Exploring Emotions in Agritourism:
A Sentiment Analysis of Online Reviews in Campania, Italy

Autor: Luisa Fernanda Arenas Estevez
Tutor: Teresa Del Giudice | CoTutor: Roberta Garibaldi
December - 2024

1. Introduction

• Context:
  Rural tourism has emerged as a key driver of local economic development and the revitalization of rural areas facing demographic and economic challenges. A distinctive model within rural tourism in Italy is Agritourism, a policy aimed at diversifying the agricultural economy. This form of tourism is exclusively carried out on agricultural holdings.

• Problem:
  Although agritourism is important for diversifying rural economies and supporting local development, there is limited understanding of how tourists perceive these experiences and the emotional value they derive from them. Understanding these perceptions is critical for improving services and evaluating the effectiveness of this public policy.

• Justification::
  In the context of a postmodern consumer, decision-making is influenced not only by functional and economic factors but also by emotions and meaningful experiences. Understanding these emotional drivers is essential to fully grasp consumer behavior. Analyzing online reviews allows us to capture these emotions and perceptions, providing valuable insights to enhance agritourism services and promote sustainable rural tourism.

• General Objective:
  

To systematically analyze the experiences of agritourism consumers in the Campania region using Sentiment Analysis (SA) on online reviews.

• Specific Objectives:
  

  • Detect emotional and thematic patterns.
  • Analyze and quantify the emotions expressed in the reviews.
  • Identify the general polarity of sentiments (positive or negative) associated with each location.

• Guiding Questions:
  

  • What aspects do tourists discuss the most in their agritourism experiences in Campania?
  • What trends and topics emerge in these reviews?
  • What emotions do tourists experience, and how are these emotions related to specific aspects of their experience?
  • How do emotions and perceptions vary across provinces?

2) Framework:

The Role of Emotions in Tourism

Emotions as Key Drivers:

• Emotions influence travelers’ decisions, perceptions, and satisfaction. Tourism involves multisensory and complex experiences beyond tangible goods (Li et al., 2015).
• Designing tourism experiences requires emotions as a central focus to enhance loyalty and recommendations (Afshardoost & Eshaghi, 2020; Volo, 2021).

Challenges in Measuring Emotions:

• Traditional methods (e.g., self-reports) are limited by cognitive biases and socially desirable responses.
• Structured formats can prevent genuine, spontaneous information (Li et al., 2015).

User-Generated Content (UGC):

• Social networks and digital platforms allow authentic capture of emotions through unstructured content.
• UGC reflects travelers’ perceptions with greater expressive power.

Sentiment Analysis in Tourism Research:

• Since 2016, SA research has grown significantly in tourism, led by China and the USA (Manosso & Ruiz, 2021).

Applications in Italy:

• Evaluation of tourist satisfaction: Examples: Albergo Diffuso (Vallone & Veglio, 2019), Wine tours in Tuscany (Barbierato et al., 2022).

• Analysis of destination reputation: Mount Etna (Graziano & Albanese, 2020), Uffizi Gallery (Collini et al., 2023).

• Assessment of critical events (e.g., COVID-19 impact) (Peña et al., 2021).

Opportunities for Agritourism:

• Limited use of SA in Agritourism
• Potential to understand emotions and perceptions shaping authentic and sustainable tourism experiences.

3) Materials and Methodology:

• 3.1 Data Collection

• 3.2 Methodology

3.2.1 Data Preprocessing

Key Steps:

• Stopwords Removal: Elimination of non-meaningful words (e.g., articles, prepositions).
• Custom Stopwords: Added “agriturismo” and other terms (solo, cosa, etc.).
• Data Cleaning: Remove numbers, punctuation, and spaces.
• Tokenization: Split text into individual words.

Example:

• Original Review: E un posto bellissimo elegante rilassante Per una giornata all insegna del relax non c è di meglio

• Clean Review: bellissimo elegante rilassante insegna relax meglio

The final result of this process was a set of unique elements or words \(𝑊 ={𝑤_1,𝑤_2,…,𝑤_𝑚}\)

where \(𝑤_𝑗∈ 𝑊\) corresponds to each unique word after text cleansing, and \(𝑚= 17122\) is the total number of unique words.

and the set of reviews \(𝑑\), where \(𝑑= {𝑑_1,_2,…,𝑑_𝑛}\) and \(n=4047\)

3.2.2 Frequency and Topic Analysis

• Document-Term Matrix (DTM)

DTM organizes the frequency of words across documents is defined as a matrix \(M\) with dimensions \(n \times m\):

\[ M = \begin{bmatrix} m_{1,1} & \cdots & m_{1,m} \\ \vdots & \ddots & \vdots \\ m_{n,1} & \cdots & m_{n,m} \end{bmatrix} \]

Where \(M_{ij}\) indicates the frequency of word \(w_j\) in review \(d_i\).

Example of a Mini DTM

• Topic Modeling with Latent Dirichlet Allocation
  
  • A probabilistic model used for identifying hidden topics in a corpus
  • Assumes each document (review) is a mixture of topics.
  • Each topic is represented by a distribution of words.

Topic Distribution per Document (\(\Gamma_{ik}\)):

\[ \Gamma_{ik} = \{\Gamma_{i1}, \Gamma_{i2}, \dots, \Gamma_{nq} \} \]

where \(k\) represents the topics, and \(\Gamma_{ik}\): is the probability that the review \(d_i\) belongs to topic \(k = {1,..𝑞}\).

Word Distribution per Topic (\(\phi_{jk}\)):

\[ \phi_{jk} = \{\phi_{j1}, \phi_{j2}, \dots, \phi_{mq} \} \]

where \(\phi_{jk}\) is the probability that the word \(w_j\) is associated with topic \(k\).

3.2.3 Sentiment Analysis (SA)

Sentiment Analysis identifies and quantifies the emotions expressed in textual data.
Tools: Syuzhet Package & NRC EmoLex Lexicon (Plutchik’s emotions).

Emotions Analyzed

Emotion	Meaning
Joy	Happiness, pleasure, satisfaction.
Sadness	Loss, disappointment, unmet expectations.
Anger	Frustration, injustice, hostility.
Fear	Perception of insecurity or threat.
Trust	Safety, acceptance, positive connections.
Disgust	Rejection of unpleasant aspects (e.g., hygiene, quality).
Surprise	Reactions to unexpected events (positive or negative).
Anticipation	Expectations, curiosity, excitement for future events.

Polarity Classification : Words categorized into : Positive: - Negative:

4) Results

The analysis was based on a total of 4,047 reviews collected from agritourisms in the Campania region, distributed across the five provinces of the region. These reviews cover a total of 251 communes

4.1 Frequency Analysis

Topic Identification

• Topic 1. Gastronomy: This topic specifically addresses culinary experience, a recurring theme in the reviews. Words such as qualità, locale, menù, piatti, vino, and antipasti highlight aspects such as menu diversity and food quality.

• Topic 2. Environment: This topic emphasizes visitors’ connection to nature and the relaxing environment offered by agritourisms. Associated with terms such as natura, verde, immerso and relax.

• Topic 3. Authenticity: This topic highlights the authenticity and quality of local products, as well as personalized interactions with the owners. Words such as prodotti, sapori, genuini, and proprietari emphasize the relevance of local foods, as well as the hosts’ hospitality.

• Topic 4. Service: This topic is dominated by terms related to hospitality and service quality. Words such as personale, gentile, accogliente and grazie reflect visitors’ appreciation for the friendly treatment and welcoming atmosphere.

• Topic 5. Facilities: This topic focuses on the physical facilities of agritourisms. Terms such as struttura, piscina, camere, bagno, and parcheggio highlight the amenities available, including accommodation spaces.

Topic Identification by Province

4.2 Sentiment Analysis

The words were classified into the eight basic emotions proposed by Plutchik. The results reveal that words related to trust are the most predominant (72.55%), followed by joy (50.32%) and anticipation (35.06%). In contrast, words linked to anger (8.89%) and disgust (8.70%) are the least frequent.

Distribution of Sentiments by Province

When analyzing the distribution of emotions by province, a general homogeneity in proportions is observed

Wordcloud by Emotion

Polarity

the sentiment analysis evaluated the valence within each review, categorizing words into two broad categories: positive and negative. The overall results show that 83.76% of the words analyzed were associated with a positive valence, while 16.24% corresponded to a negative valence

Conclusions

• Key Insights:
  

  • The analysis highlights gastronomy as a central element in agritourism experiences in Campania.
  • Positive emotions such as trust, joy, and anticipation predominate, reinforcing the overall satisfaction of visitors.
  • Service quality emerges as the predominant topic in reviews, highlighting an area of focus for improving visitor satisfaction in Campania.

• Practical Implications:
  

  • For Businesses: Understanding visitor emotions can help improve service quality, particularly in hospitality and gastronomy.
  • For Policymakers: Identifying strengths and weaknesses at the commune and provincial levels supports targeted investments and sustainable tourism development.

• Limitations:
  

  • Use of the EmoLex lexicon may have introduced translation inconsistencies, affecting emotion classification.
  • The study focuses exclusively on the Campania region, potentially limiting variability and generalization of findings.

• Future Directions:
  

  • Expand the dataset to include agritourism reviews from other Italian regions to capture cultural and geographical variations.
  • Compare findings using lexicons specifically designed for the Italian language.
  • Explore correlations between emotional metrics and other quantitative variables such as visitor demographics or economic impact indicators.

Muchas Gracias | Grazie mille | Thank you