Hugging Face's logo

Datasets:
patriziobellan
/
PET

Tasks:
Token Classification
Languages:
English
Size:
n<1K
ArXiv:
Tags:
Business Process Management
NLP
ML
DL
License:
Dataset card Files Community
6
Dataset Viewer
Full Screen

The viewer is disabled because this dataset repo requires arbitrary Python code execution. Please consider removing the loading script and relying on automated data support (you can use convert_to_parquet from the datasets library). If this is not possible, please open a discussion for direct help.

PET: A NEW DATASET FOR PROCESS EXTRACTION FROM TEXT

Dataset Card for PET

Dataset Summary

Abstract. Although there is a long tradition of work in NLP on extracting entities and relations from text, to date there exists little work on the acquisition of business processes from unstructured data such as textual corpora of process descriptions. With this work we aim at filling this gap and establishing the first steps towards bridging data-driven information extraction methodologies from Natural Language Processing and the model-based formalization that is aimed from Business Process Management. For this, we develop the first corpus of business process descriptions annotated with activities, actors, activity data, gateways and their conditions. We present our new resource to benchmark the difficulty and challenges of business process extraction from text.

Supported Tasks and Leaderboards

  • Token Classification
  • Named Entity Recognition
  • Relations Extraction

Languages

English

Dataset Structure

Test set to beanchmark Business Process Extraction from Text approaches.

Data Instances

Token Classification

For each instance, there is a document name representing the name of the document of the Friedrich et al. dataset, an integer representing the number of the sentence, a list of tokens representing the words of the sentence instance, and a list of ner tags (in IOB2 format) representing the annotation of process elements of the sentence.

Below, an example of data instance.

{
  "document name":"doc-1.1",
  "sentence-ID":1,
  "tokens":["Whenever","the","sales","department","receives","an","order",",","a","new","process","instance","is","created","."],
  "ner-tags":["O","B-Actor","I-Actor","I-Actor","B-Activity","B-Activity Data","I-Activity Data","O","O","O","O","O","O","O","O"]
}

Relations Extraction

For each instance, there is a document name representing the name of the document of the Friedrich et al. dataset, a list of tokens representing the words of the document instance, a list of interger representing the words position within each sentence of the document instance, a list of ner tags (in IOB2 format) representing the annotation of the token, a list of sentence id representing for each token the number of the sentence, and a list of relations of the document.

Below, an example of data instance.

{
  "document name": "doc-1.1", 
  "tokens": ["A", "small", "company",...],
  "tokens-IDs": [0, 1, 2, ...], 
  "ner_tags": ["O", "O", "O", ...],
  "sentence-IDs": [0, 0, 0, ...],
  "relations": {
                "source-head-sentence-ID": [1, 1, 1, ...], 
                "source-head-word-ID": [4, 4, 4, ...], 
                "relation-type": ["uses", "flow", "actor recipient", ...], 
                "target-head-sentence-ID": [1, 2, 1,...], 
                "target-head-word-ID": [5, 9, 1, ...]
                }
}

Data Fields

Token Classification

  • document name: a string used to represent the name of the document.
  • sentence-ID: an integer (starting from 0) representing the number of the sentence within the document.
  • tokens: a list of string representing the words of the sentence
  • ner-tags: a list of string representing the annotation for each word.

The allowed ner-tags are:

  • O: An O tag indicates that a token belongs to no chunk.
  • B-Actor: This tag indicates the beginning of an Actor chunk.
  • I-Actor: This tag indicates that the tag is inside an Actor chunk.
  • B-Activity: This tag indicates the beginning of an Activity chunk.
  • I-Activity: This tag indicates that the tag is inside an Activity chunk.
  • B-Activity Data: This tag indicates the beginning of an Activity Data chunk.
  • I-Activity Data: This tag indicates that the tag is inside an Activity Data chunk.
  • B-Further Specification: This tag indicates the beginning of a Further Specification chunk.
  • I-Further Specification: This tag indicates that the tag is inside a Further Specification chunk.
  • B-XOR Gateway: This tag indicates the beginning of a XOR Gateway chunk.
  • I-XOR Gateway: This tag indicates that the tag is inside a XOR Gateway chunk.
  • B-Condition Specification: This tag indicates the beginning of a Condition Specification chunk.
  • I-Condition Specification: This tag indicates that the tag is inside a Condition Specification chunk.
  • B-AND Gateway: This tag indicates the beginning of an AND Gateway chunk.
  • I-AND Gateway: This tag indicates that the tag is inside an AND Gateway chunk.

To have a complete explanation of each process element tag please refer to the research paper and the annotation guidelines.

Relations Extraction

  • document name: a string used to represent the name of the document.
  • tokens: a list of string representing the words of the document
  • tokens-IDs: a list of interger representing the word position within a sentence.
  • ner_tags: a list of string representing the annotation for each word. (see ner-tags above)
  • sentence-IDs: a list of interger representing the sentence number for each word of the document.
  • relations:: a list of document relations.
    • source-head-sentence-ID: a list of sentence ID pointing to the sentence number of the head (first token) of the source entity.
    • source-head-word-ID: a list of token ID pointing to the word ID of the head (first token) of the source entity.
    • relation-type: a list of relation tags.
    • target-head-sentence-ID: a list of sentence ID pointing to the sentence number of the head (first token) of the target entity.
    • target-head-word-ID: a list of token ID pointing to the word ID of the head (first token) of the target entity.

For instance, a relation is defined by the instances of source-head-sentence-ID, source-head-word-ID, relation-type, target-head-sentence-ID, and target-head-word-ID at the same index position. In the following example, the first relation of the first document is shown:

document_1=modelhub_dataset['test'][0]
relation = {
            'source-head-sentence-ID': document_1['relations']['source-head-sentence-ID'][0],
            'source-head-word-ID': document_1['relations']['source-head-word-ID'][0],
            'relation-type': document_1['relations']['relation-type'][0],
            'target-head-sentence-ID': document_1['relations']['target-head-sentence-ID'][0],
            'target-head-word-ID': document_1['relations']['target-head-sentence-ID'][0],
          }
print(relation)

the output is:

{'relation-type': 'uses',
 'source-head-sentence-ID': 1,
 'source-head-word-ID': 4,
 'target-head-sentence-ID': 1,
 'target-head-word-ID': 1}

That means: the entity in sentence number 1, starting at the token position 4 has a uses relation with the entity in sentence number 1 starting at token position 1

Data Splits

The data was not split. It contains the test set only.

Dataset Creation

Curation Rationale

Although there is a long tradition of work in NLP on extracting entities and relations from text to date there exists little work on the acquisition of business processes from unstructured data such as textual corpora of process descriptions. With this work we aim at filling this gap and establishing the first steps towards bridging data-driven information extraction methodologies from Natural Language Processing and the model-based formalization that is aimed from Business Process Management.

Source Data

Initial Data Collection and Normalization

The dataset construction process has been split in five main phases:

  1. Text pre-processing. As the first operation, we check the content of each document and we tokenized it. This initial check was necessary since some of the original texts were automatically translated into English by the authors of the dataset. The translations were never validated, indeed, several errors have been found and fixed.

  2. Text Annotation. Each text has been annotated by using the guidelines. The team was composed by five annotators with high expertise in BPMN. Each document has been assigned to three experts that were in change of identifying all the elements and flows with each document. In this phase, we used the the Inception tool to support annotators.

  3. Automatic annotation fixing. After the second phase, we ran an automatic procedure relying on a rule-based script to automatically fix annotations that were not compliant with the guidelines. For example, if a modal verb was erroneously included in the annotation of an Activity, the procedure removed it from the annotation. Another example is the missing of the article within an annotation related to an Actor. In this case, the script included it in the annotation. This phase allowed to remove possible annotation errors and to obtain annotations compliant with the guidelines.

  4. Agreement Computation. Here, we computed, on the annotation provided by the experts, the agreement scores for each process element and for each relation between process elements pair adopting the methodology proposed in Hripcsak et al.. We measured the agreement in terms of the F1 measure because, besides being straightforward to calculate, it is directly interpretable. Note that chance-corrected measures like k approach the F1-measure as the number of cases that raters agree are negative grows. By following such a methodology, an annotation was considered in agreement among the experts if and only if they capture the same span of words and they assign the same process element tag to the annotation.

  5. Reconciliation. The last phase consisted of the mitigation of disagreements within the annotations provided by the experts. The aim of this phase is to obtain a shared and agreed set of gold standard annotations on each text for both entities and relations. Such entities also enable the generation of the related full-connected process model flow that can be rendered by using, but not limited to, a BPMN diagram. During this last phase, among the 47 documents originally included into the dataset, 2 of them were discarded. These texts were not fully annotated by the annotators since they were not be able to completely understand which process elements were actually included in some specific parts of the text. For this reason, the final size of the dataset is 45 textual descriptions of the corresponding process models together with their annotations.

Who are the source language producers?

English

Annotations

Annotation process

You can read about the annotation process in the original paper https://arxiv.org/abs/2203.04860

Who are the annotators?

Expert Annotators

Personal and Sensitive Information

No personal or sensitive information issues.

Considerations for Using the Data

Social Impact of Dataset

The dataset has no social impact

Discussion of Biases

No bias found in the dataset

Other Known Limitations

The Further specification and AND Gateway elements obtained very poor performance on the baselines proposed in the paper. The AND Gateway is the less represented process elements in this dataset. The Further Specification process element was the most difficult element to annotate.

Additional Information

Dataset Curators

  • Patrizio Bellan (Fondazione Bruno Kessler, Trento, Italy and Free University of Bozen-Bolzano, Bolzano, Italy)
  • Mauro Dragoni (Fondazione Bruno Kessler, Trento, Italy)
  • Chiara Ghidini (Fondazione Bruno Kessler, Trento, Italy)
  • Han van der Aa (University of Mannheim, Mannheim, Germany)
  • Simone Ponzetto (University of Mannheim, Mannheim, Germany)

Licensing Information

Citation Information 

@inproceedings{DBLP:conf/aiia/BellanGDPA22,
  author       = {Patrizio Bellan and
                  Chiara Ghidini and
                  Mauro Dragoni and
                  Simone Paolo Ponzetto and
                  Han van der Aa},
  editor       = {Debora Nozza and
                  Lucia C. Passaro and
                  Marco Polignano},
  title        = {Process Extraction from Natural Language Text: the {PET} Dataset and
                  Annotation Guidelines},
  booktitle    = {Proceedings of the Sixth Workshop on Natural Language for Artificial
                  Intelligence {(NL4AI} 2022) co-located with 21th International Conference
                  of the Italian Association for Artificial Intelligence (AI*IA 2022),
                  Udine, November 30th, 2022},
  series       = {{CEUR} Workshop Proceedings},
  volume       = {3287},
  pages        = {177--191},
  publisher    = {CEUR-WS.org},
  year         = {2022},
  url          = {https://ceur-ws.org/Vol-3287/paper18.pdf},
  timestamp    = {Fri, 10 Mar 2023 16:23:01 +0100},
  biburl       = {https://dblp.org/rec/conf/aiia/BellanGDPA22.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}

@inproceedings{DBLP:conf/bpm/BellanADGP22,
  author       = {Patrizio Bellan and
                  Han van der Aa and
                  Mauro Dragoni and
                  Chiara Ghidini and
                  Simone Paolo Ponzetto},
  editor       = {Cristina Cabanillas and
                  Niels Frederik Garmann{-}Johnsen and
                  Agnes Koschmider},
  title        = {{PET:} An Annotated Dataset for Process Extraction from Natural Language
                  Text Tasks},
  booktitle    = {Business Process Management Workshops - {BPM} 2022 International Workshops,
                  M{\"{u}}nster, Germany, September 11-16, 2022, Revised Selected
                  Papers},
  series       = {Lecture Notes in Business Information Processing},
  volume       = {460},
  pages        = {315--321},
  publisher    = {Springer},
  year         = {2022},
  url          = {https://doi.org/10.1007/978-3-031-25383-6\_23},
  doi          = {10.1007/978-3-031-25383-6\_23},
  timestamp    = {Tue, 14 Feb 2023 09:47:10 +0100},
  biburl       = {https://dblp.org/rec/conf/bpm/BellanADGP22.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}

Contributions

Thanks to Patrizio Bellan for adding this dataset.

Update

  • v1.0.0: Added token classification task
  • v1.0.1: Added extraction relation task
  • v1.1.0: Fixed minor errors, fixed performs relations

Version 1.1.0 cab be found here

Annotation Guidelines

Inception Schema

The inception schema can be found here

Annotation Guidelines

The Annotation guidelines and procedures adopted to annotate the PET dataset can be downloaded here

Article

The article can be downloaded here

Python Interface

A Python interface (beta version) to interact with the dataset can be found here

You can find the BASELINES, the annotation data, and a graphical interface to visualize predictions here

Benchmarks

A Python benchmarking procedure package to test approaches on the PET dataset ca be found here

Loading data

Token-classification task 

from datasets import load_dataset
modelhub_dataset = load_dataset("patriziobellan/PET", name='token-classification')

Relations-extraction task 

from datasets import load_dataset
modelhub_dataset = load_dataset("patriziobellan/PET", name='relations-extraction')
Downloads last month
229
Edit dataset card
Homepage:
pdi.fbk.eu
Paper:
arxiv.org
Point of Contact:
Patrizio Bellan