Analyzing Hotel Reviews: Sentiment Analysis and Topic Modeling for Summarization
Click the Poster to View Full Screen, Right click to save image
Grant: Freshman Research Initiative
Olivia Tirso
College:
The Dorothy and George Hennings College of Science, Mathematics, and Technology
Major:
Computer Science
Faculty Research Advisor(s):
Daehan Kwak
Abstract:
Hotel reviews play a crucial role in shaping perceptions, informing decision-making, and influencing the overall guest experience. In many cases, it is impossible to go over all the reviews to make an informed decision. By summarizing these reviews, we can identify patterns which can help hotels understand their strengths, as well as finding solutions for areas in which consumers saw faults. Using sentiment analysis, it reads an Excel file with review data, performs sentiment analysis using the Vader sentiment analyzer, checks for the presence of specific target words in the reviews, and saves the results in a new Excel file for further analysis. Using nltk’s wordnet lemmatizer, it lemmatized and performed word counting on text data, stored in a data frame. It's useful for analyzing the frequency of different lemmatized words in a text dataset.
The results display vader scores of neutral, positive, and negative from target words that were extracted from the dataset. As well as calculating the word frequencies in a specific data frame column and then creating and displaying a word cloud visualization based on those frequencies using Python libraries like Pandas, WordCloud, and Matplotlib. The word cloud visually represents the most common words in the text data. Alongside each written review, included is a chart displaying neutral, positive, and negative scores. Additionally, a rating column is provided to classify the reviews as bad, good, or neutral based on their respective scores.
This research represents a substantial contribution to the expansive field of sentiment analysis and topic modeling. It engages in a comprehensive exploration of the practical applications of these advanced techniques, with a specific focus on their deployment within the intricate landscape of hotel reviews. By examining and dissecting the intricacies of sentiment analysis and topic modeling within this specific domain, this study not only enhances our understanding but also provides a blueprint for their successful implementation in a broader spectrum of industries and domains. It explores the application of these techniques specifically in the context of hotel reviews, which can serve as a blueprint for similar analysis in other industries. The knowledge acquired through this research extends beyond the hospitality industry, empowering businesses across diverse sectors to leverage customer reviews for informed decision-making based on data.