AUB ScholarWorks (American Univ. of Beirut)
Not a member yet
33634 research outputs found
Sort by
TIMELINEGPT: UTILIZING LLMS FOR AUTOMATIC TIMELINE GENERATION OF NEWS DATA
No full textThis thesis introduces a novel approach for automatic timeline generation for news data, an essential task in Information Retrieval. Our work addresses the challenge of generating timelines and multilingual scalability issues, mainly focusing on Arabic Timeline Generation, an under-resourced field in literature. Our approach utilizes the improved capabilities of Large Language Models(LLMs), particularly their accuracy and efficiency in text generation and summarization. We showcase the architecture of our system, highlighting the novel components that distinguish it from other approaches in the literature, particularly the integration of LLMs for event and date extraction of news sources, through utilizing the state-of-the-art GPT4 Turbo model. In this work, we demonstrate that our system outperforms existing methods in critical metrics regarding accuracy. Our system's versatility allows us to generate timelines independent of the input language, highlighting its scalability and adaptation to various applications. This is the first work that introduces the use of LLMs in the Timeline Generation task, as well as building a Timeline Generation system independent of language-specific features, highlighting the novelty of our approach. The experimental results demonstrate that our approach fills critical gaps in Timeline Generation. Thus, this work represents a substantial advancement in Timeline Generation, offering new grounds for research and application in various contexts
Dynamics Of Pattern Formation in Nonequilibrium Diffusion-Precipitation Systems
No full textSelf-organization is a fascinating natural occurrence, yielding intricate structures and patterns under non-equilibrium conditions. One well-known example is the Liesegang phenomenon, which refers to the formation of parallel bands in 1D or concentric rings in 2D, via a diffusion-precipitation mechanism. Driven by a concentration gradient, the outer electrolyte diffuses into a gel medium impregnated with an inner electrolyte containing its coprecipitate ion. The precipitation gives rise to interesting patterns of rich and diverse morphological characteristics.
The present work studies several forms of such precipitation patterns and is composed of four main sections: (1) Investigation of the pattern and boundary formation in the chemical Voronoi framework, (2) Study of the influence of a temperature gradient on several characteristics of a 1D cobalt hydroxide system, (3) Dynamical study on novel three precipitate systems, (4) Studies on the chaotic behavior in 2D lead chromate system.
In the first project, we investigate the chemical analogues of the mathematical Voronoi diagrams. While the original (traditional) systems include sources of equal concentration and equal hole diameter, we vary, separately, the concentration or the hole diameter of these sources, in four different sets of systems. We try several different parameters to characterize the formation of such sectioned or tiled diagrams. Later, we perform kinetics experiments, where we study the evolution of the intensity and velocity of the fronts with time. Additionally, we examine the pattern formation in the presence of linear interfaces.
In the second project, we report the first ever application of a temperature gradient on a Liesegang system. We subject the monotonic cobalt hydroxide system to negative (upward), from 40˚C to 15˚C, and positive (downward), from 15˚C to 40˚C, temperature gradients. The latter are also compared to tubes placed at constant temperatures of 15˚C, 18˚C and 27.5˚C, considering variations in the position of the last band, the number of bands formed, as well as the position of all the bands with band number.
In the third project, we present two new three-precipitate systems. The mere introduction of cadmium and manganese cations, individually, into systems initially containing cobalt and nickel cations yields beautiful and fascinating precipitation patterns. Later, we perform atomic absorption spectrophotometry measurements for some selected systems to determine the composition of the bands.
Finally, in the fourth project, we carry out a preliminary investigation on the chaotic behavior exhibited by the lead chromate system in 2D as a result of variations in the flow rates of the outer electrolyte
Unmasking the Silent Struggle: Leveraging Machine Learning to Uncover Depressive Patterns in Social Media
No full textIn recent years, people have been using social media platforms to express their feelings and share their mental health struggles openly and anonymously. This surge has motivated many researchers to take advantage of social media as a valuable resource of data to detect severe depression. However, existing approaches have significant limitations as they rely on datasets suffering from a wide disparity between negative instances and positive instances, ignoring the middle ground containing depression-mimicking states. This thesis considers new mimicking states (stress, anxiety, sadness, sarcasm, and complaints) that are often misidentified as real depression due to their overlapping nature in the language and expressions used. We propose an automated system for detecting severe depression that features two main modules: content level and user level depression detection. The study encompasses the extensive evaluation of seven LLMs on the content level. We observe that finetuning significantly enhances the performance of all the evaluated LLMs. RoBERTa emerged as the top performer among the tested models, achieving an impressive AUC of 98.53% on the test set. For the user-level module, we utilize the best model, RoBERTa, along two main criteria: the user’s content severity and posting history. This approach is crucial as it allows us to capture the nuanced variations in user behavior and content, thereby enhancing the accuracy and reliability of our classification system. Overall, our work illustrates the potential of using LLMs in developing an accurate depression detection system that will contribute to a reduction in the overall prevalence of untreated depression cases
Use of a Big Data Approach in Guiding Climate Adaptation and Mitigation Policies and Interventions
No full textDespite the rise in global environmental action in the past decades, CO₂ emissions are expected to progressively continue uphill with deforestation swiping millions of hectares annually, hence destroying a major CO₂ sink. Moreover, many of the past controversies on climate change arose due to contentious, differing, disputed data or its nonexistence – but big data could come to solve this challenge.
Today, big data can be used for various climate interventions. The United Nations has clearly indicated that there is a role for big data in fighting climate change and other key global issues underlying the Sustainable Development Goals. But while the role of big data in many climate action areas has become evident, a clear gap exists in understanding its ability to shape climate policy. Hence, given the rapidly expanding positive impact and influence of big data, it is thus serious to investigate its implications in the climate policy realm. Hence, this work examines the role of big climate data in guiding national climate policymaking, specifically the Paris Agreement’s Nationally Determined Contributions (NDCs).
Methodologically, the study followed a mixed methods approach, by using both quantitative and qualitative data, and the analysis of both primary and secondary data sources. Primary data analysis consisted of semi-structured interviews with climate policy experts who were selected for their expertise using purposive sampling. The interviews allowed us to understand perspectives on the ability of big climate databases to reflect NDCs and provide support in their design and implementation. Secondary data analysis involved the evaluation of selected global climate NDC databases by developing an evaluation tool from relevant sources from the literature, including peer-reviewed scientific articles, climate policy documents and international reports. For the purpose of feasibility, case studies were selected also using purposive sampling. The study focused its scope on three big climate databases (World Emissions Clock, Climate Action Tracker, Climate Watch) and two countries (Lebanon and the United Kingdom).
Overall, the results showcased a currently low level of big data guidance in all levels of climate policy, but with a potential for increase if key challenges are addressed properly. Hence, the study proposed a recommendatory framework to ensure enhanced big data guidance in national climate adaptation and mitigation policies based on expert views and insights from the literature
Context Relevance and Community Policing in Lebanon
No full textThis project tackles the implementation of the community policing model. It focuses on the importance of relevance to the social context in implementing this model. The project takes the Lebanese state security institutions as a case study. This research claims that community policing implementation should be relevant to the context in which it is deployed. Further, it argues that the matter of relevance is not only restricted to society but also includes the relevance within state security institutions that are part of the social context. Although the social context encompasses several aspects, the project focuses mainly on the institutional aspect in discussing the community policing relevance, challenges, and opportunities in the Lebanese context. Furthermore, the project is divided into three chapters. The first chapter covers the theoretical framework surrounding the topic. It clarifies the basic policing definitions and highlights the impact and importance of social context relevance in the community policing implementation process. In the second chapter, the angel is narrowed down to focus on the case of Lebanon. This chapter explains how the implementation process has gradually occurred through exploring the Lebanese experience with community policing since 2008 and through examining the policing legal framework and infrastructure for this model. It further discusses the general challenges facing community policing implementation. The third chapter focuses on the institutional aspect of community policing implementation, and the importance of ensuring relevance within state security institutions. This chapter tackles the importance of introducing quality assurance frameworks within security institutions to ensure that the institution’s policies, practices, and organization are relevant to community policing. Also, the chapter presents a set of recommendations to the Lebanese Internal Security Forces (ISF) institution that are needed for an efficient and relevant community policing implementation. The project concludes by questioning the “rendering technical” approach that is shifting security issues from the political arena, and also, by presenting a new community policing perspective
Heterogeneous Catalysis of Biodiesel Production by Corn Oil Transesterification using a Calcium Oxide - Cd Al LDH Composite Synthesized via RDF
Get PDFIn this study, a CdAl LDH precursor synthesized by RDF was impregnated with
calcium nitrate and calcinated to obtain Ca/CdAl oxide. This was used as a heterogeneous base catalyst in the transesterification of corn oil with methanol for the
purpose of biodiesel production. The catalyst was characterized by SEM/EDX,
powder XRD, and BET nitrogen adsorption-desorption technique. Thermogravimetric analysis was carried out on its LDH precursor. Transesterification reactions
were carried out using the prepared catalyst and commercially obtained calcium
oxide to compare their activity in the reaction. Proton NMR was used to examine
the obtained biodiesel product. The fatty acid methyl ester yield was measured
under a series of different reaction durations, with a 4wt% catalyst loading at 11:1
methanol:oil ratio at 62°C. The optimal biodiesel yield was determined to be 93%.
The Ca/CdAl oxide catalyst was also tested in the transesterification of waste cooking oil, where a 72.4% yield was observed
Al-Sakhawi : a biographer from Mamlūk Cairo and his dictionary of women
Get PDFText in English; with some texts in Arabic.Includes bibliographical references and index
MULTI-MODAL ARABIC NEGOTIATING BOT
No full textNegotiation is a fundamental aspect of human interaction, involving a dynamic
exchange of communication between two or more parties to reach mutually agreeable
outcomes. With recent advancements in chatbots, leveraging artificial intelligence (AI)
for negotiation has emerged as an ideal application. Despite significant progress in
English negotiation bots using deep learning and reinforcement learning, such
advancements are notably absent in other languages, particularly Arabic. Furthermore,
while previous research has primarily focused on developing high-performing neural
response generation systems for negotiation bots, the integration of multimodality into
these automated agents remains unexplored. The incorporation of multimodality is
represented in image analysis, and it contributes to a more comprehensive and userfriendly negotiation model. This thesis presents the first Arabic negotiation model,
distinguished by incorporating multimodality into negotiation models. The integration
of multimodality, particularly through image analysis, provides a more comprehensive
and user-centric approach to negotiation. Our primary objective is to develop an Arabic
multimodal negotiating bot, a seller agent capable of engaging in negotiations with
buyers in the context of item sales. This seller agent is designed to understand the
buyer's Arabic utterances and to interpret the negotiation context through images
provided by the buyer. To achieve this, we trained a Generative Pre-trained Transformer
(GPT-2) model on an Arabic dataset, integrating it with a Convolutional Neural
Network (CNN) for image analysis. The model's automatic evaluation yielded a BLEU4 score of 0.21 and a cross-entropy loss of 0.55, metrics that are promising for the first
model of its kind in Arabic. Our experiments and analyses reveal both the successes and
limitations of the designed multi-modal Arabic negotiating model, offering insights into
the inherent challenges and setting directions for future research
Experimental Validation of a Computational Fluid Dynamics Model of The Upper Respiratory Airways Utilizing a Heat Transfer Approach
No full textTobacco smoking is one of the leading causes of death and disease globally. E-CIGS (e-cigarettes) were introduced in 2004 and marketed as a safer alternative. However, there is growing concern that instead of displacing combustible cigarette use, E-CIGS are attracting large numbers of nicotine-naïve youth who may not have smoked otherwise. Factors such as attractive flavors, advertising targeting youth, and the introduction of nicotine salts are responsible for the rise in prevalence among youth. Unlike free-base nicotine, nicotine salts are nonvolatile and do not induce throat harshness when inhaled. One regulatory approach that has been proposed is to set a floor on the throat harshness of electronic cigarette aerosols to deter previously nicotine-naïve youth from using them. However, the relationship between throat harshness and various electronic cigarette variables, including nicotine concentration, nicotine salt fraction, electrical power, and inhalation patterns, has not been closely examined. Recently, a theoretical model was developed to quantify nicotine deposition and throat harshness from key electronic cigarette variables. This model is partly derived from a computational fluid dynamics simulation of flow through the upper respiratory airways.
This thesis aims to experimentally validate the CFD simulations used in the 1-D model derivation of the segmental heat transfer correlations. The study involved the construction of a physical model that replicates the complex geometry of the human airway used in CFD computations and measuring the temperature at various points in the model. At the same time, air was drawn through it at different flow rates (1 SLPM to 20 SLPM ). Then, temperature measurements, appropriately non-dimensionalized, were compared to the CFD-predicted temperatures at the same flow rates. An experimental setup was developed to mimic the idealized boundary conditions used in the CFD model. The setup rigorously treated measurement uncertainty and optimized temperature measurement locations in the flow path. Key measurement outcomes included temperature, relative temperature change across locations, and relative temperature change across flow rates.
Experimental results aligned with the CFD predictions after optimizing for the dominant source of uncertainty, the position of the thermocouples. This experimental set-up was deemed a working approach to validating this computational fluid dynamic simulation across the upper respiratory tract. However, to improve this setup, a more precise method is needed to locate the thermocouples inside the physical model after its construction
First Detection of the Plasmid-Mediated Colistin Resistance Gene, mcr-1, in Multidrug Resistant E. coli, in Backyard Animals in Lebanon
No full textThe increase in antibiotic resistance has jeopardized the use of what is dubbed as last resort antibiotic, Polymyxin E. Specifically, the dissemination of mobile colistin resistance (mcr-1) gene has imperiled the effectiveness of this drug; posing a serious challenge worldwide.
Previous studies documented the occurrence of the mobile colistin resistance gene, mcr-1, that encodes colistin (Polymyxin E) resistance in Escherichia coli isolated from vital matrices such as irrigation, sea water, aquaculture, drinking water and broiler chickens in Lebanon. To investigate the spread of mcr-1 gene to new matrices, we targeted backyard farms that have become a main food source for a large number of people in the country during the on-going economic crisis. In this study, 15 fecal samples were collected from a backyard farm in South Lebanon; 3 samples were collected from different animal species on the farm. All the samples were analyzed on agar media that was supplemented with 4 μg/ml colistin, and media without colistin to detect colistin and non- colistin resistant E. coli. Forty-five colonies were selected and purified from non-colistin plates, and twenty-seven colonies were selected and purified from colistin-containing plates. And all these colonies were selected for further analyses that included antimicrobial resistance phenotypes and gene-specific PCR analyses among others.
All the colonies that grew on colistin plates were mcr-1 positive and were multidrug-resistant and the minimum inhibitory concertation of colistin varied between 4 and 1024 μg/ml. E. coli isolates (75%, 71%, and 44 %) carried blaTEM , isolated from chickens, pigeons and sheep respectively, and only 25% isolates of the chickens carried blaCTX-M.
Plasmid transformation experiments and sequencing for mcr-1 gene were performed the transformants were colistin resistant and mcr-1 positive, indicating that the gene was plasmid-borne. Detecting mcr-1 positive E. coli and other resistant genes is a serious issue and can threaten the therapy of infectious diseases in the human and animal population.
Our findings emphasize the absolute need to tackle the surge of colistin resistance in Lebanon through a One Health approach