Skip to content

Text Classification

nyckel.TextClassificationFunction

Example:

from nyckel import Credentials, TextClassificationFunction

credentials = Credentials(client_id="...", client_secret="...")

func = TextClassificationFunction.create("IsToxic", credentials)
func.create_samples([
    ("This is a nice comment", "Not toxic"),
    ("Hello friend", "Not toxic"),
    ("I hate people like you", "Toxic"),
    ("Who is this? Go away!", "Toxic"),
])

predictions = func.invoke(["This example is fantastic!"])
Source code in src/nyckel/functions/classification/text_classification.py
class TextClassificationFunction(ClassificationFunction):
    """
    Example:

    ```py

    from nyckel import Credentials, TextClassificationFunction

    credentials = Credentials(client_id="...", client_secret="...")

    func = TextClassificationFunction.create("IsToxic", credentials)
    func.create_samples([
        ("This is a nice comment", "Not toxic"),
        ("Hello friend", "Not toxic"),
        ("I hate people like you", "Toxic"),
        ("Who is this? Go away!", "Toxic"),
    ])

    predictions = func.invoke(["This example is fantastic!"])
    ```
    """

    def __init__(self, function_id: NyckelId, credentials: Credentials):
        self._function_id = function_id

        self._function_handler = ClassificationFunctionHandler(function_id, credentials)
        self._label_handler = ClassificationLabelHandler(function_id, credentials)
        self._url_handler = ClassificationFunctionURLHandler(function_id, credentials.server_url)
        self._sample_handler = ClassificationSampleHandler(function_id, credentials)

        assert self._function_handler.get_input_modality() == "Text"

    def __str__(self) -> str:
        return self.__repr__()

    def __repr__(self) -> str:
        status_string = f"Name: {self.name}, id: {self.function_id}, url: {self._url_handler.train_page}"
        return status_string

    @property
    def function_id(self) -> str:
        return self._function_id

    @property
    def sample_count(self) -> int:
        return self._function_handler.sample_count

    @property
    def label_count(self) -> int:
        return self._function_handler.label_count

    @property
    def name(self) -> str:
        return self._function_handler.get_name()

    @classmethod
    def create(cls, name: str, credentials: Credentials) -> "TextClassificationFunction":
        return factory.ClassificationFunctionFactory.create(name, "Text", credentials)  # type:ignore

    def delete(self) -> None:
        self._function_handler.delete()

    def invoke(
        self,
        sample_data_list: List[TextSampleData],
        model_id: str = "",
    ) -> List[ClassificationPrediction]:
        return self._sample_handler.invoke(sample_data_list, lambda x: x, model_id=model_id)

    def has_trained_model(self) -> bool:
        return self._function_handler.is_trained

    def create_labels(self, labels: Sequence[Union[ClassificationLabel, str]]) -> List[NyckelId]:
        typed_labels = [
            label if isinstance(label, ClassificationLabel) else ClassificationLabel(name=label) for label in labels
        ]
        return self._label_handler.create_labels(typed_labels)

    def list_labels(self) -> List[ClassificationLabel]:
        return self._label_handler.list_labels(self.label_count)

    def read_label(self, label_id: NyckelId) -> ClassificationLabel:
        return self._label_handler.read_label(label_id)

    def update_label(self, label: ClassificationLabel) -> ClassificationLabel:
        return self._label_handler.update_label(label)

    def delete_labels(self, label_ids: List[NyckelId]) -> None:
        return self._label_handler.delete_labels(label_ids)

    def create_samples(
        self, samples: Sequence[Union[TextClassificationSample, Tuple[TextSampleData, LabelName], TextSampleData]]  # type: ignore  # noqa: E501
    ) -> List[NyckelId]:
        typed_samples = self._wrangle_post_samples_input(samples)
        typed_samples = self._strip_label_names(typed_samples)
        self._create_labels_as_needed(typed_samples)

        return self._sample_handler.create_samples(typed_samples, lambda x: x)

    def list_samples(self) -> List[TextClassificationSample]:  # type: ignore
        samples_dict_list = self._sample_handler.list_samples(self.sample_count)
        labels = self._label_handler.list_labels(None)

        label_name_by_id = {label.id: label.name for label in labels}

        return [self._sample_from_dict(entry, label_name_by_id) for entry in samples_dict_list]  # type: ignore

    def read_sample(self, sample_id: NyckelId) -> TextClassificationSample:
        sample_dict = self._sample_handler.read_sample(sample_id)

        labels = self._label_handler.list_labels(None)
        label_name_by_id = {strip_nyckel_prefix(label.id): label.name for label in labels}  # type: ignore

        return self._sample_from_dict(sample_dict, label_name_by_id)  # type: ignore

    def update_annotation(self, sample: TextClassificationSample) -> None:  # type: ignore
        self._sample_handler.update_annotation(sample)

    def delete_samples(self, sample_ids: List[NyckelId]) -> None:
        self._sample_handler.delete_samples(sample_ids)

    def _sample_from_dict(self, sample_dict: Dict, label_name_by_id: Dict[NyckelId, str]) -> TextClassificationSample:
        if "annotation" in sample_dict:
            annotation = ClassificationAnnotation(
                label_name=label_name_by_id[strip_nyckel_prefix(sample_dict["annotation"]["labelId"])],
            )
        else:
            annotation = None
        if "prediction" in sample_dict:
            prediction = ClassificationPrediction(
                confidence=sample_dict["prediction"]["confidence"],
                label_name=label_name_by_id[strip_nyckel_prefix(sample_dict["prediction"]["labelId"])],
            )
        else:
            prediction = None
        return TextClassificationSample(
            id=strip_nyckel_prefix(sample_dict["id"]),
            data=sample_dict["data"],
            external_id=sample_dict["externalId"] if "externalId" in sample_dict else None,
            annotation=annotation,
            prediction=prediction,
        )

    def _wrangle_post_samples_input(
        self, samples: Sequence[Union[TextClassificationSample, Tuple[TextSampleData, LabelName], TextSampleData]]
    ) -> List[TextClassificationSample]:
        typed_samples: List[TextClassificationSample] = []
        for sample in samples:
            if isinstance(sample, str):
                typed_samples.append(TextClassificationSample(data=sample))
            elif isinstance(sample, (list, tuple)):
                typed_samples.append(
                    TextClassificationSample(data=sample[0], annotation=ClassificationAnnotation(label_name=sample[1]))
                )
            elif isinstance(sample, TextClassificationSample):
                typed_samples.append(sample)
            else:
                raise ValueError(f"Unknown sample type: {type(sample)}")
        return typed_samples

    def _create_labels_as_needed(self, samples: List[TextClassificationSample]) -> None:
        existing_labels = self._label_handler.list_labels(None)
        existing_label_names = {label.name for label in existing_labels}
        new_label_names = {sample.annotation.label_name for sample in samples if sample.annotation}
        missing_label_names = new_label_names - existing_label_names
        missing_labels = [ClassificationLabel(name=label_name) for label_name in missing_label_names]
        if len(missing_labels) > 0:
            self._label_handler.create_labels(missing_labels)

    def _strip_label_names(self, samples: List[TextClassificationSample]) -> List[TextClassificationSample]:
        for sample in samples:
            if sample.annotation:
                sample.annotation.label_name = sample.annotation.label_name.strip()
        return samples

function_id property

function_id: str

sample_count property

sample_count: int

label_count property

label_count: int

name property

name: str

__init__

__init__(function_id: NyckelId, credentials: Credentials)
Source code in src/nyckel/functions/classification/text_classification.py
def __init__(self, function_id: NyckelId, credentials: Credentials):
    self._function_id = function_id

    self._function_handler = ClassificationFunctionHandler(function_id, credentials)
    self._label_handler = ClassificationLabelHandler(function_id, credentials)
    self._url_handler = ClassificationFunctionURLHandler(function_id, credentials.server_url)
    self._sample_handler = ClassificationSampleHandler(function_id, credentials)

    assert self._function_handler.get_input_modality() == "Text"

create classmethod

create(name: str, credentials: Credentials) -> TextClassificationFunction
Source code in src/nyckel/functions/classification/text_classification.py
@classmethod
def create(cls, name: str, credentials: Credentials) -> "TextClassificationFunction":
    return factory.ClassificationFunctionFactory.create(name, "Text", credentials)  # type:ignore

delete

delete() -> None
Source code in src/nyckel/functions/classification/text_classification.py
def delete(self) -> None:
    self._function_handler.delete()

invoke

invoke(sample_data_list: List[TextSampleData], model_id: str = '') -> List[ClassificationPrediction]
Source code in src/nyckel/functions/classification/text_classification.py
def invoke(
    self,
    sample_data_list: List[TextSampleData],
    model_id: str = "",
) -> List[ClassificationPrediction]:
    return self._sample_handler.invoke(sample_data_list, lambda x: x, model_id=model_id)

has_trained_model

has_trained_model() -> bool
Source code in src/nyckel/functions/classification/text_classification.py
def has_trained_model(self) -> bool:
    return self._function_handler.is_trained

create_labels

create_labels(labels: Sequence[Union[ClassificationLabel, str]]) -> List[NyckelId]
Source code in src/nyckel/functions/classification/text_classification.py
def create_labels(self, labels: Sequence[Union[ClassificationLabel, str]]) -> List[NyckelId]:
    typed_labels = [
        label if isinstance(label, ClassificationLabel) else ClassificationLabel(name=label) for label in labels
    ]
    return self._label_handler.create_labels(typed_labels)

list_labels

list_labels() -> List[ClassificationLabel]
Source code in src/nyckel/functions/classification/text_classification.py
def list_labels(self) -> List[ClassificationLabel]:
    return self._label_handler.list_labels(self.label_count)

read_label

read_label(label_id: NyckelId) -> ClassificationLabel
Source code in src/nyckel/functions/classification/text_classification.py
def read_label(self, label_id: NyckelId) -> ClassificationLabel:
    return self._label_handler.read_label(label_id)

update_label

update_label(label: ClassificationLabel) -> ClassificationLabel
Source code in src/nyckel/functions/classification/text_classification.py
def update_label(self, label: ClassificationLabel) -> ClassificationLabel:
    return self._label_handler.update_label(label)

delete_labels

delete_labels(label_ids: List[NyckelId]) -> None
Source code in src/nyckel/functions/classification/text_classification.py
def delete_labels(self, label_ids: List[NyckelId]) -> None:
    return self._label_handler.delete_labels(label_ids)

create_samples

create_samples(samples: Sequence[Union[TextClassificationSample, Tuple[TextSampleData, LabelName], TextSampleData]]) -> List[NyckelId]
Source code in src/nyckel/functions/classification/text_classification.py
def create_samples(
    self, samples: Sequence[Union[TextClassificationSample, Tuple[TextSampleData, LabelName], TextSampleData]]  # type: ignore  # noqa: E501
) -> List[NyckelId]:
    typed_samples = self._wrangle_post_samples_input(samples)
    typed_samples = self._strip_label_names(typed_samples)
    self._create_labels_as_needed(typed_samples)

    return self._sample_handler.create_samples(typed_samples, lambda x: x)

list_samples

list_samples() -> List[TextClassificationSample]
Source code in src/nyckel/functions/classification/text_classification.py
def list_samples(self) -> List[TextClassificationSample]:  # type: ignore
    samples_dict_list = self._sample_handler.list_samples(self.sample_count)
    labels = self._label_handler.list_labels(None)

    label_name_by_id = {label.id: label.name for label in labels}

    return [self._sample_from_dict(entry, label_name_by_id) for entry in samples_dict_list]  # type: ignore

read_sample

read_sample(sample_id: NyckelId) -> TextClassificationSample
Source code in src/nyckel/functions/classification/text_classification.py
def read_sample(self, sample_id: NyckelId) -> TextClassificationSample:
    sample_dict = self._sample_handler.read_sample(sample_id)

    labels = self._label_handler.list_labels(None)
    label_name_by_id = {strip_nyckel_prefix(label.id): label.name for label in labels}  # type: ignore

    return self._sample_from_dict(sample_dict, label_name_by_id)  # type: ignore

update_annotation

update_annotation(sample: TextClassificationSample) -> None
Source code in src/nyckel/functions/classification/text_classification.py
def update_annotation(self, sample: TextClassificationSample) -> None:  # type: ignore
    self._sample_handler.update_annotation(sample)

delete_samples

delete_samples(sample_ids: List[NyckelId]) -> None
Source code in src/nyckel/functions/classification/text_classification.py
def delete_samples(self, sample_ids: List[NyckelId]) -> None:
    self._sample_handler.delete_samples(sample_ids)

nyckel.TextClassificationSample dataclass

Source code in src/nyckel/functions/classification/classification.py
@dataclass
class TextClassificationSample:
    data: TextSampleData
    id: Optional[NyckelId] = None
    external_id: Optional[str] = None
    annotation: Optional[ClassificationAnnotation] = None
    prediction: Optional[ClassificationPrediction] = None

data instance-attribute

id class-attribute instance-attribute

id: Optional[NyckelId] = None

external_id class-attribute instance-attribute

external_id: Optional[str] = None

annotation class-attribute instance-attribute

annotation: Optional[ClassificationAnnotation] = None

prediction class-attribute instance-attribute

prediction: Optional[ClassificationPrediction] = None

__init__

__init__(data: TextSampleData, id: Optional[NyckelId] = None, external_id: Optional[str] = None, annotation: Optional[ClassificationAnnotation] = None, prediction: Optional[ClassificationPrediction] = None) -> None

nyckel.TextSampleData module-attribute

TextSampleData = str