metadata

annotations_creators:
  - other
language_creators:
  - found
language:
  - hi
license:
  - other
multilinguality:
  - monolingual
size_categories:
  - 1K<n<10K
source_datasets:
  - original
task_categories:
  - text-classification
task_ids:
  - multi-label-classification
pretty_name: Discourse Analysis dataset
tags:
  - discourse-analysis
dataset_info:
  features:
    - name: Story_no
      dtype: int32
    - name: Sentence
      dtype: string
    - name: Discourse Mode
      dtype:
        class_label:
          names:
            '0': Argumentative
            '1': Descriptive
            '2': Dialogue
            '3': Informative
            '4': Narrative
            '5': Other
  splits:
    - name: train
      num_bytes: 1998930
      num_examples: 9968
  download_size: 4176677
  dataset_size: 1998930

Dataset Card for Discourse Analysis dataset

Dataset Description
Dataset Structure
Dataset Creation
Considerations for Using the Data
Additional Information

Dataset Description

Homepage:
Repository: https://github.com/midas-research/hindi-discourse
Paper: An Annotated Dataset of Discourse Modes in Hindi Stories
Point of Contact: https://github.com/midas-research/MeTooMA

Dataset Summary

The Hindi Discourse Analysis dataset is a corpus for analyzing discourse modes present in its sentences.
It contains sentences from stories written by 11 famous authors from the 20th Century.
4-5 stories by each author have been selected which were available in the public domain resulting in a collection of 53 stories.
Most of these short stories were originally written in Hindi but some of them were written in other Indian languages and later translated to Hindi.

The corpus contains a total of 10472 sentences belonging to the following categories:

Argumentative
Descriptive
Dialogic
Informative
Narrative

Supported Tasks and Leaderboards

Discourse Analysis of Hindi.

Languages

Hindi

Dataset Structure

The dataset is structured into JSON format.

Data Instances

{'Story_no': 15, 'Sentence': ' गाँठ से साढ़े तीन रुपये लग गये, जो अब पेट में जाकर खनकते भी नहीं! जो तेरी करनी मालिक! ” “इसमें मालिक की क्या करनी है? ”', 'Discourse Mode': 'Dialogue'}

Data Fields

Sentence number, story number, sentence and discourse mode

Data Splits

Train: 9983

Dataset Creation

Curation Rationale

Present a new publicly available corpus consisting of sentences from short stories written in a low-resource language of Hindi having high quality annotation for five different discourse modes - argumentative, narrative, descriptive, dialogic and informative.
Perform a detailed analysis of the proposed annotated corpus and characterize the performance of different classification algorithms.

Source Data

Source of all the data points in this dataset is Hindi stories written by famous authors of Hindi literature.

Initial Data Collection and Normalization

All the data was collected from various Hindi websites.
We chose against crowd-sourcing the annotation pro- cess because we wanted to directly work with the an- notators for qualitative feedback and to also ensure high quality annotations.
We employed three native Hindi speakers with college level education for the an- notation task.
We first selected two random stories from our corpus and had the three annotators work on them independently and classify each sentence based on the discourse mode.
Please refer to this paper for detailed information: https://www.aclweb.org/anthology/2020.lrec-1.149/

Who are the source language producers?

Please refer to this paper for detailed information: https://www.aclweb.org/anthology/2020.lrec-1.149/

Annotations

Annotation process

The authors chose against crowd sourcing for labeling this dataset due to its highly sensitive nature.
The annotators are domain experts having degress in advanced clinical psychology and gender studies.
They were provided a guidelines document with instructions about each task and its definitions, labels and examples.
They studied the document, worked a few examples to get used to this annotation task.
They also provided feedback for improving the class definitions.
The annotation process is not mutually exclusive, implying that presence of one label does not mean the absence of the other one.

Who are the annotators?

The annotators were three native Hindi speakers with college level education.
Please refer to the accompnaying paper for a detailed annotation process.

Personal and Sensitive Information

[More Information Needed]

Considerations for Using the Data

Social Impact of Dataset

As a future work we would also like to use the presented corpus to see how it could be further used in certain downstream tasks such as emotion analysis, machine translation, textual entailment, and speech sythesis for improving storytelling experience in Hindi language.

Discussion of Biases