PIE Dataset Card for "cdcp"

This is a PyTorch-IE wrapper for the CDCP Huggingface dataset loading script.

Usage

from pie_datasets import load_dataset
from pie_documents.documents import TextDocumentWithLabeledSpansAndBinaryRelations

# load English variant
dataset = load_dataset("pie/cdcp")

# if required, normalize the document type (see section Document Converters below)
dataset_converted = dataset.to_document_type(TextDocumentWithLabeledSpansAndBinaryRelations)
assert isinstance(dataset_converted["train"][0], TextDocumentWithLabeledSpansAndBinaryRelations)

# get first relation in the first document
doc = dataset_converted["train"][0]
print(doc.binary_relations[0])
# BinaryRelation(head=LabeledSpan(start=0, end=78, label='value', score=1.0), tail=LabeledSpan(start=79, end=242, label='value', score=1.0), label='reason', score=1.0)
print(doc.binary_relations[0].resolve())
# ('reason', (('value', 'State and local court rules sometimes make default judgments much more likely.'), ('value', 'For example, when a person who allegedly owes a debt is told to come to court on a work day, they may be forced to choose between a default judgment and their job.')))

Data Schema

The document type for this dataset is CDCPDocument which defines the following data fields:

• text (str)
• id (str, optional)
• metadata (dictionary, optional)

and the following annotation layers:

• propositions (annotation type: LabeledSpan, target: text)
• relations (annotation type: BinaryRelation, target: propositions)
• urls (annotation type: Attribute, target: propositions)

See here for the annotation type definitions.

Document Converters

The dataset provides document converters for the following target document types:

• pie_documents.documents.TextDocumentWithLabeledSpansAndBinaryRelations
  • labeled_spans: LabeledSpan annotations, converted from CDCPDocument's propositions
    • labels: fact, policy, reference, testimony, value
    • if propositions contain whitespace at the beginning and/or the end, the whitespace are trimmed out.
  • binary_relations: BinaryRelation annotations, converted from CDCPDocument's relations
    • labels: reason, evidence

See here for the document type definitions.

Collected Statistics after Document Conversion

We use the script evaluate_documents.py from PyTorch-IE-Hydra-Template to generate these statistics. After checking out that code, the statistics and plots can be generated by the command:

python src/evaluate_documents.py dataset=cdcp_base metric=METRIC

where a METRIC is called according to the available metric configs in config/metric/METRIC (see metrics).

This also requires to have the following dataset config in configs/dataset/cdcp_base.yaml of this dataset within the repo directory:

_target_: src.utils.execute_pipeline
input:
  _target_: pie_datasets.DatasetDict.load_dataset
  path: pie/cdcp
  revision: 001722894bdca6df6a472d0d186a3af103e392c5

For token based metrics, this uses bert-base-uncased from transformer.AutoTokenizer (see AutoTokenizer, and bert-based-uncased to tokenize text in TextDocumentWithLabeledSpansAndBinaryRelations (see document type).

Relation argument (outer) token distance per label

The distance is measured from the first token of the first argumentative unit to the last token of the last unit, a.k.a. outer distance.

We collect the following statistics: number of documents in the split (no. doc), no. of relations (len), mean of token distance (mean), standard deviation of the distance (std), minimum outer distance (min), and maximum outer distance (max). We also present histograms in the collapsible, showing the distribution of these relation distances (x-axis; and unit-counts in y-axis), accordingly.

Command

python src/evaluate_documents.py dataset=cdcp_base metric=relation_argument_token_distances

train (580 documents)

	len	max	mean	min	std
ALL	2204	240	48.839	8	31.462
evidence	94	196	66.723	14	42.444
reason	2110	240	48.043	8	30.64

Histogram (split: train, 580 documents)

test (150 documents)

	len	max	mean	min	std
ALL	648	212	51.299	8	31.159
evidence	52	170	73.923	20	39.855
reason	596	212	49.326	8	29.47

Histogram (split: test, 150 documents)

Span lengths (tokens)

The span length is measured from the first token of the first argumentative unit to the last token of the particular unit.

We collect the following statistics: number of documents in the split (no. doc), no. of spans (len), mean of number of tokens in a span (mean), standard deviation of the number of tokens (std), minimum tokens in a span (min), and maximum tokens in a span (max). We also present histograms in the collapsible, showing the distribution of these token-numbers (x-axis; and unit-counts in y-axis), accordingly.

Command

python src/evaluate_documents.py dataset=cdcp_base metric=span_lengths_tokens

statistics	train	test
no. doc	580	150
len	3901	1026
mean	19.441	18.758
std	11.71	10.388
min	2	3
max	142	83

Histogram (split: train, 580 documents)

Histogram (split: test, 150 documents)

Token length (tokens)

The token length is measured from the first token of the document to the last one.

We collect the following statistics: number of documents in the split (no. doc), mean of document token-length (mean), standard deviation of the length (std), minimum number of tokens in a document (min), and maximum number of tokens in a document (max). We also present histograms in the collapsible, showing the distribution of these token lengths (x-axis; and unit-counts in y-axis), accordingly.

Command

python src/evaluate_documents.py dataset=cdcp_base metric=count_text_tokens

statistics	train	test
no. doc	580	150
mean	130.781	128.673
std	101.121	98.708
min	13	15
max	562	571

Histogram (split: train, 580 documents)

Histogram (split: test, 150 documents)