Multilabel multiclass sentiment and emotion dataset from indonesian mobile application review

Area of Science:

• Natural Language Processing (NLP)
• Data Science
• Computational Linguistics

Background:

• Mobile app reviews offer valuable user feedback on application performance and aesthetics.
• User reviews contain rich textual data reflecting sentiments and emotions, suitable for dataset creation.
• Existing Indonesian textual datasets for multi-label, multi-class sentiment analysis are limited.

Purpose of the Study:

• To create a novel multi-label, multi-class Indonesian-language dataset from public mobile app reviews.
• To annotate the dataset with sentiment (positive, negative, neutral) and emotion (anger, fear, sad, happy, love, neutral) values.
• To support and advance sentiment analysis and text classification research in Indonesian.

Main Methods:

• Collection of public mobile application reviews.
• Data pre-processing, including cleaning and handling.
• Annotation of reviews with 3 sentiment categories and 6 emotion categories.

Main Results:

• A comprehensive Indonesian dataset of mobile app reviews was successfully generated.
• The dataset is annotated for multi-label, multi-class sentiment and emotion analysis.
• The dataset addresses the scarcity of Indonesian resources for advanced text classification tasks.

Conclusions:

• The developed dataset is a valuable resource for NLP research, particularly for Indonesian sentiment and emotion analysis.
• This work contributes to the field by providing a much-needed dataset for multi-label text classification.
• The dataset facilitates further research into understanding user emotions and sentiments expressed in mobile app reviews.