argureviews / README.md
nihiluis's picture
Update README.md
55e4d0f verified
metadata
dataset_info:
  features:
    - name: type
      dtype: string
    - name: text
      dtype: string
    - name: annotator
      dtype: string
    - name: component
      dtype: string
    - name: specificity
      dtype: string
    - name: sentiment
      dtype: string
    - name: aspect
      dtype: string
    - name: id
      dtype: string
    - name: sidx
      dtype: float64
  splits:
    - name: train
      num_bytes: 1328357
      num_examples: 7266
  download_size: 534609
  dataset_size: 1328357
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

Dataset Card for "argureviews"

Dataset for basic argumentation in online reviews

The dataset aims to annotate online review sentences for basic argumentative quality, sentiment and aspect of interest. It covers 1016 online reviews with 7286 sentences for the following domains: products from Amazon, local services, restaurant and hotels from Yelp and brokerage apps from the Google Play Store.

The label set descriptions are as follows. The respective DeBERTa models are linked as well.

  • Argument component: Distinguishes the argumentative component that is used. Can be one of: claim, premise, background.
  • Specificity: Differentiates between generic statements and more thoughtful user statements. Can be one of: general, specific, experience.
  • Sentiment: A positive, balanced, negative or neutral argumentative statement about the reviewed item.
  • Aspect: Provides more insight into what aspect of interest the argumentative statement covers. Can be one or more of: general sentiment, price, delivery, function and quality, fun and usage, style, installation, customer service and none. Only available for the Amazon review subset.