1
|
|
|
from __future__ import annotations |
2
|
|
|
|
3
|
|
|
import os |
4
|
|
|
from typing import TYPE_CHECKING, Any |
5
|
|
|
|
6
|
|
|
from stwfsapy.predictor import StwfsapyPredictor |
7
|
|
|
|
8
|
|
|
from annif.exception import NotInitializedException, NotSupportedException |
9
|
|
|
from annif.suggestion import SubjectSuggestion |
10
|
|
|
from annif.util import atomic_save, boolean |
11
|
|
|
|
12
|
|
|
from . import backend |
13
|
|
|
|
14
|
|
|
if TYPE_CHECKING: |
15
|
|
|
from annif.corpus.document import DocumentCorpus |
16
|
|
|
|
17
|
|
|
_KEY_CONCEPT_TYPE_URI = "concept_type_uri" |
18
|
|
|
_KEY_SUBTHESAURUS_TYPE_URI = "sub_thesaurus_type_uri" |
19
|
|
|
_KEY_THESAURUS_RELATION_TYPE_URI = "thesaurus_relation_type_uri" |
20
|
|
|
_KEY_THESAURUS_RELATION_IS_SPECIALISATION = "thesaurus_relation_is_specialisation" |
21
|
|
|
_KEY_REMOVE_DEPRECATED = "remove_deprecated" |
22
|
|
|
_KEY_HANDLE_TITLE_CASE = "handle_title_case" |
23
|
|
|
_KEY_EXTRACT_UPPER_CASE_FROM_BRACES = "extract_upper_case_from_braces" |
24
|
|
|
_KEY_EXTRACT_ANY_CASE_FROM_BRACES = "extract_any_case_from_braces" |
25
|
|
|
_KEY_EXPAND_AMPERSAND_WITH_SPACES = "expand_ampersand_with_spaces" |
26
|
|
|
_KEY_EXPAND_ABBREVIATION_WITH_PUNCTUATION = "expand_abbreviation_with_punctuation" |
27
|
|
|
_KEY_SIMPLE_ENGLISH_PLURAL_RULES = "simple_english_plural_rules" |
28
|
|
|
_KEY_USE_TXT_VEC = "use_txt_vec" |
29
|
|
|
|
30
|
|
|
|
31
|
|
|
class StwfsaBackend(backend.AnnifBackend): |
32
|
|
|
name = "stwfsa" |
33
|
|
|
|
34
|
|
|
STWFSA_PARAMETERS = { |
35
|
|
|
_KEY_CONCEPT_TYPE_URI: str, |
36
|
|
|
_KEY_SUBTHESAURUS_TYPE_URI: str, |
37
|
|
|
_KEY_THESAURUS_RELATION_TYPE_URI: str, |
38
|
|
|
_KEY_THESAURUS_RELATION_IS_SPECIALISATION: boolean, |
39
|
|
|
_KEY_REMOVE_DEPRECATED: boolean, |
40
|
|
|
_KEY_HANDLE_TITLE_CASE: boolean, |
41
|
|
|
_KEY_EXTRACT_UPPER_CASE_FROM_BRACES: boolean, |
42
|
|
|
_KEY_EXTRACT_ANY_CASE_FROM_BRACES: boolean, |
43
|
|
|
_KEY_EXPAND_AMPERSAND_WITH_SPACES: boolean, |
44
|
|
|
_KEY_EXPAND_ABBREVIATION_WITH_PUNCTUATION: boolean, |
45
|
|
|
_KEY_SIMPLE_ENGLISH_PLURAL_RULES: boolean, |
46
|
|
|
_KEY_USE_TXT_VEC: bool, |
47
|
|
|
} |
48
|
|
|
|
49
|
|
|
DEFAULT_PARAMETERS = { |
50
|
|
|
_KEY_CONCEPT_TYPE_URI: "http://www.w3.org/2004/02/skos/core#Concept", |
51
|
|
|
_KEY_SUBTHESAURUS_TYPE_URI: "http://www.w3.org/2004/02/skos/core#Collection", |
52
|
|
|
_KEY_THESAURUS_RELATION_TYPE_URI: "http://www.w3.org/2004/02/skos/core#member", |
53
|
|
|
_KEY_THESAURUS_RELATION_IS_SPECIALISATION: True, |
54
|
|
|
_KEY_REMOVE_DEPRECATED: True, |
55
|
|
|
_KEY_HANDLE_TITLE_CASE: True, |
56
|
|
|
_KEY_EXTRACT_UPPER_CASE_FROM_BRACES: True, |
57
|
|
|
_KEY_EXTRACT_ANY_CASE_FROM_BRACES: False, |
58
|
|
|
_KEY_EXPAND_AMPERSAND_WITH_SPACES: True, |
59
|
|
|
_KEY_EXPAND_ABBREVIATION_WITH_PUNCTUATION: True, |
60
|
|
|
_KEY_SIMPLE_ENGLISH_PLURAL_RULES: False, |
61
|
|
|
_KEY_USE_TXT_VEC: False, |
62
|
|
|
} |
63
|
|
|
|
64
|
|
|
MODEL_FILE = "stwfsa_predictor.zip" |
65
|
|
|
|
66
|
|
|
_model = None |
67
|
|
|
|
68
|
|
|
def initialize(self, parallel: bool = False) -> None: |
69
|
|
|
if self._model is None: |
70
|
|
|
path = os.path.join(self.datadir, self.MODEL_FILE) |
71
|
|
|
self.debug(f"Loading STWFSA model from {path}.") |
72
|
|
|
if os.path.exists(path): |
73
|
|
|
self._model = StwfsapyPredictor.load(path) |
74
|
|
|
self.debug("Loaded model.") |
75
|
|
|
else: |
76
|
|
|
raise NotInitializedException( |
77
|
|
|
f"Model not found at {path}", backend_id=self.backend_id |
78
|
|
|
) |
79
|
|
|
|
80
|
|
|
def _load_data(self, corpus: DocumentCorpus) -> tuple[list[str], list[list[str]]]: |
81
|
|
|
if corpus == "cached": |
82
|
|
|
raise NotSupportedException( |
83
|
|
|
"Training stwfsa project from cached data not supported." |
84
|
|
|
) |
85
|
|
|
if corpus.is_empty(): |
86
|
|
|
raise NotSupportedException( |
87
|
|
|
"Cannot train stwfsa project with no documents." |
88
|
|
|
) |
89
|
|
|
self.debug("Transforming training data.") |
90
|
|
|
X = [] |
91
|
|
|
y = [] |
92
|
|
|
for doc in corpus.documents: |
93
|
|
|
X.append(doc.text) |
94
|
|
|
y.append( |
95
|
|
|
[ |
96
|
|
|
self.project.subjects[subject_id].uri |
97
|
|
|
for subject_id in doc.subject_set |
98
|
|
|
] |
99
|
|
|
) |
100
|
|
|
return X, y |
101
|
|
|
|
102
|
|
|
def _train( |
103
|
|
|
self, |
104
|
|
|
corpus: DocumentCorpus, |
105
|
|
|
params: dict[str, Any], |
106
|
|
|
jobs: int = 0, |
107
|
|
|
) -> None: |
108
|
|
|
X, y = self._load_data(corpus) |
109
|
|
|
new_params = { |
110
|
|
|
key: self.STWFSA_PARAMETERS[key](val) |
111
|
|
|
for key, val in params.items() |
112
|
|
|
if key in self.STWFSA_PARAMETERS |
113
|
|
|
} |
114
|
|
|
p = StwfsapyPredictor( |
115
|
|
|
graph=self.project.vocab.as_graph(), |
116
|
|
|
langs=frozenset([params["language"]]), |
117
|
|
|
**new_params, |
118
|
|
|
) |
119
|
|
|
p.fit(X, y) |
120
|
|
|
self._model = p |
121
|
|
|
atomic_save( |
122
|
|
|
p, |
123
|
|
|
self.datadir, |
124
|
|
|
self.MODEL_FILE, |
125
|
|
|
lambda model, store_path: model.store(store_path), |
126
|
|
|
) |
127
|
|
|
|
128
|
|
|
def _suggest(self, text: str, params: dict[str, Any]) -> list[SubjectSuggestion]: |
129
|
|
|
self.debug(f'Suggesting subjects for text "{text[:20]}..." (len={len(text)})') |
130
|
|
|
result = self._model.suggest_proba([text])[0] |
131
|
|
|
suggestions = [] |
132
|
|
|
for uri, score in result: |
133
|
|
|
subject_id = self.project.subjects.by_uri(uri) |
134
|
|
|
if subject_id is not None: |
135
|
|
|
suggestions.append( |
136
|
|
|
SubjectSuggestion(subject_id=subject_id, score=score) |
137
|
|
|
) |
138
|
|
|
return suggestions |
139
|
|
|
|