|
|
by_lang[lang][0] |
|
89
|
|
|
if by_lang[lang] # correct lang |
|
90
|
|
|
else ( |
|
91
|
|
|
by_lang[None][0] |
|
92
|
|
|
if by_lang[None] # no language |
|
93
|
|
|
else self.graph.namespace_manager.qname(concept) |
|
94
|
|
|
) |
|
95
|
|
|
) |
|
96
|
|
|
for lang in self.languages |
|
97
|
|
|
} |
|
98
|
|
|
|
|
99
|
|
|
@property |
|
100
|
|
|
def subjects(self) -> Iterator[Subject]: |
|
101
|
|
|
for concept in self.concepts: |
|
102
|
|
|
labels = self._concept_labels(concept) |
|
103
|
|
|
|
|
104
|
|
|
notation = self.graph.value(concept, SKOS.notation, None, any=True) |
|
105
|
|
|
if notation is not None: |
|
106
|
|
|
notation = str(notation) |
|
107
|
|
|
|
|
108
|
|
|
yield Subject(uri=str(concept), labels=labels, notation=notation) |
|
109
|
|
|
|
|
110
|
|
|
@property |
|
111
|
|
|
def concepts(self) -> Iterator[URIRef]: |
|
112
|
|
|
for concept in self.graph.subjects(RDF.type, SKOS.Concept): |
|
113
|
|
|
if (concept, OWL.deprecated, rdflib.Literal(True)) in self.graph: |
|
114
|
|
|
continue |
|
115
|
|
|
yield concept |
|
116
|
|
|
|
|
117
|
|
|
def get_concept_labels( |
|
118
|
|
|
self, |
|
119
|
|
|
concept: URIRef, |
|
120
|
|
|
label_types: Sequence[URIRef], |
|
121
|
|
|
) -> collections.defaultdict[str | None, list[str]]: |
|
122
|
|
|
"""return all the labels of the given concept with the given label |
|
123
|
|
|
properties as a dict-like object where the keys are language codes |
|
124
|
|
|
and the values are lists of labels in that language""" |
|
125
|
|
|
labels_by_lang = collections.defaultdict(list) |
|
126
|
|
|
|
|
127
|
|
|
for label_type in label_types: |
|
128
|
|
|
for label in self.graph.objects(concept, label_type): |
|
129
|
|
|
labels_by_lang[label.language].append(str(label)) |
|
130
|
|
|
|
|
131
|
|
|
return labels_by_lang |
|
132
|
|
|
|
|
133
|
|
|
@staticmethod |
|
134
|
|
|
def is_rdf_file(path: str) -> bool: |
|
135
|
|
|
"""return True if the path looks like an RDF file that can be loaded |
|
136
|
|
|
as SKOS""" |
|
137
|
|
|
|
|
138
|
|
|
fmt = rdflib.util.guess_format(path) |
|
139
|
|
|
return fmt is not None |
|
140
|
|
|
|
|
141
|
|
|
def save_skos(self, path: str) -> None: |
|
142
|
|
|
"""Save the contents of the subject vocabulary into a SKOS/Turtle |
|
143
|
|
|
file with the given path name.""" |
|
144
|
|
|
|
|
145
|
|
|
if self.path.endswith(".ttl"): |
|
146
|
|
|
# input is already in Turtle syntax, no need to reserialize |
|
147
|
|
|
if not os.path.exists(path) or not os.path.samefile(self.path, path): |
|
148
|
|
|
shutil.copyfile(self.path, path) |
|
149
|
|
|
else: |
|
150
|
|
|
# need to serialize into Turtle |
|
151
|
|
|
self.graph.serialize(destination=path, format="turtle") |
|
152
|
|
|
# also dump the graph in joblib format which is faster to load |
|
153
|
|
|
import joblib |
|
154
|
|
|
|
|
155
|
|
|
annif.util.atomic_save( |
|
156
|
|
|
self.graph, |
|
157
|
|
|
*os.path.split(path.replace(".ttl", ".dump.gz")), |
|
158
|
|
|
method=joblib.dump, |
|
159
|
|
|
) |
|
160
|
|
|
|
NatLibFi /
Annif
