标签

Labels 管道使用文本分类模型为输入文本应用标签。此管道可以使用零样本模型（动态标注）或标准文本分类模型（固定标注）对文本进行分类。

示例

以下是使用此管道的一个简单示例。

from txtai.pipeline import Labels

# Create and run pipeline
labels = Labels()
labels(
    ["Great news", "That's rough"],
    ["positive", "negative"]
)

有关更详细的示例，请参阅下面的链接。

Notebook	描述
使用零样本分类应用标签	使用零样本学习进行标注、分类和主题建模

配置驱动的示例

管道通过 Python 或配置运行。可以在配置中使用管道的小写名称实例化管道。配置驱动的管道使用工作流或API运行。

config.yml

# Create pipeline using lower case class name
labels:

# Run pipeline with workflow
workflow:
  labels:
    tasks:
      - action: labels
        args: [["positive", "negative"]]

使用工作流运行

from txtai import Application

# Create and run pipeline with workflow
app = Application("config.yml")
list(app.workflow("labels", ["Great news", "That's rough"]))

使用 API 运行

CONFIG=config.yml uvicorn "txtai.api:app" &

curl \
  -X POST "http://localhost:8000/workflow" \
  -H "Content-Type: application/json" \
  -d '{"name":"labels", "elements": ["Great news", "Thats rough"]}'

方法

此管道的 Python 文档。

__init__(path=None, quantize=False, gpu=True, model=None, dynamic=True, **kwargs)

源代码位于 txtai/pipeline/text/labels.py

def __init__(self, path=None, quantize=False, gpu=True, model=None, dynamic=True, **kwargs):
    super().__init__("zero-shot-classification" if dynamic else "text-classification", path, quantize, gpu, model, **kwargs)

    # Set if labels are dynamic (zero shot) or fixed (standard text classification)
    self.dynamic = dynamic

__call__(text, labels=None, multilabel=False, flatten=None, workers=0, **kwargs)

将文本分类器应用于文本。返回按最高分数排序的 (id, score) 列表，其中 id 是标签中的索引。对于零样本分类，需要一个标签列表。对于文本分类模型，标签列表是可选的，否则返回所有已训练的标签。

此方法支持字符串或列表作为文本输入。如果输入是字符串，则返回类型为一维的 (id, score) 列表。如果文本是列表，则返回二维的 (id, score) 列表，每行对应一个字符串。

参数

名称	类型	描述	默认值
text		字符串|列表	必需
labels		标签列表	None
multilabel		如果为 True，则标签是独立的；如果为 False，则分数按文本项归一化，总和为 1；如果为 None，则返回原始分数。	False
flatten		如果存在，则将输出展平为标签列表。接受布尔值或浮点值，仅保留大于该值的分数。	None
workers		用于处理数据的并发工作者数量，默认为 None	0
kwargs		附加关键字参数	{}

返回值

类型	描述
	根据 flatten 参数，返回 (id, score) 列表或标签列表

源代码位于 txtai/pipeline/text/labels.py

def __call__(self, text, labels=None, multilabel=False, flatten=None, workers=0, **kwargs):
    """
    Applies a text classifier to text. Returns a list of (id, score) sorted by highest score,
    where id is the index in labels. For zero shot classification, a list of labels is required.
    For text classification models, a list of labels is optional, otherwise all trained labels are returned.

    This method supports text as a string or a list. If the input is a string, the return
    type is a 1D list of (id, score). If text is a list, a 2D list of (id, score) is
    returned with a row per string.

    Args:
        text: text|list
        labels: list of labels
        multilabel: labels are independent if True, scores are normalized to sum to 1 per text item if False, raw scores returned if None
        flatten: flatten output to a list of labels if present. Accepts a boolean or float value to only keep scores greater than that number.
        workers: number of concurrent workers to use for processing data, defaults to None
        kwargs: additional keyword args

    Returns:
        list of (id, score) or list of labels depending on flatten parameter
    """

    if self.dynamic:
        # Run zero shot classification pipeline
        results = self.pipeline(text, labels, multi_label=multilabel, truncation=True, num_workers=workers)
    else:
        # Set classification function based on inputs
        function = "none" if multilabel is None else "sigmoid" if multilabel or len(self.labels()) == 1 else "softmax"

        # Run text classification pipeline
        results = self.pipeline(text, top_k=None, function_to_apply=function, num_workers=workers, **kwargs)

    # Convert results to a list if necessary
    if isinstance(text, str):
        results = [results]

    # Build list of outputs and return
    outputs = self.outputs(results, labels, flatten)
    return outputs[0] if isinstance(text, str) else outputs