|
1
|
|
|
"""Project management functionality for Annif""" |
|
2
|
|
|
|
|
3
|
|
|
import collections |
|
4
|
|
|
import configparser |
|
5
|
|
|
import logging |
|
|
|
|
|
|
6
|
|
|
from flask import current_app |
|
7
|
|
|
import annif |
|
8
|
|
|
import annif.hit |
|
9
|
|
|
import annif.backend |
|
10
|
|
|
from annif import logger |
|
11
|
|
|
|
|
12
|
|
|
|
|
13
|
|
|
class AnnifProject: |
|
|
|
|
|
|
14
|
|
|
"""Class representing the configuration of a single Annif project.""" |
|
15
|
|
|
|
|
16
|
|
|
def __init__(self, project_id, config): |
|
17
|
|
|
self.project_id = project_id |
|
18
|
|
|
self.language = config['language'] |
|
19
|
|
|
self.backends = self._initialize_backends(config['backends']) |
|
20
|
|
|
|
|
21
|
|
|
def _initialize_backends(self, backends_configuration): |
|
|
|
|
|
|
22
|
|
|
backends = [] |
|
23
|
|
|
for backenddef in backends_configuration.split(','): |
|
24
|
|
|
bedefs = backenddef.strip().split(':') |
|
25
|
|
|
backend_id = bedefs[0] |
|
26
|
|
|
if len(bedefs) > 1: |
|
27
|
|
|
weight = float(bedefs[1]) |
|
28
|
|
|
else: |
|
29
|
|
|
weight = 1.0 |
|
30
|
|
|
backend = annif.backend.get_backend(backend_id) |
|
31
|
|
|
backends.append((backend, weight)) |
|
32
|
|
|
return backends |
|
33
|
|
|
|
|
34
|
|
|
def _analyze_with_backends(self, text, backend_params): |
|
35
|
|
|
if backend_params is None: |
|
36
|
|
|
backend_params = {} |
|
37
|
|
|
hits_by_uri = collections.defaultdict(list) |
|
38
|
|
|
for backend, weight in self.backends: |
|
39
|
|
|
beparams = backend_params.get(backend.backend_id, {}) |
|
40
|
|
|
hits = [hit for hit in backend.analyze(text, params=beparams) |
|
41
|
|
|
if hit.score > 0.0] |
|
42
|
|
|
logger.debug( |
|
43
|
|
|
'Got {} hits from backend {}'.format( |
|
|
|
|
|
|
44
|
|
|
len(hits), backend.backend_id)) |
|
45
|
|
|
for hit in hits: |
|
46
|
|
|
hits_by_uri[hit.uri].append((hit.score * weight, hit)) |
|
47
|
|
|
return hits_by_uri |
|
48
|
|
|
|
|
49
|
|
|
@classmethod |
|
50
|
|
|
def _merge_hits(cls, hits_by_uri): |
|
51
|
|
|
merged_hits = [] |
|
52
|
|
|
for score_hits in hits_by_uri.values(): |
|
53
|
|
|
total = sum([sh[0] for sh in score_hits]) |
|
54
|
|
|
hit = annif.hit.AnalysisHit( |
|
55
|
|
|
score_hits[0][1].uri, score_hits[0][1].label, total) |
|
56
|
|
|
merged_hits.append(hit) |
|
57
|
|
|
return merged_hits |
|
58
|
|
|
|
|
59
|
|
|
@classmethod |
|
60
|
|
|
def _filter_hits(cls, hits, limit, threshold): |
|
61
|
|
|
hits.sort(key=lambda hit: hit.score, reverse=True) |
|
62
|
|
|
hits = hits[:limit] |
|
63
|
|
|
logger.debug( |
|
64
|
|
|
'{} hits after applying limit {}'.format( |
|
|
|
|
|
|
65
|
|
|
len(hits), limit)) |
|
66
|
|
|
hits = [hit for hit in hits if hit.score >= threshold] |
|
67
|
|
|
logger.debug( |
|
68
|
|
|
'{} hits after applying threshold {}'.format( |
|
|
|
|
|
|
69
|
|
|
len(hits), threshold)) |
|
70
|
|
|
return hits |
|
71
|
|
|
|
|
72
|
|
|
def analyze(self, text, limit=10, threshold=0.0, backend_params=None): |
|
73
|
|
|
"""Analyze the given text by passing it to backends and joining the |
|
74
|
|
|
results. Returns a list of AnalysisHit objects ordered by decreasing |
|
75
|
|
|
score. The limit parameter defines the maximum number of hits to return. |
|
76
|
|
|
Only hits whose score is over the threshold are returned.""" |
|
77
|
|
|
|
|
78
|
|
|
logger.debug('Analyzing text "{}..." (len={})'.format( |
|
|
|
|
|
|
79
|
|
|
text[:20], len(text))) |
|
80
|
|
|
hits_by_uri = self._analyze_with_backends(text, backend_params) |
|
81
|
|
|
merged_hits = self._merge_hits(hits_by_uri) |
|
82
|
|
|
logger.debug('{} hits after merging'.format(len(merged_hits))) |
|
|
|
|
|
|
83
|
|
|
return self._filter_hits(merged_hits, limit, threshold) |
|
84
|
|
|
|
|
85
|
|
|
def load_subjects(self, subjects): |
|
|
|
|
|
|
86
|
|
|
for backend, weight in self.backends: |
|
|
|
|
|
|
87
|
|
|
logger.debug( |
|
88
|
|
|
'Loading subjects for backend {}'.format( |
|
|
|
|
|
|
89
|
|
|
backend.backend_id)) |
|
90
|
|
|
backend.load_subjects(subjects) |
|
91
|
|
|
|
|
92
|
|
|
def dump(self): |
|
93
|
|
|
"""return this project as a dict""" |
|
94
|
|
|
return {'project_id': self.project_id, |
|
95
|
|
|
'language': self.language, |
|
96
|
|
|
'backends': [{'backend_id': be[0].backend_id, |
|
97
|
|
|
'weight': be[1]} for be in self.backends] |
|
98
|
|
|
} |
|
|
|
|
|
|
99
|
|
|
|
|
100
|
|
|
|
|
101
|
|
|
def get_projects(): |
|
102
|
|
|
"""return the available projects as a dict of project_id -> AnnifProject""" |
|
103
|
|
|
projects_file = current_app.config['PROJECTS_FILE'] |
|
104
|
|
|
config = configparser.ConfigParser() |
|
105
|
|
|
with open(projects_file) as projf: |
|
106
|
|
|
config.read_file(projf) |
|
107
|
|
|
|
|
108
|
|
|
# create AnnifProject objects from the configuration file |
|
109
|
|
|
projects = {} |
|
110
|
|
|
for project_id in config.sections(): |
|
111
|
|
|
projects[project_id] = AnnifProject(project_id, config[project_id]) |
|
112
|
|
|
return projects |
|
113
|
|
|
|
|
114
|
|
|
|
|
115
|
|
|
def get_project(project_id): |
|
116
|
|
|
"""return the definition of a single Project by project_id""" |
|
117
|
|
|
projects = get_projects() |
|
118
|
|
|
try: |
|
119
|
|
|
return projects[project_id] |
|
120
|
|
|
except KeyError: |
|
121
|
|
|
raise ValueError("No such project {}".format(project_id)) |
|
122
|
|
|
|
NatLibFi /
Annif
Loading history...