Giguru Scheuer commited on
Commit
445f168
·
1 Parent(s): a50bb4c

Added download sizes and implemented generate_examples for 'test_collection'

Browse files
Files changed (1) hide show
  1. trec-cast-2019-multi-turn.py +25 -10
trec-cast-2019-multi-turn.py CHANGED
@@ -49,17 +49,13 @@ _URLs = {
49
  'topics': _URL+"cast2019_test_annotated.tsv",
50
  'qrels': _URL+"2019qrels.txt",
51
  'test_collection': {
52
- 'msmarco': 'https://msmarco.blob.core.windows.net/msmarcoranking/collection.tar.gz',
53
  'car': "http://trec-car.cs.unh.edu/datareleases/v2.0/paragraphCorpus.v2.0.tar.xz",
 
54
  }
55
  }
56
 
57
 
58
  class TrecCast2019MultiTurn(datasets.GeneratorBasedBuilder):
59
- """
60
-
61
- """
62
-
63
  VERSION = datasets.Version("1.0.0")
64
 
65
  # This is an example of a dataset with multiple configurations.
@@ -90,24 +86,29 @@ class TrecCast2019MultiTurn(datasets.GeneratorBasedBuilder):
90
 
91
  def _info(self):
92
  # This is the name of the configuration selected in BUILDER_CONFIGS above
 
93
  if self.config.name == "topics":
94
  features = datasets.Features({
95
  "qid": datasets.Value("string"),
96
  "history": datasets.features.Sequence(feature=datasets.Value('string')),
97
  "query": datasets.Value("string"),
98
  })
 
99
  elif self.config.name == "qrels":
100
  features = datasets.Features({
101
  "qid": datasets.Value("string"),
102
  "qrels": datasets.features.Sequence(feature=datasets.Features({
103
  'docno': datasets.Value("string"),
104
- 'rank': datasets.Value("string"),
105
  })),
106
  })
 
107
  elif self.config.name == 'test_collection':
108
  features = datasets.Features({
109
- "docid": datasets.Value("string"),
 
110
  })
 
111
  return datasets.DatasetInfo(
112
  # This is the description that will appear on the datasets page.
113
  description=_DESCRIPTION,
@@ -123,6 +124,7 @@ class TrecCast2019MultiTurn(datasets.GeneratorBasedBuilder):
123
  license=_LICENSE,
124
  # Citation for the dataset
125
  citation=_CITATION,
 
126
  )
127
 
128
  def _split_generators(self, dl_manager):
@@ -158,8 +160,8 @@ class TrecCast2019MultiTurn(datasets.GeneratorBasedBuilder):
158
  for row in qrels_file:
159
  qid = row[0]
160
  docno = row[2]
161
- rank = row[3]
162
- qrels[qid].append({'docno': docno, 'rank': rank})
163
 
164
  for qid in qrels.keys():
165
  yield qid, {'qid': qid, 'qrels': qrels[qid]}
@@ -178,6 +180,19 @@ class TrecCast2019MultiTurn(datasets.GeneratorBasedBuilder):
178
  query = queries[idx]
179
  qid = f"{conversation_id}_{str(idx+1)}"
180
  yield qid, ({'query': query, 'history': queries[:idx], 'qid': qid})
181
-
 
 
 
 
 
 
 
 
 
 
 
 
 
182
  else:
183
  raise NotImplementedError(f"'{split}' is not yet implemented")
 
49
  'topics': _URL+"cast2019_test_annotated.tsv",
50
  'qrels': _URL+"2019qrels.txt",
51
  'test_collection': {
 
52
  'car': "http://trec-car.cs.unh.edu/datareleases/v2.0/paragraphCorpus.v2.0.tar.xz",
53
+ 'msmarco': 'https://msmarco.blob.core.windows.net/msmarcoranking/collection.tar.gz',
54
  }
55
  }
56
 
57
 
58
  class TrecCast2019MultiTurn(datasets.GeneratorBasedBuilder):
 
 
 
 
59
  VERSION = datasets.Version("1.0.0")
60
 
61
  # This is an example of a dataset with multiple configurations.
 
86
 
87
  def _info(self):
88
  # This is the name of the configuration selected in BUILDER_CONFIGS above
89
+ download_size = None
90
  if self.config.name == "topics":
91
  features = datasets.Features({
92
  "qid": datasets.Value("string"),
93
  "history": datasets.features.Sequence(feature=datasets.Value('string')),
94
  "query": datasets.Value("string"),
95
  })
96
+ download_size = 1138032
97
  elif self.config.name == "qrels":
98
  features = datasets.Features({
99
  "qid": datasets.Value("string"),
100
  "qrels": datasets.features.Sequence(feature=datasets.Features({
101
  'docno': datasets.Value("string"),
102
+ 'relevance': datasets.Value("string"),
103
  })),
104
  })
105
+ download_size = 8010
106
  elif self.config.name == 'test_collection':
107
  features = datasets.Features({
108
+ "docno": datasets.Value("string"),
109
+ "text": datasets.Value("string"),
110
  })
111
+ download_size = 5085726092 + 1035009698
112
  return datasets.DatasetInfo(
113
  # This is the description that will appear on the datasets page.
114
  description=_DESCRIPTION,
 
124
  license=_LICENSE,
125
  # Citation for the dataset
126
  citation=_CITATION,
127
+ download_size=download_size
128
  )
129
 
130
  def _split_generators(self, dl_manager):
 
160
  for row in qrels_file:
161
  qid = row[0]
162
  docno = row[2]
163
+ relevance = row[3]
164
+ qrels[qid].append({'docno': docno, 'relevance': relevance})
165
 
166
  for qid in qrels.keys():
167
  yield qid, {'qid': qid, 'qrels': qrels[qid]}
 
180
  query = queries[idx]
181
  qid = f"{conversation_id}_{str(idx+1)}"
182
  yield qid, ({'query': query, 'history': queries[:idx], 'qid': qid})
183
+ elif split == 'test_collection':
184
+ with open(file['msmarco']+"/collection.tsv") as f:
185
+ msmarco = csv.reader(f, delimiter="\t")
186
+ for line in msmarco:
187
+ docid, text = line
188
+ docid = f"MARCO_{docid}"
189
+ yield docid, ({"docno": docid, "text": text})
190
+
191
+ with open(file['car']+"/collection.tsv", 'rb') as f:
192
+ car = csv.reader(f, delimiter="\t")
193
+ for line in car:
194
+ docid, text = line.split(" ")
195
+ docid = f"CAR_{docid}"
196
+ yield docid, ({"docno": docid, "text": text})
197
  else:
198
  raise NotImplementedError(f"'{split}' is not yet implemented")