πŸ“„ Tab FactΒΆ

The problem of verifying whether a textual hypothesis holds the truth based on the given evidence, also known as fact verification, plays an important role in the study of natural language understanding and semantic representation. However, existing studies are restricted to dealing with unstructured textual evidence (e.g., sentences and passages, a pool of passages), while verification using structured forms of evidence, such as tables, graphs, and databases, remains unexplored. TABFACT is large scale dataset with 16k Wikipedia tables as evidence for 118k human annotated statements designed for fact verification with semi-structured evidence… See the full description on the dataset page: https://huggingface.co/datasets/ibm/tab_fact

Tags: arxiv:1909.02164, license:cc-by-4.0, region:us, task_categories:text-classification

cards.tab_fact

type: TaskCard
loader: 
  type: LoadHF
  path: ibm/tab_fact
  streaming: False
  data_classification_policy: 
    - public
preprocess_steps: 
  - type: Rename
    field_to_field: 
      table: text_a
      statement: text_b
  - type: MapInstanceValues
    mappers: 
      label: 
        0: refuted
        1: entailed
  - type: Set
    fields: 
      type_of_relation: entailment
      text_a_type: Table
      text_b_type: Statement
      classes: 
        - refuted
        - entailed
task: tasks.classification.multi_class.relation
templates: templates.classification.multi_class.relation.all
[source]

Explanation about TaskCardΒΆ

TaskCard delineates the phases in transforming the source dataset into 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 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 LoadHFΒΆ

Loads datasets from the HuggingFace Hub.

It supports loading with or without streaming, and it can filter datasets upon loading.

Args:

path: The path or identifier of the dataset on the HuggingFace Hub. name: An optional dataset name. data_dir: Optional directory to store downloaded data. split: Optional specification of which split to load. data_files: Optional specification of particular data files to load. revision: Optional. The revision of the dataset. Often the commit id. Use in case you want to set the dataset version. streaming: Bool indicating if streaming should be used. filtering_lambda: A lambda function for filtering the data after loading. num_proc: Optional integer to specify the number of processes to use for parallel dataset loading.

Example:

Loading glue’s mrpc dataset

load_hf = LoadHF(path='glue', name='mrpc')

Explanation about SetΒΆ

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 β€˜/’ to access inner fields

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

Examples:

# Add a β€˜classes’ field with a value of a list β€œpositive” and β€œnegative” to all streams Set(fields={β€œclasses”: [β€œpositive”,”negatives”]})

# Add a β€˜start’ field under the β€˜span’ field with a value of 0 to all streams Set(fields={β€œspan/start”: 0}

# Add a β€˜classes’ field with a value of a list β€œpositive” and β€œnegative” to β€˜train’ stream Set(fields={β€œclasses”: [β€œpositive”,”negatives”], apply_to_stream=[β€œtrain”]})

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

Explanation about RenameΒΆ

Renames fields.

Move value from one field to another, potentially, if field name contains a /, from one branch into another. Remove the from field, potentially part of it in case of / in from_field.

Examples:

Rename(field_to_field={β€œb”: β€œc”}) will change inputs [{β€œa”: 1, β€œb”: 2}, {β€œa”: 2, β€œb”: 3}] to [{β€œa”: 1, β€œc”: 2}, {β€œa”: 2, β€œc”: 3}]

Rename(field_to_field={β€œb”: β€œc/d”}) will change inputs [{β€œa”: 1, β€œb”: 2}, {β€œa”: 2, β€œb”: 3}] to [{β€œa”: 1, β€œc”: {β€œd”: 2}}, {β€œa”: 2, β€œc”: {β€œd”: 3}}]

Rename(field_to_field={β€œb”: β€œb/d”}) will change inputs [{β€œa”: 1, β€œb”: 2}, {β€œa”: 2, β€œb”: 3}] to [{β€œa”: 1, β€œb”: {β€œd”: 2}}, {β€œa”: 2, β€œb”: {β€œd”: 3}}]

Rename(field_to_field={β€œb/c/e”: β€œb/d”}) will change inputs [{β€œa”: 1, β€œb”: {β€œc”: {β€œe”: 2, β€œf”: 20}}}] to [{β€œa”: 1, β€œb”: {β€œc”: {β€œf”: 20}, β€œd”: 2}}]

Explanation about MapInstanceValuesΒΆ

A class used to map instance values into other values.

This class is a type of InstanceOperator, it maps values of instances in a stream using predefined mappers.

Attributes:
mappers (Dict[str, Dict[str, Any]]): The mappers to use for mapping instance values.

Keys are the names of the fields to undergo mapping, and values are dictionaries that define the mapping from old values to new values.

strict (bool): If True, the mapping is applied strictly. That means if a value

does not exist in the mapper, it will raise a KeyError. If False, values that are not present in the mapper are kept as they are.

process_every_value (bool): If True, all fields to be mapped should be lists, and the mapping

is to be applied to their individual elements. If False, mapping is only applied to a field containing a single value.

Examples:

MapInstanceValues(mappers={β€œa”: {β€œ1”: β€œhi”, β€œ2”: β€œbye”}}) replaces β€˜1’ with β€˜hi’ and β€˜2’ with β€˜bye’ in field β€˜a’ in all instances of all streams: instance {β€œa”:”1”, β€œb”: 2} becomes {β€œa”:”hi”, β€œb”: 2}.

MapInstanceValues(mappers={β€œa”: {β€œ1”: β€œhi”, β€œ2”: β€œbye”}}, process_every_value=True) Assuming field β€˜a’ is a list of values, potentially including β€œ1”-s and β€œ2”-s, this replaces each such β€œ1” with β€œhi” and β€œ2” – with β€œbye” in all instances of all streams: instance {β€œa”: [β€œ1”, β€œ2”], β€œb”: 2} becomes {β€œa”: [β€œhi”, β€œbye”], β€œb”: 2}.

MapInstanceValues(mappers={β€œa”: {β€œ1”: β€œhi”, β€œ2”: β€œbye”}}, strict=True) To ensure that all values of field β€˜a’ are mapped in every instance, use strict=True. Input instance {β€œa”:”3”, β€œb”: 2} will raise an exception per the above call, because β€œ3” is not a key in the mapper of β€œa”.

MapInstanceValues(mappers={β€œa”: {str([1,2,3,4]): β€˜All’, str([]): β€˜None’}}, strict=True) replaces a list [1,2,3,4] with the string β€˜All’ and an empty list by string β€˜None’. Note that mapped values are defined by their string representation, so mapped values must be converted to strings.

References: templates.classification.multi_class.relation.all, tasks.classification.multi_class.relation

Read more about catalog usage here.