Analysing top hospitality values mentioned in Accor Hotels reviews

This project was completed as part of my cursus at Ironhack (a 9-week intensive coding bootcamp).

The objective of this project was to understand which hospitality values are the most sought-after by the customers of Accor Hotels premium brands, from the scrapping of online hotel reviews.

PROJECT SCOPE

In total, 11,694 reviews on 265 hotels were scrapped from a well-known hotel review website.

• Locations: Top 10 French cities
  • Paris, Marseille, Lyon, Toulouse, Nice, Nantes, Strasbourg, Montpellier, Bordeaux, Lille
• Brands: Accor Hotels
  • Premium brands: Raffles, Sofitel, Pullman, MGallery
  • Intermediary brands: Novotel, Mercure, Adagio, Mama Shelter
  • Economic brands: Ibis, hotelF1
• Values: Top 10 hospitality values, each one of them being characterized by a set of keywords (keywords can be found in the data.py file of this repository, under the values_dict variable)
  • Authenticity, Adventure, Community, Creativity, Cool, Fairness, Friendliness, Inner Harmony, Instantaneity, Love, Indulgement, Safety, Service, Status, Trustworthiness

METHODOLOGY

The scraping pipeline was composed of three main steps:

• STEP #1 - Acquisition: scrapping of hotel reviews

  • 1 hotel_table per city, including city, hotel name, hotel rating, number of hotel reviews and hotel url
  • 1 filtered_hotel_table per city, removing non-Accor Hotels brands from the scraping output
  • 1 review_table per city, inclidung city, hotel name, individual rating, review text

• STEP #2 - Preparation of collected data

  • 1 scrapped_table for France: aggregation of the filtered_hotel_tables and review_tables of all cities, filtered on the selected range of Accor Hotel brands (premium, intermediary or economic)
  • 1 study_table: transformed scrapped_table, featuring a count of values occurrences

• STEP #3 - Reporting

  • 1 bar chart summarizing the percentage of total reviews mentioning each one of the hospitality values

REPOSITORY STRUCTURE

This repository is composed of three types of Python files:

• Project resources:
  • libraries.py
  • data.py (on Top French cities, Accor Hotel brands and hospitality values)
• Executable pipelines:
  • pipeline-acquisition.py
  • pipeline-preparation-reporting.py
• Functions used in executable pipelines:
  • functions-acquisition.py
  • functions-preparation-reporting.py

A directory containing pipeline outputs was also added to this repository (scraping date: 2019.01.25). The pipeline-outputs directory contains:

• the output tables of the acquisition step
• the output tables of the preparation step and associated bar charts, on two distinct study perimeters: premium brands, and intermediary brands

HOW DID WE SPOT VALUES IN HOTEL REVIEWS?

This project was a great opportunity to get acquainted with Natural Language Processing (NLP) terminology and associated Python library, nltk.

To detect values occurrences in hotel reviews, two parallel workflows were performed:

• On values keywords: stem values keywords (i.e. extract the root of the words) with the nltk library
• On hotel reviews: original reviews were transformed in a three step process: 1. translate reviews from French to English with the googletrans library, as nltk achieves a better performance on English texts, 2. remove stop words and punctuation with the nltk library, and 3. stem transformed reviews with the nltk library.

As a final step, the stemmed forms of values keywords and hotel reviews were matched to count values occurrences.

RESULTS

• Regardless of the hotel range, friendliness, instantaneity and service/creativity are the Top 4 values that are the most sought-after by Accor Hotels customers.
• The service value seem to have a greater importance for the customers of premium brands than intermediary brands (appears respectively in 48% vs. 30% of hotel reviews).
• Quite logically, intermediary customers also seem to be more attentive to price fairness, while premium customers give a greater importance to status.

The two below charts show the proportion of hotel reviews (in %) mentioning each one of the values:

Analysis conducted on 949 reviews of premium Accor Hotels brands:

Analysis conducted on 4,575 reviews of intermediary Accor Hotels brands:

MAIN LIBRARIES

Requests, BeautifulSoup, GoogleTranslate, NLTK