Dataset summary
This dataset contains 1789 data instances with problem identification, missing resource, time-dependent questions and answers pairs for disaster management. The prominent fields of interest are:
- problem: a description of the main problem a human may face in the incident depicted in the image
- context: A textual description of the incident situation. It is automatically generated by captioning an image from Incidents 1M
- missing resource: the main missing resource (service, amenity) in the described incident situation
- solutions: a set of three sentences, one for each timeframe (immediately, after one week, after few days) addressing how to replace or deal with the missing resource while dealing with the problem identified earlier.
- explanations: natural language rationale to justify the validity of the solution
- scenario : list of natural language questions automatically generated, one per each timeframe, incident and place labels. Additionally, the fields with the metadata are:
- id: an identifier form the source data corresponding to the image id.
- ann_id : an identifier for the set of annotators, either 0 or 1. The tuple (ann_id, id) is a unique identifier for the dataset instance. If you are having problems with downloading the dataset from the huggingface hub, please download it from here.
Languages
English
Additional Information
You can download the data via:
from datasets import load_dataset
response_dataset = load_dataset(adial/response)
Please find more information about the code and how the data was collected on here.
Licensing Information
Creative Commons Attribution Non Commercial Share Alike 4.0
Tags
annotations_creators:
- crowdsourced
language:
- en
language_creators:
- crowdsourced
- machine-generated
multilinguality:
- monolingual
tags:
- qa
- disaster management
- commonsense
task_categories:
- question-answering
task_ids:
- open-domain-qa