def load_wikIR(collection_path):
left = pd.read_csv(collection_path + '/training/queries.csv',
index_col='id_left')
right = pd.read_csv(collection_path + '/documents.csv',
index_col='id_right')
relation = pd.read_csv(collection_path + '/training/BM25.qrels.csv',
index_col=0)
train_raw = mz.DataPack(left=left, right=right, relation=relation)
left = pd.read_csv(collection_path + '/validation/queries.csv',
index_col='id_left')
right = pd.read_csv(collection_path + '/documents.csv',
index_col='id_right')
relation = pd.read_csv(collection_path + '/validation/BM25.qrels.csv',
index_col=0)
validation_raw = mz.DataPack(left=left, right=right, relation=relation)
left = pd.read_csv(collection_path + '/test/queries.csv',
index_col='id_left')
right = pd.read_csv(collection_path + '/documents.csv',
index_col='id_right')
relation = pd.read_csv(collection_path + '/test/BM25.qrels.csv',
index_col=0)
test_raw = mz.DataPack(left=left, right=right, relation=relation)
return train_raw, validation_raw, test_raw
def pack(df: pd.DataFrame, selected_columns_left: List[str],
selected_columns_right: List[str]) -> 'matchzoo.DataPack':
"""
This is a customized class to avoid overriding code of Learning to Rank project
Pack a :class:`DataPack` using `df`.
The `df` must have `text_left` and `text_right` columns. Optionally,
the `df` can have `id_left`, `id_right` to index `text_left` and
`text_right` respectively. `id_left`, `id_right` will be automatically
generated if not specified.
:param df: Input :class:`pandas.DataFrame` to use.
:param selected_columns_left: `List[str]` a list of columns selected for building data pack
:param selected_columns_right: `List[str]` a list of columns selected for building data pack
Examples::
>>> import matchzoo as mz
>>> import pandas as pd
>>> df = pd.DataFrame(data={'text_left': list('AABC'),
... 'text_right': list('abbc'),
... 'label': [0, 1, 1, 0]})
>>> mz.pack(df).frame()
id_left text_left id_right text_right label
0 L-0 A R-0 a 0
1 L-0 A R-1 b 1
2 L-1 B R-1 b 1
3 L-2 C R-2 c 0
"""
if 'text_left' not in df or 'text_right' not in df:
raise ValueError(
'Input data frame must have `text_left` and `text_right`.')
# Gather IDs
if 'id_left' not in df:
id_left = _gen_ids(df, 'text_left', 'L-')
else:
id_left = df['id_left']
if 'id_right' not in df:
id_right = _gen_ids(df, 'text_right', 'R-')
else:
id_right = df['id_right']
# Build Relation
relation = pd.DataFrame(data={'id_left': id_left, 'id_right': id_right})
for col in df:
if col not in ['id_left', 'id_right', 'text_left', 'text_right']:
relation[col] = df[col]
# Build Left and Right
left = _merge(df, id_left, selected_columns_left)
right = _merge(df, id_right, selected_columns_right)
return matchzoo.DataPack(relation, left, right)
def pack(
df: pd.DataFrame,
task: typing.Union[str, BaseTask] = 'ranking',
) -> 'matchzoo.DataPack':
"""
Pack a :class:`DataPack` using `df`.
The `df` must have `text_left` and `text_right` columns. Optionally,
the `df` can have `id_left`, `id_right` to index `text_left` and
`text_right` respectively. `id_left`, `id_right` will be automatically
generated if not specified.
:param df: Input :class:`pandas.DataFrame` to use.
:param task: Could be one of `ranking`, `classification` or a
:class:`matchzoo.engine.BaseTask` instance.
Examples::
>>> import matchzoo as mz
>>> import pandas as pd
>>> df = pd.DataFrame(data={'text_left': list('AABC'),
... 'text_right': list('abbc'),
... 'label': [0, 1, 1, 0]})
>>> mz.pack(df, task='classification').frame()
id_left text_left id_right text_right label
0 L-0 A R-0 a 0
1 L-0 A R-1 b 1
2 L-1 B R-1 b 1
3 L-2 C R-2 c 0
>>> mz.pack(df, task='ranking').frame()
id_left text_left id_right text_right label
0 L-0 A R-0 a 0.0
1 L-0 A R-1 b 1.0
2 L-1 B R-1 b 1.0
3 L-2 C R-2 c 0.0
"""
if 'text_left' not in df or 'text_right' not in df:
raise ValueError(
'Input data frame must have `text_left` and `text_right`.')
# Gather IDs
if 'id_left' not in df:
id_left = _gen_ids(df, 'text_left', 'L-')
else:
id_left = df['id_left']
if 'id_right' not in df:
id_right = _gen_ids(df, 'text_right', 'R-')
else:
id_right = df['id_right']
# Build Relation
relation = pd.DataFrame(data={'id_left': id_left, 'id_right': id_right})
for col in df:
if col not in ['id_left', 'id_right', 'text_left', 'text_right']:
relation[col] = df[col]
if 'label' in relation:
if task == 'classification' or isinstance(
task, matchzoo.tasks.Classification):
relation['label'] = relation['label'].astype(int)
elif task == 'ranking' or isinstance(task, matchzoo.tasks.Ranking):
relation['label'] = relation['label'].astype(float)
else:
raise ValueError(f"{task} is not a valid task.")
# Build Left and Right
left = _merge(df, id_left, 'text_left', 'id_left')
right = _merge(df, id_right, 'text_right', 'id_right')
return matchzoo.DataPack(relation, left, right)
left_train = pd.merge(train_id_left, left, how="inner", on="id_left")[["id_left", "text_left"]]
left_train.set_index("id_left", inplace=True)
left_vali = pd.merge(vali_id_left, left, how="inner", on="id_left")[["id_left", "text_left"]]
left_vali.set_index("id_left", inplace=True)
left_test = pd.merge(test_id_left, left, how="inner", on="id_left")[["id_left", "text_left"]]
left_test.set_index("id_left", inplace=True)
right_train = pd.merge(relation_train.id_right.drop_duplicates(), right, how="inner", on="id_right")[["id_right", "text_right"]].drop_duplicates()
right_train.set_index("id_right", inplace=True)
right_vali = pd.merge(relation_vali.id_right.drop_duplicates(), right, how="inner", on="id_right")[["id_right", "text_right"]].drop_duplicates()
right_vali.set_index("id_right", inplace=True)
right_test = pd.merge(relation_test.id_right.drop_duplicates(), right, how="inner", on="id_right")[["id_right", "text_right"]].drop_duplicates()
right_test.set_index("id_right", inplace=True)
print('data loading ...')
train_pack_raw = mz.DataPack(relation=relation_train,left=left_train,right=right_train)
dev_pack_raw = mz.DataPack(relation=relation_vali,left=left_vali,right=right_vali)
test_pack_raw = mz.DataPack(relation=relation_test,left=left_test,right=right_test)
print('data loaded as `train_pack_raw` `dev_pack_raw` `test_pack_raw`')
ranking_task = mz.tasks.Ranking(losses=mz.losses.RankHingeLoss())
#ranking_task = mz.tasks.Ranking(losses=mz.losses.RankCrossEntropyLoss())
ranking_task.metrics = [
mz.metrics.NormalizedDiscountedCumulativeGain(k=3),
mz.metrics.NormalizedDiscountedCumulativeGain(k=5),
mz.metrics.MeanAveragePrecision()
]
print("`ranking_task` initialized with metrics", ranking_task.metrics)
preprocessor = mz.preprocessors.BasicPreprocessor(
truncated_length_left = 10,