mirror of https://github.com/doccano/doccano.git
pythondatasetsactive-learningtext-annotationdatasetnatural-language-processingdata-labelingmachine-learningannotation-tool
You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
83 lines
2.9 KiB
83 lines
2.9 KiB
"""
|
|
Convert a dataset to the specified format.
|
|
"""
|
|
import abc
|
|
|
|
import pandas as pd
|
|
|
|
from data_export.models import DATA
|
|
|
|
|
|
class Formatter(abc.ABC):
|
|
def __init__(self, target_column: str = "labels", **kwargs):
|
|
self.target_column = target_column
|
|
self.mapper = kwargs
|
|
|
|
def format(self, dataset: pd.DataFrame) -> pd.DataFrame:
|
|
if self.target_column not in dataset.columns:
|
|
return dataset
|
|
return self.apply(dataset)
|
|
|
|
@abc.abstractmethod
|
|
def apply(self, dataset: pd.DataFrame) -> pd.DataFrame:
|
|
raise NotImplementedError("Please implement this method in the subclass.")
|
|
|
|
|
|
class JoinedCategoryFormatter(Formatter):
|
|
def apply(self, dataset: pd.DataFrame) -> pd.DataFrame:
|
|
"""Format the label column to `LabelA#LabelB` format."""
|
|
dataset[self.target_column] = dataset[self.target_column].apply(
|
|
lambda labels: "#".join(sorted(label.to_string() for label in labels))
|
|
)
|
|
return dataset
|
|
|
|
|
|
class ListedCategoryFormatter(Formatter):
|
|
def apply(self, dataset: pd.DataFrame) -> pd.DataFrame:
|
|
"""Format the label column to `['LabelA', 'LabelB']` format."""
|
|
dataset[self.target_column] = dataset[self.target_column].apply(
|
|
lambda labels: sorted([label.to_string() for label in labels])
|
|
)
|
|
return dataset
|
|
|
|
|
|
class FastTextCategoryFormatter(Formatter):
|
|
def apply(self, dataset: pd.DataFrame) -> pd.DataFrame:
|
|
"""Format the label column to `__label__LabelA __label__LabelB` format.
|
|
Also, drop the columns except for `data` and `self.target_column`.
|
|
"""
|
|
dataset = dataset[[DATA, self.target_column]]
|
|
dataset[self.target_column] = dataset[self.target_column].apply(
|
|
lambda labels: " ".join(sorted(f"__label__{label.to_string()}" for label in labels))
|
|
)
|
|
dataset[self.target_column] = dataset[self.target_column].fillna("")
|
|
dataset = dataset[self.target_column] + " " + dataset[DATA]
|
|
return dataset
|
|
|
|
|
|
class TupledSpanFormatter(Formatter):
|
|
def apply(self, dataset: pd.DataFrame) -> pd.DataFrame:
|
|
"""Format the span column to `(start_offset, end_offset, label)` format"""
|
|
dataset[self.target_column] = dataset[self.target_column].apply(
|
|
lambda spans: sorted(span.to_tuple() for span in spans)
|
|
)
|
|
return dataset
|
|
|
|
|
|
class DictFormatter(Formatter):
|
|
def apply(self, dataset: pd.DataFrame) -> pd.DataFrame:
|
|
"""Format the column to `{key: value}` format"""
|
|
dataset[self.target_column] = dataset[self.target_column].apply(
|
|
lambda labels: [label.to_dict() for label in labels]
|
|
)
|
|
return dataset
|
|
|
|
|
|
class RenameFormatter(Formatter):
|
|
def format(self, dataset: pd.DataFrame) -> pd.DataFrame:
|
|
return self.apply(dataset)
|
|
|
|
def apply(self, dataset: pd.DataFrame) -> pd.DataFrame:
|
|
"""Rename columns"""
|
|
dataset.rename(columns=self.mapper, inplace=True)
|
|
return dataset
|