Skip to main content
Have a personal or library account? Click to login
Leveraging Large Language Models for Aspect-Based Sentiment Analysis: A Restaurant Recommendation System for Entrepreneurs in Lisbon Cover

Leveraging Large Language Models for Aspect-Based Sentiment Analysis: A Restaurant Recommendation System for Entrepreneurs in Lisbon

By:  and    
Open Access
|Aug 2026

Figures & Tables

Table 1:

Dimensions and main attributes

DimensionCriteria
TangibilityPhysical appearance of facilities and equipment;
Staff uniforms and appearance;
Menu presentation.
ReliabilityPreparation and service times;
Accuracy of orders and billing;
Consistency of dishes, flavour, and service.
ResponsivenessStaff availability;
Fast service;
Handling of special requests.
AssuranceAccurate and transparent information;
Staff training and experience;
Ease of interaction with staff.
EmpathySensitivity to customer needs;
Ability to anticipate needs;
Support in case of problems.

[i] Adapted from (Stevens et al.,1995)

Table 2:

Consolidated attributes for Aspect-Based Sentiment Analysis

AttributeDescription
1. Food TasteAssesses the quality of food flavour, including authenticity and balance of ingredients.
2. Food PortionRefers to the quantity served, considering appropriateness in terms of price and expectations.
3. ServiceSeeks speed, efficiency, courtesy, and professionalism of the staff.
4. PriceEvaluates the adequacy of prices in relation to the quality of dishes and service.
5. Ambiance/AtmosphereConsiders decoration, lighting, music, and spatial arrangement.
6. Food PresentationObserves visual preparation, care, and creativity in the presentation of dishes.
7. Nutritious FoodAnalyses the nutritional quality of food options, such as the use of fresh ingredients and preparation methods.
8. Restaurant ReputationBased on reviews, recommendations, awards received, and public recognition of the restaurant.
9. CleanlinessAssess the hygiene of the restaurant, including dining and food preparation areas.
10. Variety of Healthy MealsMeasures the diversity of healthy options available on the menu, as well as other menu alternatives.

[i] Adapted from (Choi et al., 2009)

Table 3:

List of keywords and frequencies

AttributeKeywordsFrequency
1. Food Tastetasty, delicious, seasoning, authentic, nice, balanced.112
2. Food Portionportion, quantity, sufficient, generous, large.119
3. Serviceservice, quick, efficient, friendly, kind, professional.312
4. Priceprice, cost, expensive, cheap, affordable.264
5. Ambiance/Atmosphereambience, atmosphere, cosy, decoration, lighting, music.217
6. Food Presentationpresentation, visual, plating, creative, decorated.45
7. Nutritious Foodnutritious, healthy, fresh, natural ingredients.98
8. Restaurant Reputationfamous, recommended, awarded, well-known, reputation.19
9. Cleanlinessclean, hygienic, tidy, cleanliness.7
10. Variety of Healthy Mealsvariety, healthy options, healthy menu, alternatives.53
Table 4:

Agreement Rates Between Models and Human Evaluator

LLM modelAgreement Rate
ChatGPT 3.5 Turbo64,4%
ChatGPT 4o_2024-05-1381,6%
Mistral 7B Instruct Free60,6%
Figure 1:

Prototype visual components (a) map of the locations of competing restaurants within a chosen category; (b) information and summarised sentiment analysis results for a specific restaurant

Figure 2:

Radar chart comparison of multiple restaurants

DOI: https://doi.org/10.2478/ejthr-2026-0001 | Journal eISSN: 2182-4924 | Journal ISSN: 2182-4916
Language: English
Page range: 1 - 14
Submitted on: May 8, 2025
Accepted on: Jul 22, 2025
Published on: Aug 17, 2026
Published by: Polytechnic Institute of Leiria
In partnership with: Paradigm Publishing Services
Publication frequency: 2 issues per year

© 2026 Paulo Carrasco, Pedro Esteves, published by Polytechnic Institute of Leiria
This work is licensed under the Creative Commons Attribution 4.0 License.