Dataset for the article “AI-Generated Refined Summaries for Green Consumption: Mining and Optimizing the Key Features of Green Claims Using Large Language Models”

It is mainly used to examine the predictive effects of semantic features in green claims and non-green control claims on consumers’ purchase intention. The data were compiled from real market product claim texts and their matched non-green control texts, together with text-level purchase intention ratings obtained from the questionnaire study. After data cleaning, text pairing, semantic feature extraction, vector representation, and manual review, the original texts were transformed into a machine learning input table.

The data file is in Excel format and contains one worksheet named “Machine Learning Input Table.” The main variables include text ID, pair ID, text-level mean purchase intention, high versus low purchase intention classification, product category, text length, and three types of machine learning input features. Type A variables include 150 continuous features obtained after dimensionality reduction of the overall semantic representation of each text.

Type B variables include quantified scores for 153 interpretable semantic features. Type C variables include dummy-coded product category variables. Product categories include clothing and home goods, personal care and daily necessities, digital appliances, transportation, food and beverages, and stationery and packaging.Study 3 DataThis dataset contains the experimental data used to test the intervention effect of AIGRS in Study 3.

It is mainly used to examine whether intervention texts generated based on key semantic features can increase consumers’ purchase intention for green products. The data were collected from questionnaire experiments. The experimental materials included three product categories: bowls, batteries, and cars.

For each product category, both a control text and an intervention text were created, and participants’ purchase intention ratings under different text conditions were recorded. The data file is in Excel format and contains two worksheets, named “Study 3_Experiment 1” and “Study 3_Experiment 2.” Experiment 1 was used to test the positive intervention effect of AIGRS, whereas Experiment 2 was used to test the control effect of de-greenified parallel rewritten texts.

The main variables include participant ID, product presentation order, gender, age, occupation, monthly income, academic major, and the mean purchase intention for bowls, batteries, and cars under the control and intervention conditions. “Study 3_Experiment 1” contains data from 58 participants, and “Study 3_Experiment 2” contains data from 57 participants.