IndonesianFoodSME_CustomerReviews_UmayumchaProbolinggo_2026
Description
This dataset contains customer reviews of Umayumcha Probolinggo, a micro and small food enterprise (UMKM) in Indonesia. It includes reviewer names, rating scores (3–5 stars), review text, and the business location. The dataset is curated for research in text summarization using the IndoBERT model, providing a real-world corpus of Indonesian language customer feedback. Columns: Nama – Reviewer’s name Rating – Customer rating (3, 4, or 5) Teks_Ulasan – Review text in Indonesian Tanggal – Date of the review (if available) UMKM_Name – Business name (“Umayumcha Probolinggo”) Purpose: The dataset is intended to facilitate research in natural language processing (NLP), specifically for developing and evaluating automatic summarization models in Indonesian language customer reviews. It supports experiments in understanding and summarizing large volumes of consumer feedback to improve business insight and decision-making.
Files
Steps to reproduce
1. Download the Dataset Download the CSV file: IndonesianFoodSME_CustomerReviews_UmayumchaProbolinggo_2026.csv. 2. Load the Dataset Open the dataset using spreadsheet software such as Excel, or a data analysis environment like Python, R, or MATLAB. 3. Data Cleaning Check for and remove duplicate reviews. Remove rows without review text or other essential columns. Ensure that dates and ratings are consistent and properly formatted. 4. Text Preprocessing for NLP Clean the review text from special characters, symbols, or irrelevant emojis. Normalize informal or nonstandard words into standard Indonesian. Optionally, split text into sentences or paragraphs for easier processing. 5. Prepare for IndoBERT Model Apply IndoBERT tokenization to convert the text into model-readable representations. Prepare sequences of tokens to feed into the model for summarization. 6. Text Summarization Use an encoder-decoder transformer architecture with attention to generate summaries. Ensure that key words and the main context of each review are preserved so the summary remains coherent and informative. 7. Evaluation (Optional) If available, compare automatic summaries with manually created summaries for reference. Use text evaluation metrics such as ROUGE to measure how well the summaries retain essential information. 8. Output The cleaned dataset and generated summaries can be used for NLP research, model training, or analysis of customer reviews for SMEs. The resulting summaries help SMEs understand customer opinions and satisfaction efficiently.