📄 qa¶

Note

ID: cards.coqa.qa | Type: TaskCard

{
    "loader": {
        "path": "stanfordnlp/coqa",
        "type": "load_hf"
    },
    "preprocess_steps": [
        "splitters.small_no_test",
        {
            "fields": {
                "context_type": "story"
            },
            "type": "add_fields"
        },
        {
            "fields": [
                "questions",
                "answers/input_text"
            ],
            "to_field": "dialog",
            "type": "zip_field_values"
        },
        {
            "field": "dialog",
            "process_every_value": true,
            "type": "dictify",
            "with_keys": [
                "user",
                "system"
            ]
        },
        {
            "field": "dialog",
            "type": "duplicate_by_sub_lists"
        },
        {
            "field": "dialog",
            "item": -1,
            "to_field": "last_turn",
            "type": "get"
        },
        {
            "field_to_field": {
                "last_turn/system": "answer",
                "last_turn/user": "question"
            },
            "type": "copy_fields"
        },
        {
            "field": "answer",
            "inside": "list",
            "to_field": "answers",
            "type": "wrap"
        },
        {
            "context_field": "story",
            "field": "dialog",
            "to_field": "context",
            "type": "serialize_dialog"
        }
    ],
    "task": "tasks.qa.with_context.extractive",
    "templates": "templates.qa.with_context.all",
    "type": "task_card"
}

Explanation about TaskCard¶

TaskCard delineates the phases in transforming the source dataset into a model-input, and specifies the metrics for evaluation of model-output.

Attributes:
loader: specifies the source address and the loading operator that can access that source and transform it into a unitxt multistream.

preprocess_steps: list of unitxt operators to process the data source into a model-input.

task: specifies the fields (of the already (pre)processed instance) making the inputs, the fields making the outputs, and the metrics to be used for evaluating the model output.

templates: format strings to be applied on the input fields (specified by the task) and the output fields. The template also carries the instructions and the list of postprocessing steps, to be applied to the model output.

Explanation about SerializeDialog¶

Serializes dialog data for feeding into a model.

This class takes structured dialog data and converts it into a text format according to a specified template. It allows for the inclusion or exclusion of system responses and can operate on a per-turn basis or aggregate the entire dialog.

Attributes:
field (str): The field in the input data that contains the dialog. to_field (Optional[str]): The field in the output data where the serialized dialog will be stored. last_user_turn_to_field (Optional[str]): Field to store the last user turn. last_system_turn_to_field (Optional[str]): Field to store the last system turn. context_field (Optional[str]): Field that contains additional context to be prepended to the dialog.

Explanation about ZipFieldValues¶

Zips values of multiple fields in a given instance, similar to list(zip(*fields)).

The value in each of the specified ‘fields’ is assumed to be a list. The lists from all ‘fields’ are zipped, and stored into ‘to_field’.

If ‘longest’=False, the length of the zipped result is determined by the shortest input value. If ‘longest’=False, the length of the zipped result is determined by the longest input, padding shorter inputs with None -s.

Explanation about CopyFields¶

Copies values from specified fields to specified fields.

Args (of parent class):
field_to_field (Union[List[List], Dict[str, str]]): A list of lists, where each sublist contains the source field and the destination field, or a dictionary mapping source fields to destination fields.

Examples:
An input instance {“a”: 2, “b”: 3}, when processed by CopyField(field_to_field={“a”: “b”} would yield {“a”: 2, “b”: 2}, and when processed by CopyField(field_to_field={“a”: “c”} would yield {“a”: 2, “b”: 3, “c”: 2}

with field names containing / , we can also copy inside the field: CopyFields(field_to_field={“a/0”: “a”}) would process instance {“a”: [1, 3]} into {“a”: 1}

Explanation about AddFields¶

Adds specified fields to each instance in a given stream or all streams (default) If fields exist, updates them.

Args:

fields (Dict[str, object]): The fields to add to each instance.
Use ‘/’ to access inner fields

use_deepcopy (bool) : Deep copy the input value to avoid later modifications

Examples:
# Add a ‘classes’ field with a value of a list “positive” and “negative” to all streams AddFields(fields={“classes”: [“positive”,”negatives”]})

# Add a ‘start’ field under the ‘span’ field with a value of 0 to all streams AddFields(fields={“span/start”: 0}

# Add a ‘classes’ field with a value of a list “positive” and “negative” to ‘train’ stream AddFields(fields={“classes”: [“positive”,”negatives”], apply_to_stream=[“train”]})

# Add a ‘classes’ field on a given list, prevent modification of original list # from changing the instance. AddFields(fields={“classes”: alist}), use_deepcopy=True) # if now alist is modified, still the instances remain intact.

References: tasks.qa.with_context.extractive, templates.qa.with_context.all, splitters.small_no_test