← All proceedings

TheIIER INTERNATIONAL CONFERENCE

Proceedings cover: TheIIER INTERNATIONAL CONFERENCE

Madrid, Spain 10-08-2025

6 papers

  1. Local Benchmarking of Sentence Embedding Models for Retrieval-Augmented Generation

    Murat Karakaya

    Paper ID II-CSISMDRD-100825-29311 · Pages 1-6

    Abstract

    Retrieval-Augmented Generation (RAG) systems heavily depend on high-quality embedding models to effectively retrieve semantically relevant text chunks from document corpora. Current benchmarking practices predominantly utilize static, general-purpose datasets, which may inadequately represent the nuances and specificities of domain-focused applications. To address this limitation, we introduce a fully automated benchmarking pipeline enabling practitioners to evaluate sentence-transformer embedding models directly on their customized document collections. Our system leverages synthetic query generation coupled with LLM-based automated relevance judgments, thus simulating realistic retrieval scenarios without manual annotation efforts. In our study, we benchmarked several prominent Sentence Transformers embedding models on a specialized technical corpus, rigorously analyzing their retrieval performance using five distinct metrics. The proposed pipeline further provides detailed statistical comparisons, visual performance diagnostics, and practical throughput assessments, significantly aiding in embedding model selection for real-world RAG system deployments.

    Retrieval-Augmented Generation, Embedding Models, Sentence Transformers, Information Retrieval, Benchmarking, Synthetic Queries.

    Download PDF
  2. WAYVORA: An AI-Powered Mobile Travel Planner Integrating Real-Time Prices and User Preferences

    Abdullah Mahmood Najmnajm, Yaseen Abdubaker Sadeq, Guzin Turkmen

    Paper ID II-CSISMDRD-100825-29904 · Pages 7-13

    Abstract

    WAYVORA is an AI-driven mobile travel planner that uses a lightweight React-Native client and a cloud micro service stackcentered on Gemini 1.5 Flashto turn free-form dialogue into end-to-end, budget-aware itineraries. The system fuses live price feeds, preference profiles, and a multi-objective optimizer so that suggestions satisfy spending caps, timing constraints, and user interests in a single conversational loop. In a two-stage study with 24 participants, WAYVORA enabled travelers to complete identical planning tasks 31 % faster than Google Trips (p < 0.01), achieved a System Usability Scale score of 78.5 versus 68.2, and kept itineraries within budget in 87 % of trials while earning a mean satisfaction rating of 4.4/5. Performance profiling on mid-range Android devices showed a median end-to-end latency of 610 ms over 4 G and a 19 MB APK footprint, demonstrating practical deployability. These results suggest that tightly integrating conversational LLM reasoning with cost-conscious optimization can streamline mobile travel planning and offer a viable blueprint for next-generation e-tourism services.

    Keywords - Budget-Aware Itinerary Generation; Conversational LLM Interface; Multi-Objective optimization; User-Centered Evaluation

    Download PDF
  3. The Issue of Creativity, Creative Thinking, Skill and Image in the Work of a Television Journalist

    Sevinch Khushnazarova

    Paper ID II-ELTMDRD-100825-005 · Pages 14-16

    Abstract

    This article discusses what a television journalist should be today, and the high and reliable demands placed on representatives of this field today. The article also contains quotes from the opinions and comments of national and foreign journalists based on scientific foundations.

    Keywords: Image, Creative Thinking, Television Journalist, Camera, Portrait

    Download PDF
  4. Exploring the Role of Artificial Intelligence in Supply Chain Finance: A Systematic Review

    Maroua Zineelabidine, Abir Attahiri

    Paper ID II-ABMeBMDRD-100825-12 · Pages 17

    Abstract

    This paper provides a systematic review of the academic literature on the use of Artificial Intelligence (AI) in the field of Supply Chain Finance (SCF). As supply chains become increasingly complex and data-driven, AI technologies have emerged as powerful tools to enhance financial flows, optimize working capital, and improve decision-making across supply chain partners. Based on a rigorous selection of peer-reviewed articles published between 2010 and 2025, this study synthesizes key findings and classifies AI applications in SCF into four main categories: credit risk assessment and financing decisions, invoice and payment automation, fraud detection, and dynamic discounting and trade finance optimization. The review highlights how machine learning, natural language processing, and other AI techniques contribute to increased transparency, cost efficiency, and resilience in SCF practices. However, the analysis also reveals critical research gaps, including data governance, ethical concerns, regulatory constraints, and the need for industry-specific AI models. The paper concludes by proposing a future research agenda and offering managerial implications for leveraging AI to create more agile and financially sustainable supply chains.

    Keywords - Artificial Intelligence, Supply Chain Finance, Machine Learning, Risk Assessment, Invoice Automation, Trade Finance, Systematic Literature Review.

    Download PDF
  5. Insertion of the Nigerian Economy Into the Development Dynamics of the Brics Main Challenges

    Lic. Nwaebonyi Chukwuemeka Isaac, Zulma D. Ramirez Cruz, Yerenis Torres Cala, Dr. C. Roberto Munoz Gonzalez

    Paper ID II-SSEMDRD-100825-29488 · Pages 18

    Abstract

    Strengthening economic, commercial and financial relations between African countries and the BRICS countries constitutes a viable alternative considering the fact that BRICS countries represents a collective effort by emerging economies to create a more balanced world economic order and new forms of global governance.

    The inclusion of Nigeria as an associate country of the BRICS represents a socio-economic strength and a development opportunity for the African country.

    The objective of this research is to explain the main challenges for Nigeria's competitive insertion in the developmental dynamics of the BRICS countries.

    The research demanded the use of theoretical methods such as historical-logical, analytical-synthetic and inductive-deductive; empirical methods necessary for the analysis of documents, scientific observation, as well as the review and compilation of data.

    In general, the investigation carried out revealed the strengths and weaknesses that Nigeria has for its progressive and competitive insertion into the BRICS, but also the enormous opportunities to access financing, investments and competitive and fair markets, in the perspectives of the country's socio-economic development.

    It is concluded that the main challenges for competitive insertion in the BRICS are associated with the problems of structural economic deformation of the country, but also with inefficient processes of organization and management of the economy and its sectors, in the same way there are states of corruption that can hinder more dynamic advances for competitive insertion in the Group.

    Keywords: development, emerging countries, BRICS, Nigeria, international economy

    Abstract: (faithful English translation of the summary section and the same format as the previous one, adding italics

    Keywords: Development, Emerging Countries, BRICS, Nigeria, International Economics

    Download PDF
  6. Leveraging Transformer-Based Architectures for Multi-Label Complaint Detection and Sentiment Analysis in Hotel Feedback Systems

    Milena Nikolić, Nevena Minić, Marina Marjanović

    Paper ID II-CSISCANES-130825-002 · Pages 52-59

    Abstract

    In the competitive hotel industry, understanding guest feedback is essential for improving service quality. This paper proposes a comprehensive intelligent system that applies transformer-based architectures to perform sentiment analysis and multi-label classification of consumer complaints in hotel review data. Using BERT and RoBERTa, we fine-tune models for twotasks: classifying sentiment (positive, neutral, negative) and detecting multiple complaint categories such as cleanliness, staff behavior,food quality,and booking issues. A shared encoder supports both tasks, trained on an extended TripAdvisor dataset with additional evaluationon the Consumer Complaint Database from Kaggle. Comparative experiments show that the fine-tuned RoBERTa model outperforms BERT, achieving an F1 score above 0.928 for sentiment classification and 87% mean precision across complaint categories. The system also integrates SHAP and LIME to provide insights into model decisions. It enables transparentreview analysis, helping hotel managers detect issues promptly and enhance service quality.This solution supports scalable and interpretable feedback analytics for smart hospitality systems.

    Keywords - Intelligent Systems, Hotel Reviews, Sentiment Analysis, Classification, NLP, Deep Learning, Complaint Detection

    Download PDF