-
Notifications
You must be signed in to change notification settings - Fork 0
/
competensor.py
308 lines (255 loc) · 10.2 KB
/
competensor.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
from io import StringIO
import datetime
import pandas as pd
from itertools import repeat
from joblib import load
from json import loads, dumps
from nameko.dependency_providers import Config
from nameko.rpc import rpc
from nameko_redis import Redis
import numpy as np
from uuid import NAMESPACE_DNS, uuid5
from utils.embeddings import UniversalSentenceEncoder\
as _UniversalSentenceEncoder
from utils.db import JDXDatabase
from nameko.extensions import DependencyProvider
from services.decomposition import Sentencizer
import web_pdb
class SetupFrameworks(DependencyProvider):
def __init__(self):
pass
def start(self):
setup = BuildCompetensorFrameworks()
setup.with_providers_and_store_frameworks()
setup.and_convert_frameworks_to_embeddings()
class GetRawText(DependencyProvider):
def __init__(self,
tablename=None,
uri=None):
self.uri = uri
self.tablename = tablename
def setup(self):
self.uri = self.uri or self.container.config['JDX_DB_URI']
self.jdx_database = JDXDatabase(uri=self.uri)
def get_dependency(self, worker_ctx):
#return self.jdx_database.get_raw_text_from_pipeline
return self.jdx_database
class Decomposer(DependencyProvider):
def get_dependency(self, worker_ctx):
return Sentencizer()
class UniversalSentenceEncoder(DependencyProvider):
def __init__(self,
module_url="./tensorflow_models/universal_sentence_encoder/",
encoder_type="DAN"):
self.module_url = module_url
self.encoder_type = encoder_type
def setup(self):
self.universal_sentence_encoder = _UniversalSentenceEncoder(
module_url=self.module_url,
encoder_type=self.encoder_type
)
def get_dependency(self, worker_ctx):
return self.universal_sentence_encoder
class Model(DependencyProvider):
def __init__(self,
model_name="linear_5_19.joblib",
model_path="./"):
self.model = None
self.model_name = model_name
if model_name:
self.model = load(
model_path+model_name
)
def get_model_name(self):
return self.model_name
def get_dependency(self, worker_ctx):
return self.model
class Competensor:
name = "competensor"
config = Config()
universal_sentence_encoder = UniversalSentenceEncoder()
model = Model()
get_raw_text = GetRawText()
frameworks = Redis("frameworks")
embeddings = Redis("embeddings")
match_table = Redis("match_table")
components = Decomposer()
def construct_sentences(self,
data_frame=None,
sentence_cols=["substatement"]):
competency_query =\
'jdx_property_a == "competency" | \
jdx_property_b == "competency" | \
jdx_property_c == "competency"'
return pd.Series(
data_frame.query(competency_query)[sentence_cols]
.values
.flatten()
)
def save_match_table(self, job_description_id, substatement_map, embedding_indices):
def get_match_table_row(substatement_map, embedding_indices):
for index, substatementID in zip(embedding_indices, substatement_map.keys()):
item = substatement_map[substatementID]
matches = item['matches']
for data in matches:
yield (substatementID,
data['recommendationID'],
data['name'],
data['definedTermSet'],
data['termCode'],
data['value'],
item['substatement'],
index)
# self.match_table.append(
# job_description_id,
# pd.DataFrame(
# data=get_match_table_row(substatement_map, embedding_indices)
# ).to_csv(index=False, header=False)
# )
self.match_table.set(
job_description_id,
pd.DataFrame(
data=get_match_table_row(substatement_map, embedding_indices)
).to_csv(index=False, header=False)
)
def get_all_sentences(self,
job_description_id):
sentences_df = pd.read_csv(
StringIO(
self.components.decompose(
self.get_raw_text.get_raw_text_from_pipeline(
job_description_id)))
)
return sentences_df
def get_all_sentence_embeddings(self,
job_description_id,
sentences_df):
sentence_embeddings =\
self.universal_sentence_encoder\
.embed(
sentences=self.construct_sentences(
data_frame=sentences_df)
)
return sentence_embeddings
def for_all_frameworks_find_matches_and_store(
self,
framework_names,
sentence_embeddings,
sentences_df,
threshold,
substatement_map,
job_description_id):
for framework_name in framework_names:
framework_df = pd.read_csv(
StringIO(
self.frameworks.get(
framework_name))
)
framework_embeddings = pd.read_csv(
StringIO(
self.embeddings.get(
framework_name))
).iloc[:, :-1].values
assert sentence_embeddings.shape[1] == framework_embeddings.shape[1]
similarities =\
np.inner(
sentence_embeddings,
framework_embeddings
)
lower_triangle = np.tril_indices(n=similarities.shape[0],
m=similarities.shape[1])
similarities[lower_triangle] = 0
sentence_and_framework_statement_indices =\
np.argwhere(similarities > threshold)
# values =\
# self.model.predict(
# similarities[
# similarities > threshold
# ].flatten()
# .reshape((-1, 1))
# )
values = repeat(-1, len(sentence_embeddings))
flattened_similarities =\
similarities[
similarities > threshold
].flatten()
if len(flattened_similarities):
reshaped_similarities =\
flattened_similarities.reshape((-1, 1))
values = \
self.model.predict(
reshaped_similarities
)
embedding_indices_for_save_match_table = []
for value, (sentence_idx, framework_idx) in zip(
values, sentence_and_framework_statement_indices):
substatementID = str(
uuid5(
NAMESPACE_DNS, sentences_df.loc[sentence_idx,"substatement"]))
item = substatement_map.get(substatementID, {"matches": []})
item["substatement"] = sentences_df.loc[sentence_idx, "substatement"]
if value >= threshold:
item["matches"].append(
{
"recommendationID":
framework_df.loc[framework_idx, "uuid"],
"name":
framework_df.loc[framework_idx, "framework_statement"],
"description":
"(framework term descriptions TBD later)",
"definedTermSet":
framework_name,
"termCode": "",
# framework_df.loc[framework_idx, "numeric_tag"],
"value":
value
}
)
substatement_map[substatementID] = item
embedding_indices_for_save_match_table.append(framework_idx)
self.save_match_table(job_description_id, substatement_map, embedding_indices_for_save_match_table)
match_table = [
{
"substatementID": key,
"substatement": value["substatement"],
"matches": value["matches"]
}
for key, value in substatement_map.items()
]
return match_table
def for_all_substatements_attach_matches(self,
match_table):
attached_matches = []
attached_matches.extend(match_table)
return attached_matches
@rpc
def get_match_table_and_jsonld(self,
job_description_id,
framework_names,
threshold):
if not threshold:
threshold = 0.44
#web_pdb.set_trace()
sentences_df = self.get_all_sentences(
job_description_id)
sentence_embeddings = self.get_all_sentence_embeddings(
job_description_id,
sentences_df)
substatement_map = {}
match_table =\
self.for_all_frameworks_find_matches_and_store(
job_description_id=job_description_id,
substatement_map=substatement_map,
sentences_df=sentences_df,
sentence_embeddings=sentence_embeddings,
framework_names=framework_names,
threshold=threshold
)
return {
"matchTable":
self.for_all_substatements_attach_matches(
match_table
),
"pipelineID": job_description_id,
"timestamp": str(datetime.datetime.utcnow())
}