Absabank-Imm

The Swedish ABSAbank-Imm 1.1 is an annotated corpus designed for aspect-based sentiment analysis related to immigration in Sweden. Let's delve into the details:

  1. Title and Subtitle:

    • Title: Swedish ABSAbank-Imm v1.1
    • Subtitle: An annotated Swedish corpus for aspect-based sentiment analysis (a version of Absabank)

  2. Creation and Purpose:

    • Created by: Aleksandrs Berdicevskis
    • Purpose: To analyze sentiments expressed toward immigration in Sweden.
    • Description: The dataset is a subset of the original Swedish ABSAbank, specifically tailored for aspect-based sentiment analysis related to immigration. In this subset, paragraphs are manually labeled with sentiment scores on a scale from 1 (very negative) to 5 (very positive).

  3. Dataset Details:

    • Size: It contains 4872 short texts (paragraphs).
    • Language: Swedish
    • Tokens: Approximately 199,000 tokens.
    • Content: The original ABSAbank had two layers of annotation: token-level and text-level. In ABSAbank-Imm, only the paragraph-level annotation is preserved. Annotators labeled paragraphs related to immigration with sentiment values.

  4. Usage:

    • Applications: Machine Learning, Aspect-based Sentiment Analysis, Stance classification, Evaluation of language models.
    • Intended Task: Given a text or paragraph, label the sentiment expressed toward immigration in Sweden.
    • Recommended Evaluation Measures: Krippendorff's alpha (official SuperLim measure), Spearman's rho, or other correlation coefficients.
    • Recommended Split: Consecutive split into train, dev, and test sets.

  5. Citation and Related Datasets:

    • Cite as: Consider citing ¹.
    • Related Datasets: Part of the SuperLim collection (derived from ABSAbank) ¹.

(1) Svensk ABSAbank-Imm 1.1 | Språkbanken Text - Göteborgs universitet. https://spraakbanken.gu.se/resurser/absabank-imm. (2) ScandEval/absabank-imm-mini · Datasets at Hugging Face. https://huggingface.co/datasets/ScandEval/absabank-imm-mini. (3) ScandEval/absabank-imm-mini at main - Hugging Face. https://huggingface.co/datasets/ScandEval/absabank-imm-mini/tree/main. (4) undefined. https://spraakbanken.gu.se/en/resources/superlim. (5) undefined. https://spraakbanken.gu.se/en/resources/swe-absa-bank.