detecting novelty seeking from online travel reviews

detecting novelty seeking from online travel reviews

1. DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING (DATA SCIENCE) MAJOR PROJECT on DETECTING NOVELTY SEEKING FROM ONLINE T RAVEL REVIEWS: A DEEP LEARNING APPROACH HOD Dr.L.CHANDRA SEKHAR REDDY Assistant Professor P.LAXMI PRASANNA WAJEEHA SAMREEN SHAFIA AHMADI 21H51A6744 21H51A6746 21H51A6720 Seminar coordinator Guide Mr. M. PARAMESWAR Mrs VENKATA LAKSHMI Assistant Professor Assistant Professor CMR COLLEGE OF ENGINEERING & TECHNOLOGY (UGC AUTONOMOUS) KANDLAKOYA, MEDCHAL ROAD, HYDERABAD – 501401. Affiliated to JNTU Hyderabad, Approved by AICTE New Delhi & Accredited by NAAC with ‘A+’ Grade

2. TABLE OF CONTENTS ABSTRACT INTRODUCTION LITERATURE SURVEY PROBLEM DEFINITION EXISTING SYSTEMS PROPOSED SYSTEM ADVANTAGES OF PROPOSED SYSTEMS DESIGN ALGORITHMS IMPLEMENTATION STEPS EXPECTED OUTPUT 1

3. Online travel reviews provide insights into traveler preferences and personality traits, particularly Novelty-Seeking (NS), which influences tourism motivation and destination choices. However, manual analysis is inefficient due to high data volume. This study proposes a deep learning-based approach using BERT ( Bidirectional Encoder Representations from Transformers) - BiGRU ( Bidirectional Gated Recurrent Unit) to automatically detect NS traits from 3,000 TripAdvisor reviews. A multi-dimensional classification framework categorizes NS into four dimensions: relaxation seeking, experience seeking, arousal seeking, and boredom alleviation. The model is expected to achieve high accuracy, enabling better tourism marketing, personalized recommendations, and behavioral analysis, demonstrating the power of deep learning in personality detection. ABSTRACT 2

4. INTRODUCTION Traveler preferences play a vital role in personalized recommendations and marketing in tourism. Novelty-Seeking (NS), which reflects a traveler’s desire for new experiences, significantly influences destination choices. Traditional survey-based methods are time-consuming and subjective, while manual analysis of large-scale travel reviews is impractical. To address this, NLP and deep learning provide an efficient automated solution. This project proposes a BERT- BiGRU model to classify NS traits from 3,000 TripAdvisor reviews. Using a multi-dimensional framework, it enables accurate personality detection, benefiting tourism marketing and personalized recommendations. 3

5. LITERATURE SURVEY 4 [1] Hu, Y., & Ritchie, J. R. B. Measuring destination attractiveness: A contextual approach. Journal of Travel Research, 32(2), 25–34 (1993). This foundational study introduces the concept of novelty in tourism and provides a framework for measuring destination attractiveness, which is critical for understanding novelty-seeking behavior . [2] Lee, T. H., & Crompton, J. Measuring novelty seeking in tourism. Annals of Tourism Research, 19(4), 732–751 (1992). A seminal work that develops a scale for measuring novelty-seeking behavior in tourism. It identifies four dimensions of novelty-seeking and is widely used in tourism research. [3] Devlin, J., Chang, M. W., Lee, K., & Toutanova, K. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. Proceedings of NAACL-HLT, 2019.

6. LITERATURE SURVEY 5 [4] Liu, B. Sentiment Analysis and Opinion Mining. Synthesis Lectures on Human Language Technologies, 5(1), 1–167 (2012). A comprehensive guide to sentiment analysis techniques, covering text preprocessing, feature extraction, and classification. This work is essential for analyzing travel reviews to detect novelty-seeking behavior . [5] Xiang, Z., & Gretzel , U. Role of social media in online travel information search. Tourism Management, 31(2), 179–188 (2010). Explores how social media and online reviews influence travel decisions. This study highlights the importance of analyzing user-generated content to identify patterns in tourist behavior , including novelty-seeking.

7. PROBLEM DEFINITION Manually analyzing large volumes of online travel reviews to identify Novelty-Seeking (NS) traits is impractical, time-consuming, and subjective. Traditional methods like self-report surveys are not scalable and prone to bias, while existing deep learning models struggle with accuracy in personality detection. To overcome these challenges, this project develops a BERT- BiGRU -based model to automatically classify NS traits from TripAdvisor reviews. 6

8. EXISTING SYSTEMS Manual Classification Introduction: - In this approach, human experts manually analyze travel reviews to identify Novelty-Seeking (NS) traits based on language patterns and expressions. Merits: - Can provide detailed and context-aware insights. - Ensures high interpretability as experts classify reviews based on experience. Demerits : - Time-consuming and not scalable for large datasets. - Prone to human bias and subjectivity leading to inconsistencies. 7

9. 8 2. Self-Report Surveys Introduction: - Travelers fill out psychological questionnaires to assess their NS traits based on predefined personality scales. Merits: - Structured and theoretically validated personality measurement. Easy to implement for small groups of participants. Demerits: Responses may be biased, as participants might misrepresent their personality. - Not suitable for large-scale and real-time personality detection

10. PROPOSED SYSTEM This project proposes a deep learning-based system to automatically detect Novelty-Seeking (NS) traits from online travel reviews using BERT- BiGRU . The model processes 3,000 TripAdvisor reviews and classifies them into four NS dimensions: relaxation seeking, experience seeking, arousal seeking, and boredom alleviation. The system follows a structured approach: BERT extracts contextual embeddings from the reviews, capturing the meaning of words based on their context. These embeddings are then processed by BiGRU , which analyzes sequential patterns in text to classify reviews based on NS personality traits. By automating this process, the system offers higher accuracy, scalability, and efficiency compared to traditional methods. It enables personalized travel recommendations, targeted marketing strategies, and enhanced user behavior analysis, making it a valuable tool for tourism analytics. 9

11. ADVANTAGES OF PROPOSED SOLUTION High Accuracy – The BERT- BiGRU model provides precise classification of NS traits, outperforming traditional methods. Scalability – The system can efficiently process *large datasets, making it suitable for real-world applications. Automation – Eliminates manual effort and human bias, enabling faster and more reliable personality detection. Personalization – Helps in customizing travel recommendations based on traveler preferences, improving user experience . 10

12. ALGORITHMS 1.BERT (Bidirectional Encoder Representations from Transformers) 11 BERT is a state-of-the-art transformer model designed for natural language processing tasks. It employs a bidirectional attention mechanism, meaning it processes both the preceding and succeeding words in a sequence to better understand context. BERT generates contextual embeddings that capture the nuanced meaning of words within sentences, which is crucial for accurately identifying psychological traits like NS. These embeddings allow the model to grasp deeper semantic relationships, ensuring more precise classification.

13. 2. BiGRU (Bidirectional Gated Recurrent Unit) BiGRU is a variant of RNNs that improves upon standard RNNs by addressing issues like vanishing gradients. It processes sequences in both forward and backward directions, capturing dependencies over long sequences. BiGRU uses gates (update and reset) to control the flow of information, enabling the model to maintain critical features while discarding irrelevant information. This makes BiGRU highly effective in handling sequential data, especially when applied after BERT’s feature extraction process. . 12 ALGORITHMS

14. DESIGN 13

15. 14 IMPLEMENTATION STEPS Data Collection & Preprocessing – Travel reviews from TripAdvisor are collected, cleaned, and preprocessed by removing noise, stopwords , and irrelevant text. Feature Extraction using BERT – BERT converts text into meaningful embeddings, capturing the context and semantics of words. Classification using BiGRU – The BiGRU model processes these embeddings, analyzing sequential text patterns to classify reviews into NS dimensions. Model Training & Testing – The system is trained on labeled data, evaluated using precision, recall, and F1-score, and optimized for high accuracy. Prediction & Analysis – The trained model predicts NS traits from new travel reviews, enabling applications in tourism marketing and recommendation systems

16. 15 EXPECTED OUTPUT Accurate NS Trait Classification- The system should correctly classify travel reviews into four Novelty-Seeking (NS) dimensions: relaxation seeking, experience seeking, arousal seeking, and boredom alleviation. High Performance Metrics – The model is expected to achieve high precision, recall, and F1-score, ensuring reliable classification. Comparison Graph – A performance comparison graph should be generated, showcasing the accuracy and efficiency of the BERT- BiGRU model against other deep learning models like CNN and LSTM

17. REFERENCES [1] Hu, Y., & Ritchie, J. R. B. Measuring destination attractiveness: A contextual approach. Journal of Travel Research, 32(2), 25–34 (1993). [2] Lee, T. H., & Crompton, J. Measuring novelty seeking in tourism. Annals of Tourism Research, 19(4), 732–751 (1992) [3] Zhang, K., Chen, Y., & Li, C. Discovering the tourists’ behaviors and perceptions in a tourism destination by analyzing photos and geotags posted on social media 11, 132–143 (2019). [4] Devlin, J., Chang, M. W., Lee, K., & Toutanova, K. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. Proceedings of NAACL-HLT, 2019. [5] Hochreiter, S., & Schmidhuber , J. Long Short-Term Memory. Neural Computation, 9(8), 1735–1780 (1997).. . 16

18. THANK YOU !!