1,720,985 research outputs found
Going Beyond Counting First Authors in Author Co-citation Analysis
Get PDFThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
Intelligence artificielle pour la configuration et l’achat de campagnes publicitaires en ligne
No full textAvec l'évolution du marketing moderne, la publicité numérique est devenue un élément central, permettant aux entreprises d'atteindre des audiences plus larges et diversifiées avec une précision inédite. Les plateformes de publicité numérique offrent des avantages comme des coûts contrôlables, un ciblage précis et la mesurabilité des performances. Cependant, la complexité croissante de ces écosystèmes pose des défis majeurs pour l'optimisation des campagnes. Les méthodes traditionnelles, dépendantes de l'expertise humaine, deviennent insuffisantes face à cette complexité. D'où une dépendance accrue aux solutions algorithmiques et à l'intelligence artificielle pour améliorer les résultats.Les travaux actuels se concentrent principalement sur l'optimisation des campagnes après leur lancement, telles que la prévision des indicateurs clés de performance (KPI) et des coûts pour optimiser les agents de bidding en temps réel (RTB) qui visent à affiner l'allocation budgetaire et maximiser l'efficacité des campagnes en cours. Malgré les progrès réalisés, l'efficacité de ces méthodes dépend de la bonne configuration préalable des campagnes publicitaires, spécifiquement en termes de ciblage d'audiences et un bon paramètrage. Une tâche que nous appelons Conception de Stratégie Publicitaire.Malheureusement, en raison de l'échelle et de la complexité inhérentes à la tâche, il y a un manque considérable dans l'optimisation des campagnes avant leur lancement. La conception de stratégie publicitaire repose encore fortement sur de l'expertise humaine, ce qui conduit souvent à un ciblage sous-optimal, et dégrade la prise de décision ainsi que la performance globale de la campagne, soulignant un domaine potentiel d'amélioration.Cette thèse vise à exploiter les méthodes d'intelligence artificielle pour la configuration et l'optimisation des campagnes publicitaires numériques. À cette fin, nous intégrons des approches d'apprentissage profond dans les phases initiales de planification de la campagne dans la tâche de conception de stratégie publicitaire. Dans cette thèse, nous contribuons d'abord un système novateur et un modèle de réseau de neurones génératif qui exploite le mécanisme d'attention des transformers pour contextuellement générer des stratégies publicitaires optimales tout en évitant l'explosion combinatoire. Nous évaluons nos résultats sur un ensemble de données public iPinYou ainsi que sur les données de l'entreprise en mesurant la proximité des stratégies générées avec les ensembles de données (Distance Cosinus et Hamming) ainsi que leur performance KPI estimée. En l'absence de méthodes directement comparables, nous avons comparé nos résultats à des méthodes principales d'autres domaines, adaptées à cette tâche spécifique. Nous affinons ensuite notre contribution, en améliorant la diversité générative, en améliorant la robustesse contre le mode-collapse — une condition où le modèle tend à générer une gamme limitée de sorties — et en introduisant un mode exploratoir au moment de l'inférence via des techniques de quantification vectorielle et l'apprentissage de métriques. Nous proposons également un protocole d'évaluation amélioré pour notre système. Nous proposons finalement une méthodologie novatrice axée sur des tokens de signalisation pour le contrôle génératif flexible dans les modèles basés sur les transformers qui prends en entrée des signaux suggestifs, ce qui permet à notre modèle de considérer les préférences utilisateur tout en conservant l'autonomie de s'en écarter si elles ne produisent pas de résultats optimaux, les intégrant dans le processus génératif comme des paramètres suggestifs plutôt que des directives strictes. Des expériences étendues ont été menées pour évaluer l'efficacité de notre approche, qui a produit des résultats exceptionnels et confirmé son applicabilité dans divers domaines utilisant des modèles de transformateurs.In the evolving landscape of modern marketing, digital advertising has emerged as a pivotal component, enabling businesses to expand their reach to larger, more diverse audiences with unprecedented precision. Digital advertising platforms offer advantages such as controllable costs, accurate audience targeting, and measurable feedback. However, the escalating complexity of digital advertising ecosystems poses significant challenges in optimizing the performance of advertising campaigns. Traditional methodologies, heavily reliant on human expertise, are increasingly inadequate in addressing the multifaceted nature of these digital environments. Consequently, there is a growing dependence on algorithmic solutions and artificial intelligence (AI) to navigate this complexity and enhance campaign outcomes.Current works predominantly focus on post-launch campaign optimization, such as key performance indicator (KPI) and cost forecasting to optimize real-time bidding (RTB) agents which aim to refine budget allocation and maximize the effectiveness of ongoing campaigns. Despite the achieved progress, the effectiveness of these methods depends on the accurate configuration of advertising campaigns, specifically in terms of targeting the appropriate audiences with the correct parameters. A task that we call Advertising Strategy Design.Unfortunately, due to the inherent scale and complexity of the task, there is a noticeable gap in pre-launch campaign optimization process. Advertising strategy design still relies heavily on human expertise, which often leads to sub-optimal targeting and decision-making. This affects the overall campaign performance, underscoring a potential area of improvement.This thesis aims at leveraging artificial intelligence methods for the configuration and optimization of digital advertising campaigns. For this purpose, we integrate deep learning approaches in the initial phases of campaign planning in the task of advertising strategy design. In this thesis, we first contribute a novel framework and generative neural network model which leverages the attention mechanism through transformers to contextually generate optimal advertising strategies while avoiding combinatorial explosion. We evaluate our results on a public dataset iPinYou as well as the company's private dataset by measuring the closeness of the generated strategies to the datasets (using Cosine and Hamming distances) as well as their estimated KPI performance. In the absence of directly comparable methods, we benchmarked our results against prominent methods from other fields, adapted for this specific task. We further refined our approach by enhancing the generative diversity, improving robustness against mode collapse—a condition where the model tends toward generating a limited range of outputs—and introducing an inference-time exploration mode employing vector quantization techniques and learned metrics. An improved evaluation protocol for our framework was also developed. We finally propose a novel token-driven methodology for flexible generative control in transformer-based models. This method includes a suggestive input mechanism that allows the model to take user preferences into account while maintaining the freedom to deviate from them if they do not lead to optimal outcomes, treating these inputs as guiding suggestions rather than strict rules. Extensive experiments were conducted to assess the effectiveness of our approach, which yielded outstanding results and confirmed its applicability across various domains utilizing transformer models
Variations on the Author
Get PDF“Variations on the Author” discusses two of Eduardo Coutinho’s recent films (Um Dia na Vida, from 2010, and Últimas Conversas, posthumously released in 2015) and their contribution to the general question of documentary authorship. The director’s filmography is characterized by a consistent yet self-effacing form of authorial self-inscription: Coutinho often features as an interviewer that rather than express opinions propels discourses; an interviewer that is good at listening. This mode of self-inscription characterizes him as an author who is not expressive but who is nonetheless markedly present on the screen. In Um Dia na Vida, however, Coutinho is completely absent form the image, while Últimas Conversas, on the contrary, includes a confessional prologue that moves the director from the margins to the center of his films. This article examines the ways in which these works stand out in the filmography of a director who offers new insights into the notion of cinematic authorship
Appropriate Similarity Measures for Author Cocitation Analysis
Get PDFWe provide a number of new insights into the methodological discussion about author cocitation analysis. We first argue that the use of the Pearson correlation for measuring the similarity between authors’ cocitation profiles is not very satisfactory. We then discuss what kind of similarity measures may be used as an alternative to the Pearson correlation. We consider three similarity measures in particular. One is the well-known cosine. The other two similarity measures have not been used before in the bibliometric literature. Finally, we show by means of an example that our findings have a high practical relevance.information science;Pearson correlation;cosine;similarity measure;author cocitation analysis
Methods for Job Recommandation on Social Networks
No full textNous sommes à l’aube d’une nouvelle ère du data mining, celle du stockage, traitement, analyse et exploitation des données massives que l’on appelle Big Data. Les données sont devenues une nouvelle matière première, très prisée par les entreprises de tout type et de toute taille à travers le monde ; elles permettent d’analyser, de comprendre, de modéliser et d’expliquer certains phénomènes comme le comportement et les préférences des utilisateurs ou clients d’une entreprise donnée. La compréhension des préférences des utilisateurs et des clients d’une entreprise permet de leur proposer de la publicité ciblée afin d’augmenter les ventes et la satisfaction des clients et ainsi pouvoir améliorer les revenues de l’entreprise, ce que les géants du Web comme Google, Facebook, LinkedIn et Twitter ont bien compris. Cette thèse de doctorat a été réalisée dans le cadre d’une convention CIFRE entre le laboratoire L2TI de l’université Paris 13 et la start-up franco-américaineWork4 qui développe des applications de recrutement sur Facebook. Son objectif principal était la mise au point d’un ensemble d’algorithmes et méthodes pour proposer aux utilisateurs des réseaux sociaux les offres d’emploi les plus pertinentes. Le développement de nos algorithmes de recommandation a nécessité de surmonter de nombreuses difficultés telles que le préservation de la vie privée des utilisateurs des réseaux sociaux, le traitement des données bruitées et incomplètes des utilisateurs et des offres d’emploi, la difficulté de traitement des données multi-langues et, plus généralement, la difficulté d’extraire automatiquement les offres d’emploi pertinentes pour un utilisateur donné parmi un ensemble d’offres d’emploi. Les systèmes développés durant cette thèse sont principalement basés sur les techniques de systèmes de recommandation, de recherche documentaire,de fouille de données et d’apprentissage artificiel ; ils ont été validés sur des jeux de données réels collectés par l’entreprise Work4. Dans le cadre de cette étude, les utilisateurs d’un réseau social sont liés à trois types entités : les offres d’emploi qui leur sont pertinentes, les autres utilisateurs du réseau social auxquels ils se sont liés d’amitié et les données personnelles qu’ils ont publiées sur leurs profils. Les profils des utilisateurs des réseaux sociaux et la description de nos offres d’emploi sont constitués de plusieurs champs contenant des informations textuelles.We are entering a new era of data mining in which the main challenge is the storing andprocessing of massive data : this is leading to a new promising research and industry field called Big data. Data are currently a new raw material coveted by businesses of all sizes and all sectors. They allow organizations to analyze, understand, model and explain phenomen a such as the behavior of their users or customers. Some companies like Google, Facebook,LinkedIn and Twitter are using user data to determine their preferences in order to make targeted advertisements to increase their revenues.This thesis has been carried out in collaboration between the laboratory L2TI andWork4, a French-American startup that offers Facebook recruitment solutions. Its main objective was the development of systems recommending relevant jobs to social network users ; the developed systems have been used to advertise job positions on social networks. After studying the literature about recommender systems, information retrieval, data mining and machine learning, we modeled social users using data they posted on their profiles, those of their social relationships together with the bag-of-words and ontology-based models. We measure the interests of users for jobs using both heuristics and models based on machine learning. The development of efficient job recommender systems involved to tackle the problem of categorization and summarization of user profiles and job descriptions. After developing job recommender systems on social networks, we developed a set of systems called Work4 Oracle that predict the audience (number of clicks) of job advertisements posted on Facebook, LinkedIn or Twitter. The analysis of the results of Work4 Oracle allows us to find and quantify factors impacting the popularity of job ads posted on social networks, these results have been compared to those of the literature of Human Resource Management. All our proposed systems deal with privacy preservation by only using the data that social network users explicitly allowed to access to ; they also deal with noisy and missing data of social network users and have been validated on real-world data provided by Work4
Dispelling the Myths Behind First-author Citation Counts
Get PDFWe conducted a full-scale evaluative citation analysis study of scholars in the XML research field to explore just how different from each other author rankings resulting from different citation counting methods actually are, and to demonstrate the capability of emerging data and tools on the Web in supporting more realistic citation counting methods. Our results contest some common arguments for the continued
use of first-author citation counts in the evaluation of scholars, such as high correlations between author rankings by first-author citation counts and other citation
counting methods, and high costs of using more realistic citation counting methods that are not well-supported by the ISI databases. It is argued that increasingly available digital full text research papers make it possible for citation analysis studies to go beyond what the ISI databases have directly supported and to employ more
sophisticated methods
koamabayili/VECTRON-author-checklist: VECTRON author checklist
No full textWe have done our best to complete the author checklist relating to the use of animals in the hut study. Note that the objective for the hut study was to evaluate the IRS treatment applications for residual efficacy against Anopheles mosquitoes, including the local An. coluzzii mosquito population. Cows were only used to attract mosquitoes into the huts and no tests were carried out directly on the cows. The author checklist is intended for use with studies where experiments are carried out on animals, which is why we have had such difficulty in completing this for the hut study, as many of the questions do not relate to how the cows were used
Author-wise bibliometric analysis based on entropy.
No full textAuthor-wise bibliometric analysis based on entropy.</p
- …
