Abstract
A dataset of 2,686,354 reviews and 12,353,382 sentences posted on Airbnb from 15 cities in the United States is studied. The primary objective is to use a sentiment classifier to quantify the percentage of positive and negative reviews and sentences. Results indicate that 98.1% of reviews and 76.4% of the sentences are positive while only 1.06% of the reviews and 4.7% of the sentences are negative. A major contribution of this work is the provision of review datasets and sentences and their sentiments as identified by a classifier. These datasets can be used in other studies to examine several research questions related to the sharing economy. To increase reliability and credibility of these datasets, extensive evaluation of the performance of the sentiment classifier when used on data from Airbnb was conducted. Results of these evaluation procedures show that the sentiment classifier performs with relatively high precision.
Recommended Citation
Alsudais, Abdulkareem and Teubner, Timm, "Large-Scale Sentiment Analysis on Airbnb Reviews from 15 Cities" (2019). AMCIS 2019 Proceedings. 29.
https://aisel.aisnet.org/amcis2019/data_science_analytics_for_decision_support/data_science_analytics_for_decision_support/29
Large-Scale Sentiment Analysis on Airbnb Reviews from 15 Cities
A dataset of 2,686,354 reviews and 12,353,382 sentences posted on Airbnb from 15 cities in the United States is studied. The primary objective is to use a sentiment classifier to quantify the percentage of positive and negative reviews and sentences. Results indicate that 98.1% of reviews and 76.4% of the sentences are positive while only 1.06% of the reviews and 4.7% of the sentences are negative. A major contribution of this work is the provision of review datasets and sentences and their sentiments as identified by a classifier. These datasets can be used in other studies to examine several research questions related to the sharing economy. To increase reliability and credibility of these datasets, extensive evaluation of the performance of the sentiment classifier when used on data from Airbnb was conducted. Results of these evaluation procedures show that the sentiment classifier performs with relatively high precision.