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
|
# ---------------------------------------------------------
# Copyright (c) Microsoft Corporation. All rights reserved.
# ---------------------------------------------------------
import pytest
import os
import json
import pathlib
import random
from azure.ai.evaluation import evaluate
from azure.ai.evaluation._exceptions import EvaluationException
from azure.ai.evaluation._evaluators._document_retrieval import (
DocumentRetrievalEvaluator,
RetrievalGroundTruthDocument,
RetrievedDocument
)
def _get_file(name):
"""Get the file from the unittest data folder."""
data_path = os.path.join(pathlib.Path(__file__).parent.resolve(), "data")
return os.path.join(data_path, name)
@pytest.fixture()
def doc_retrieval_eval_data():
filename = _get_file("beir_document_retrieval_evaluation.jsonl")
records = []
with open(filename) as f:
records.extend(
[json.loads(x) for x in f.readlines()]
)
return filename, records
@pytest.fixture()
def bad_doc_retrieval_eval_data():
filename = _get_file("bad_input_document_retrieval_evaluation.jsonl")
records = []
with open(filename) as f:
records.extend(
[json.loads(x) for x in f.readlines()]
)
return filename, records
def test_success(doc_retrieval_eval_data):
_, records = doc_retrieval_eval_data
evaluator = DocumentRetrievalEvaluator(ground_truth_label_min=0, ground_truth_label_max=3)
for record in records:
result = evaluator(**record)
assert isinstance(result, dict)
assert "ndcg@3" in result
def test_groundtruth_min_gte_max():
expected_exception_msg = "The ground truth label maximum must be strictly greater than the ground truth label minimum."
with pytest.raises(EvaluationException) as exc_info:
DocumentRetrievalEvaluator(ground_truth_label_min=2, ground_truth_label_max=1)
assert expected_exception_msg in str(
exc_info._excinfo[1]
)
with pytest.raises(EvaluationException) as exc_info:
DocumentRetrievalEvaluator(ground_truth_label_max=0, ground_truth_label_min=0)
assert expected_exception_msg in str(
exc_info._excinfo[1]
)
def test_incorrect_groundtruth_min():
expected_exception_msg = ("A query relevance label less than the configured minimum value was detected in the evaluation input data. "
"Check the range of ground truth label values in the input data and set the value of ground_truth_label_min to "
"the appropriate value for your data.")
data_groundtruth_min = 0
configured_groundtruth_min = 1
groundtruth_docs = [
RetrievalGroundTruthDocument({"document_id": f"doc_{x}", "query_relevance_label": x}) for x in range(data_groundtruth_min, 5)
]
retrieved_docs = [
RetrievedDocument({"document_id": f"doc_{x}", "relevance_score": random.uniform(-10, 10)}) for x in range(data_groundtruth_min, 5)
]
evaluator = DocumentRetrievalEvaluator(ground_truth_label_min=configured_groundtruth_min, ground_truth_label_max=4)
with pytest.raises(EvaluationException) as exc_info:
evaluator(retrieval_ground_truth=groundtruth_docs, retrieved_documents=retrieved_docs)
assert expected_exception_msg in str(
exc_info._excinfo[1]
)
def test_incorrect_groundtruth_max():
expected_exception_msg = ("A query relevance label greater than the configured maximum value was detected in the evaluation input data. "
"Check the range of ground truth label values in the input data and set the value of ground_truth_label_max to "
"the appropriate value for your data.")
data_groundtruth_max = 5
configured_groundtruth_max = 4
groundtruth_docs = [
RetrievalGroundTruthDocument({"document_id": f"doc_{x}", "query_relevance_label": x}) for x in range(0, data_groundtruth_max + 1)
]
retrieved_docs = [
RetrievedDocument({"document_id": f"doc_{x}", "relevance_score": random.uniform(-10, 10)}) for x in range(0, data_groundtruth_max + 1)
]
evaluator = DocumentRetrievalEvaluator(ground_truth_label_min=0, ground_truth_label_max=configured_groundtruth_max)
with pytest.raises(EvaluationException) as exc_info:
evaluator(retrieval_ground_truth=groundtruth_docs, retrieved_documents=retrieved_docs)
assert expected_exception_msg in str(
exc_info._excinfo[1]
)
def test_thresholds(doc_retrieval_eval_data):
_, records = doc_retrieval_eval_data
record = records[-1]
custom_threshold_subset = {
"ndcg_threshold": 0.7,
"xdcg_threshold": 0.7,
"fidelity_threshold": 0.7,
}
custom_threshold_superset = {
"ndcg_threshold": 0.7,
"xdcg_threshold": 0.7,
"fidelity_threshold": 0.7,
"top1_relevance_threshold": 70,
"top3_max_relevance_threshold": 70,
"total_retrieved_documents_threshold": 10,
"total_ground_truth_documents_threshold": 10
}
for threshold in [custom_threshold_subset, custom_threshold_superset]:
evaluator = DocumentRetrievalEvaluator(ground_truth_label_min=0, ground_truth_label_max=2, **threshold)
results = evaluator(**record)
expected_keys = [
"ndcg@3", "ndcg@3_result", "ndcg@3_threshold", "ndcg@3_higher_is_better",
"xdcg@3", "xdcg@3_result", "xdcg@3_threshold", "xdcg@3_higher_is_better",
"fidelity", "fidelity_result", "fidelity_threshold", "fidelity_higher_is_better",
"top1_relevance", "top1_relevance_result", "top1_relevance_threshold", "top1_relevance_higher_is_better",
"top3_max_relevance", "top3_max_relevance_result", "top3_max_relevance_threshold", "top3_max_relevance_higher_is_better",
"total_retrieved_documents", "total_retrieved_documents_result", "total_retrieved_documents_threshold", "total_retrieved_documents_higher_is_better",
"total_ground_truth_documents", "total_ground_truth_documents_result", "total_ground_truth_documents_threshold", "total_ground_truth_documents_higher_is_better",
"holes", "holes_result", "holes_threshold", "holes_higher_is_better",
"holes_ratio", "holes_ratio_result", "holes_ratio_threshold", "holes_ratio_higher_is_better"
]
assert set(expected_keys) == set(results.keys())
def test_invalid_input(bad_doc_retrieval_eval_data):
filename, records = bad_doc_retrieval_eval_data
for record in records:
expected_exception_msg = record.pop("expected_exception")
with pytest.raises(EvaluationException) as exc_info:
evaluator = DocumentRetrievalEvaluator(ground_truth_label_min=0, ground_truth_label_max=2)
evaluator(**record)
assert expected_exception_msg in str(
exc_info._excinfo[1]
)
def test_qrels_results_limit():
groundtruth_docs = [
RetrievalGroundTruthDocument({"document_id": f"doc_{x}", "query_relevance_label": random.choice([0, 1, 2, 3, 4])}) for x in range(0, 10000)
]
retrieved_docs = [
RetrievedDocument({"document_id": f"doc_{x}", "relevance_score": random.uniform(-10, 10)}) for x in range(0, 10000)
]
evaluator = DocumentRetrievalEvaluator()
evaluator(retrieval_ground_truth=groundtruth_docs, retrieved_documents=retrieved_docs)
def test_qrels_results_exceeds_max_allowed():
expected_exception_msg = "'retrieval_ground_truth' and 'retrieved_documents' inputs should contain no more than 10000 items."
groundtruth_docs = [
RetrievalGroundTruthDocument({"document_id": f"doc_{x}", "query_relevance_label": random.choice([0, 1, 2, 3, 4])}) for x in range(0, 10001)
]
retrieved_docs = [
RetrievedDocument({"document_id": f"doc_{x}", "relevance_score": random.uniform(-10, 10)}) for x in range(0, 10001)
]
evaluator = DocumentRetrievalEvaluator()
with pytest.raises(EvaluationException) as exc_info:
evaluator(retrieval_ground_truth=groundtruth_docs, retrieved_documents=retrieved_docs)
assert expected_exception_msg in str(
exc_info._excinfo[1]
)
def test_no_retrieved_documents():
groundtruth_docs = [
RetrievalGroundTruthDocument({"document_id": f"doc_{x}", "query_relevance_label": random.choice([0, 1, 2, 3, 4])}) for x in range(0, 9)
]
retrieved_docs = []
evaluator = DocumentRetrievalEvaluator()
result = evaluator(retrieval_ground_truth=groundtruth_docs, retrieved_documents=retrieved_docs)
assert result["ndcg@3"] == 0
assert result["holes"] == 0
def test_no_labeled_retrieved_documents():
groundtruth_docs = [
RetrievalGroundTruthDocument({"document_id": f"doc_{x}", "query_relevance_label": random.choice([0, 1, 2, 3, 4])}) for x in range(0, 9)
]
retrieved_docs = [
RetrievedDocument({"document_id": f"doc_{x}_nolabel", "relevance_score": random.uniform(-10, 10)}) for x in range(0, 9)
]
evaluator = DocumentRetrievalEvaluator()
result = evaluator(retrieval_ground_truth=groundtruth_docs, retrieved_documents=retrieved_docs)
assert result["ndcg@3"] == 0
assert result["holes"] == len(retrieved_docs)
|
