Aligning "no-QA sentences" denotation to QANom - records for sentences with no QAs would have both 'question' and 'answers' as empty lists
Browse files- qa_discourse.py +10 -3
qa_discourse.py
CHANGED
@@ -61,7 +61,7 @@ COLUMNS = ['qasrl_id', 'sentence', 'worker_id', 'full_question', 'full_answer',
|
|
61 |
class QaDiscourse(datasets.GeneratorBasedBuilder):
|
62 |
"""QA-Discourse: Discourse Relations as Question-Answer Pairs. """
|
63 |
|
64 |
-
VERSION = datasets.Version("1.0.
|
65 |
|
66 |
BUILDER_CONFIGS = [
|
67 |
datasets.BuilderConfig(
|
@@ -135,7 +135,10 @@ class QaDiscourse(datasets.GeneratorBasedBuilder):
|
|
135 |
|
136 |
def _generate_examples(self, filepaths: List[str]):
|
137 |
|
138 |
-
"""
|
|
|
|
|
|
|
139 |
|
140 |
# merge annotations from sections
|
141 |
df = pd.concat([pd.read_csv(fn, sep='\t', error_bad_lines=False) for fn in filepaths]).reset_index(drop=True)
|
@@ -143,10 +146,14 @@ class QaDiscourse(datasets.GeneratorBasedBuilder):
|
|
143 |
for counter, row in df.iterrows():
|
144 |
# Prepare question (3 "slots" and question mark)
|
145 |
question = [row.question_start, row.question_aux, row.question_body.rstrip('?'), '?']
|
|
|
|
|
|
|
|
|
146 |
|
147 |
yield counter, {
|
148 |
"sentence": row.sentence,
|
149 |
"sent_id": row.qasrl_id,
|
150 |
"question": question,
|
151 |
-
"answers":
|
152 |
}
|
|
|
61 |
class QaDiscourse(datasets.GeneratorBasedBuilder):
|
62 |
"""QA-Discourse: Discourse Relations as Question-Answer Pairs. """
|
63 |
|
64 |
+
VERSION = datasets.Version("1.0.2")
|
65 |
|
66 |
BUILDER_CONFIGS = [
|
67 |
datasets.BuilderConfig(
|
|
|
135 |
|
136 |
def _generate_examples(self, filepaths: List[str]):
|
137 |
|
138 |
+
"""
|
139 |
+
Yields QA-Discourse examples from a tsv file.
|
140 |
+
Sentences with no QAs will yield an ``empty QA'' record, where both 'question' and 'answers' are empty lists.
|
141 |
+
"""
|
142 |
|
143 |
# merge annotations from sections
|
144 |
df = pd.concat([pd.read_csv(fn, sep='\t', error_bad_lines=False) for fn in filepaths]).reset_index(drop=True)
|
|
|
146 |
for counter, row in df.iterrows():
|
147 |
# Prepare question (3 "slots" and question mark)
|
148 |
question = [row.question_start, row.question_aux, row.question_body.rstrip('?'), '?']
|
149 |
+
answer = [row.answer]
|
150 |
+
if row.question_start == "_": # sentence has no QAs
|
151 |
+
question = []
|
152 |
+
answer = []
|
153 |
|
154 |
yield counter, {
|
155 |
"sentence": row.sentence,
|
156 |
"sent_id": row.qasrl_id,
|
157 |
"question": question,
|
158 |
+
"answers": answer,
|
159 |
}
|