Showing 1–17 of 17 results for author: Dumitrescu, D

Let's CONFER: A Dataset for Evaluating Natural Language Inference Models on CONditional InFERence and Presupposition

Authors: Tara Azin, Daniel Dumitrescu, Diana Inkpen, Raj Singh

Abstract: Natural Language Inference (NLI) is the task of determining whether a sentence pair represents entailment, contradiction, or a neutral relationship. While NLI models perform well on many inference tasks, their ability to handle fine-grained pragmatic inferences, particularly presupposition in conditionals, remains underexplored. In this study, we introduce CONFER, a novel dataset designed to evalu… ▽ More Natural Language Inference (NLI) is the task of determining whether a sentence pair represents entailment, contradiction, or a neutral relationship. While NLI models perform well on many inference tasks, their ability to handle fine-grained pragmatic inferences, particularly presupposition in conditionals, remains underexplored. In this study, we introduce CONFER, a novel dataset designed to evaluate how NLI models process inference in conditional sentences. We assess the performance of four NLI models, including two pre-trained models, to examine their generalization to conditional reasoning. Additionally, we evaluate Large Language Models (LLMs), including GPT-4o, LLaMA, Gemma, and DeepSeek-R1, in zero-shot and few-shot prompting settings to analyze their ability to infer presuppositions with and without prior context. Our findings indicate that NLI models struggle with presuppositional reasoning in conditionals, and fine-tuning on existing NLI datasets does not necessarily improve their performance. △ Less

Submitted 6 June, 2025; originally announced June 2025.

Comments: This paper is published in the Proceedings of the 38th Canadian Conference on Artificial Intelligence (CAIAC 2025). Please cite the conference version at https://caiac.pubpub.org/pub/keh8ij01

Generative Adversarial Training for Text-to-Speech Synthesis Based on Raw Phonetic Input and Explicit Prosody Modelling

Authors: Tiberiu Boros, Stefan Daniel Dumitrescu, Ionut Mironica, Radu Chivereanu

Abstract: We describe an end-to-end speech synthesis system that uses generative adversarial training. We train our Vocoder for raw phoneme-to-audio conversion, using explicit phonetic, pitch and duration modeling. We experiment with several pre-trained models for contextualized and decontextualized word embeddings and we introduce a new method for highly expressive character voice matching, based on discre… ▽ More We describe an end-to-end speech synthesis system that uses generative adversarial training. We train our Vocoder for raw phoneme-to-audio conversion, using explicit phonetic, pitch and duration modeling. We experiment with several pre-trained models for contextualized and decontextualized word embeddings and we introduce a new method for highly expressive character voice matching, based on discreet style tokens. △ Less

Submitted 14 October, 2023; originally announced October 2023.

The birth of Romanian BERT

Authors: Stefan Daniel Dumitrescu, Andrei-Marius Avram, Sampo Pyysalo

Abstract: Large-scale pretrained language models have become ubiquitous in Natural Language Processing. However, most of these models are available either in high-resource languages, in particular English, or as multilingual models that compromise performance on individual languages for coverage. This paper introduces Romanian BERT, the first purely Romanian transformer-based language model, pretrained on a… ▽ More Large-scale pretrained language models have become ubiquitous in Natural Language Processing. However, most of these models are available either in high-resource languages, in particular English, or as multilingual models that compromise performance on individual languages for coverage. This paper introduces Romanian BERT, the first purely Romanian transformer-based language model, pretrained on a large text corpus. We discuss corpus composition and cleaning, the model training process, as well as an extensive evaluation of the model on various Romanian datasets. We open source not only the model itself, but also a repository that contains information on how to obtain the corpus, fine-tune and use this model in production (with practical examples), and how to fully replicate the evaluation process. △ Less

Submitted 18 September, 2020; originally announced September 2020.

Comments: 5 pages (4 + reference page), accepted in Findings of EMNLP 2020

Introducing RONEC -- the Romanian Named Entity Corpus

Authors: Stefan Daniel Dumitrescu, Andrei-Marius Avram

Abstract: We present RONEC - the Named Entity Corpus for the Romanian language. The corpus contains over 26000 entities in ~5000 annotated sentences, belonging to 16 distinct classes. The sentences have been extracted from a copy-right free newspaper, covering several styles. This corpus represents the first initiative in the Romanian language space specifically targeted for named entity recognition. It is… ▽ More We present RONEC - the Named Entity Corpus for the Romanian language. The corpus contains over 26000 entities in ~5000 annotated sentences, belonging to 16 distinct classes. The sentences have been extracted from a copy-right free newspaper, covering several styles. This corpus represents the first initiative in the Romanian language space specifically targeted for named entity recognition. It is available in BRAT and CoNLL-U Plus formats, and it is free to use and extend at github.com/dumitrescustefan/ronec . △ Less

Submitted 27 April, 2020; v1 submitted 3 September, 2019; originally announced September 2019.

Comments: 8 pages + annex, accepted to LREC2020 in the main conference

Tools and resources for Romanian text-to-speech and speech-to-text applications

Authors: Tiberiu Boros, Stefan Daniel Dumitrescu, Vasile Pais

Abstract: In this paper we introduce a set of resources and tools aimed at providing support for natural language processing, text-to-speech synthesis and speech recognition for Romanian. While the tools are general purpose and can be used for any language (we successfully trained our system for more than 50 languages and participated in the Universal Dependencies Shared Task), the resources are only releva… ▽ More In this paper we introduce a set of resources and tools aimed at providing support for natural language processing, text-to-speech synthesis and speech recognition for Romanian. While the tools are general purpose and can be used for any language (we successfully trained our system for more than 50 languages and participated in the Universal Dependencies Shared Task), the resources are only relevant for Romanian language processing. △ Less

Submitted 15 February, 2018; originally announced February 2018.

Evolving TSP heuristics using Multi Expression Programming

Authors: Mihai Oltean, D. Dumitrescu

Abstract: Multi Expression Programming (MEP) is an evolutionary technique that may be used for solving computationally difficult problems. MEP uses a linear solution representation. Each MEP individual is a string encoding complex expressions (computer programs). A MEP individual may encode multiple solutions of the current problem. In this paper MEP is used for evolving a Traveling Salesman Problem (TSP) h… ▽ More Multi Expression Programming (MEP) is an evolutionary technique that may be used for solving computationally difficult problems. MEP uses a linear solution representation. Each MEP individual is a string encoding complex expressions (computer programs). A MEP individual may encode multiple solutions of the current problem. In this paper MEP is used for evolving a Traveling Salesman Problem (TSP) heuristic for graphs satisfying triangle inequality. Evolved MEP heuristic is compared with Nearest Neighbor Heuristic (NN) and Minimum Spanning Tree Heuristic (MST) on some difficult problems in TSPLIB. For most of the considered problems the evolved MEP heuristic outperforms NN and MST. The obtained algorithm was tested against some problems in TSPLIB. The results emphasizes that evolved MEP heuristic is a powerful tool for solving difficult TSP instances. △ Less

Submitted 8 September, 2015; originally announced September 2015.

Comments: International Conference on Computational Sciences, ICCS'04, 6-9 June, Krakow, Poland, Edited by M. Bubak, G.van Albada, P. Sloot, and J. Dongarra, Vol II, pp. 670-673, Springer-Verlag, Berlin, 2004. Source code available for download at: http://www.cs.ubbcluj.ro/~moltean/evolve_heuristics.html

Characterization and Detection of epsilon-Berge Zhukovskii Equilibria

Authors: Noémi Gaskó, Mihai Suciu, Rodica Ioana Lung, D. Dumitrescu

Abstract: Berge equilibrium in the sense of Zhukovskii (Berge-Zhukovskii) is an alternate solution concept in non-cooperative game theory that formalizes cooperation in a noncooperative setting. In this paper the epsilon-Berge-Zhukovskii equilibrium is introduced and characterized by using a generative relation. A computational method for detecting epsilon-Berge-Zhukovskii equilibrium based on evolutionary… ▽ More Berge equilibrium in the sense of Zhukovskii (Berge-Zhukovskii) is an alternate solution concept in non-cooperative game theory that formalizes cooperation in a noncooperative setting. In this paper the epsilon-Berge-Zhukovskii equilibrium is introduced and characterized by using a generative relation. A computational method for detecting epsilon-Berge-Zhukovskii equilibrium based on evolutionary multiobjective optimization algorithms is presented. Numerical examples are used to illustrate the results obtained. △ Less

Submitted 2 May, 2014; originally announced May 2014.

Computing Strong Nash Equilibria for Multiplayer Games

Authors: Noémi Gaskó, Rodica Ioana Lung, D. Dumitrescu

Abstract: An heuristic approach to compute strong Nash (Aumann) equilibria is presented. The method is based on differential evolution and three variants of a generative relation for strong Nash equilibria characterization. Numerical experiments performed on the minimum effort game for up to 150 players illustrate the efficiency of the approach. The advantages and disadvantages of each variant is discussed… ▽ More An heuristic approach to compute strong Nash (Aumann) equilibria is presented. The method is based on differential evolution and three variants of a generative relation for strong Nash equilibria characterization. Numerical experiments performed on the minimum effort game for up to 150 players illustrate the efficiency of the approach. The advantages and disadvantages of each variant is discussed in terms of precision and running time. △ Less

Submitted 28 August, 2014; v1 submitted 1 May, 2014; originally announced May 2014.

Beyond Nash Equilibrium in Open Spectrum Sharing: Lorenz Equilibrium in Discrete Games

Authors: Ligia Cremene, D. Dumitrescu

Abstract: A new game theoretical solution concept for open spectrum sharing in cognitive radio (CR) environments is presented, the Lorenz equilibrium (LE). Both Nash and Pareto solution concepts have limitations when applied to real world problems. Nash equilibrium (NE) rarely ensures maximal payoff and it is frequently Pareto inefficient. The Pareto set is usually a large set of solutions, often too hard t… ▽ More A new game theoretical solution concept for open spectrum sharing in cognitive radio (CR) environments is presented, the Lorenz equilibrium (LE). Both Nash and Pareto solution concepts have limitations when applied to real world problems. Nash equilibrium (NE) rarely ensures maximal payoff and it is frequently Pareto inefficient. The Pareto set is usually a large set of solutions, often too hard to process. The Lorenz equilibrium is a subset of Pareto efficient solutions that are equitable for all players and ensures a higher payoff than the Nash equilibrium. LE induces a selection criterion of NE, when several are present in a game (e.g. many-player discrete games) and when fairness is an issue. Besides being an effective NE selection criterion, the LE is an interesting game theoretical situation per se, useful for CR interaction analysis. △ Less

Submitted 5 April, 2013; originally announced April 2013.

Comments: 5 pages, 4 figures

Cognitive Radio Simultaneous Spectrum Access/ One-shot Game Modelling

Authors: Ligia C. Cremene, D. Dumitrescu, Reka Nagy, Noemi Gasko

Abstract: The aim of this work is to asses simultaneous spectrum access situations that may occur in Cognitive Radio (CR) environments. The approach is that of one shot, noncooperative games describing CR interactions. Open spectrum access scenarios are modelled based on continuous and discrete reformulations of the Cournot game theoretical model. CR interaction situations are described by Nash and Pareto e… ▽ More The aim of this work is to asses simultaneous spectrum access situations that may occur in Cognitive Radio (CR) environments. The approach is that of one shot, noncooperative games describing CR interactions. Open spectrum access scenarios are modelled based on continuous and discrete reformulations of the Cournot game theoretical model. CR interaction situations are described by Nash and Pareto equilibria. Also, the heterogeneity of players is captured by the new concept of joint Nash-Pareto equilibrium, allowing CRs to be biased toward different types of equilibrium. Numerical simulations reveal equilibrium situations that may be reached in simultaneous access scenarios of two and three users. △ Less

Submitted 22 September, 2012; originally announced September 2012.

Comments: 6 double-column pages, 8 figures, CSNDSP 2012. arXiv admin note: substantial text overlap with arXiv:1207.3365, arXiv:1209.5387, arXiv:1209.5013

Game Theoretical Modelling for Dynamic Spectrum Access in TV Whitespace

Authors: Ligia C. Cremene, D. Dumitrescu, Reka Nagy, Marcel Cremene

Abstract: In order to assess TV whitespace access scenarios, three oligopoly game models are considered and reformulated in terms of radio access: Cournot, Stackelberg, and Bertrand. Besides revisiting the relevance of Nash and Pareto equilibria, a new equilibrium concept, the joint Nash-Pareto equilibrium, is considered. An evolutionary game equilibria detection method is used. The analysis of the simulati… ▽ More In order to assess TV whitespace access scenarios, three oligopoly game models are considered and reformulated in terms of radio access: Cournot, Stackelberg, and Bertrand. Besides revisiting the relevance of Nash and Pareto equilibria, a new equilibrium concept, the joint Nash-Pareto equilibrium, is considered. An evolutionary game equilibria detection method is used. The analysis of the simulation results brings relevant insights on the issue of autonomy vs. regulation in emerging cognitive radio environments. △ Less

Submitted 22 September, 2012; originally announced September 2012.

Comments: 5 double-column pages, 6 figures, CrownCom 2011. arXiv admin note: substantial text overlap with arXiv:1207.3365, arXiv:1209.5013

Analysis of Cognitive Radio Scenes Based on Non-cooperative Game Theoretical Modelling

Authors: Ligia Cremene, Dumitru Dumitrescu

Abstract: A noncooperative game theoretical approach for analysing opportunistic spectrum access (OSA) in cognitive radio (CR) environments is proposed. New concepts from game theory are applied to spectrum access analysis in order to extract rules of behaviour for an emerging environment. In order to assess OSA scenarios of CRs, two oligopoly game models are reformulated in terms of resource access: Courno… ▽ More A noncooperative game theoretical approach for analysing opportunistic spectrum access (OSA) in cognitive radio (CR) environments is proposed. New concepts from game theory are applied to spectrum access analysis in order to extract rules of behaviour for an emerging environment. In order to assess OSA scenarios of CRs, two oligopoly game models are reformulated in terms of resource access: Cournot and Stackelberg games. Five CR scenes are analysed: simultaneous access of unlicensed users (commons regime) with symmetric and asymmetric costs, with and without bandwidth constraints and sequential access (licensed against unlicensed). Several equilibrium concepts are studied as game solutions: Nash, Pareto and the joint NashPareto equilibrium. The latter captures a game situation where players are non-homogeneous users, exhibiting different types of rationality, Nash and Pareto. This enables a more realistic modelling of interactions on a CR scene. An evolutionary game equilibrium detection method is used. The Nash equilibrium indicates the maximum number of channels a CR may access without decreasing its payoff. The Pareto equilibrium describes a larger range of payoffs, capturing unbalanced as well as equitable solutions. The analysis of the Stackelberg modelling shows that payoffs are maximised for all users if the incumbents are Nash oriented and the new entrants are Pareto driven. △ Less

Submitted 18 October, 2012; v1 submitted 22 September, 2012; originally announced September 2012.

Comments: 8 double-column pages, 10 figures. arXiv admin note: text overlap with arXiv:1209.5387

Distributed Pharaoh System for Network Routing

Authors: Camelia-M. Pintea, D. Dumitrescu

Abstract: In this paper it is introduced a biobjective ant algorithm for constructing low cost routing networks. The new algorithm is called the Distributed Pharaoh System (DPS). DPS is based on AntNet algorithm. The algorithm is using Pharaoh Ant System (PAS) with an extra-exploration phase and a 'no-entry' condition in order to improve the solutions for the Low Cost Network Routing problem. Additionally i… ▽ More In this paper it is introduced a biobjective ant algorithm for constructing low cost routing networks. The new algorithm is called the Distributed Pharaoh System (DPS). DPS is based on AntNet algorithm. The algorithm is using Pharaoh Ant System (PAS) with an extra-exploration phase and a 'no-entry' condition in order to improve the solutions for the Low Cost Network Routing problem. Additionally it is used a cost model for overlay network construction that includes network traffic demands. The Pharaoh ants (Monomorium pharaonis) includes negative pheromones with signals concentrated at decision points where trails fork. The negative pheromones may complement positive pheromone or could help ants to escape from an unnecessarily long route to food that is being reinforced by attractive signals. Numerical experiments were made for a random 10-node network. The average node degree of the network tested was 4.0. The results are encouraging. The algorithm converges to the shortest path while converging on a low cost overlay routing network topology. △ Less

Submitted 27 August, 2012; originally announced August 2012.

Comments: 4 pages, 4 figures

Journal ref: Automat. Comput. Appl. Math. 16(1-2) (2007) 27-34

Sensitive Ants in Solving the Generalized Vehicle Routing Problem

Authors: Camelia-M. Pintea, Camelia Chira, D. Dumitrescu, Petrica C. Pop

Abstract: The idea of sensitivity in ant colony systems has been exploited in hybrid ant-based models with promising results for many combinatorial optimization problems. Heterogeneity is induced in the ant population by endowing individual ants with a certain level of sensitivity to the pheromone trail. The variable pheromone sensitivity within the same population of ants can potentially intensify the sear… ▽ More The idea of sensitivity in ant colony systems has been exploited in hybrid ant-based models with promising results for many combinatorial optimization problems. Heterogeneity is induced in the ant population by endowing individual ants with a certain level of sensitivity to the pheromone trail. The variable pheromone sensitivity within the same population of ants can potentially intensify the search while in the same time inducing diversity for the exploration of the environment. The performance of sensitive ant models is investigated for solving the generalized vehicle routing problem. Numerical results and comparisons are discussed and analysed with a focus on emphasizing any particular aspects and potential benefits related to hybrid ant-based models. △ Less

Submitted 27 August, 2012; originally announced August 2012.

Comments: 5 pages

MSC Class: 68T20

Journal ref: INT J COMPUT COMMUN, VI(4):731-738 (2011)

New results of ant algorithms for the Linear Ordering Problem

Authors: Camelia-M. Pintea, Camelia Chira, D. Dumitrescu

Abstract: Ant-based algorithms are successful tools for solving complex problems. One of these problems is the Linear Ordering Problem (LOP). The paper shows new results on some LOP instances, using Ant Colony System (ACS) and the Step-Back Sensitive Ant Model (SB-SAM). Ant-based algorithms are successful tools for solving complex problems. One of these problems is the Linear Ordering Problem (LOP). The paper shows new results on some LOP instances, using Ant Colony System (ACS) and the Step-Back Sensitive Ant Model (SB-SAM). △ Less

Submitted 27 August, 2012; originally announced August 2012.

Comments: 5 pages, 5 figures Zbl:06048718

MSC Class: 68T20

Journal ref: An. Univ. Vest Timis., Ser. Mat.-Inform. 48, No. 3, 139-150 (2010)

Emergence of Techno-Social Norms in Cognitive Radio Environments

Authors: Ligia Cremene, D. Dumitrescu

Abstract: The aim of this paper is to explore the potential of Game Theory (GT) in extracting rules of behaviour for emerging Cognitive Radio environments. We revisit the commons approach to unlicensed spectrum and try to show that a commons can be basically regulated from the inside out. GT simulations of CR interactions reveal the emergence of certain equilibria mirroring behaviours/trends?. Once these ?t… ▽ More The aim of this paper is to explore the potential of Game Theory (GT) in extracting rules of behaviour for emerging Cognitive Radio environments. We revisit the commons approach to unlicensed spectrum and try to show that a commons can be basically regulated from the inside out. GT simulations of CR interactions reveal the emergence of certain equilibria mirroring behaviours/trends?. Once these ?trends identified, norms may be expressed and then embedded into machines (CRs). Internalized norms may thus become the alternative to external enforcement of rules. We call these emerging norms techno-social norms (TSNs). TSNs could eventually become a means of regulating the use of unlicensed spectrum and making open spectrum access feasible. Open spectrum access scenarios are considered and analysis is performed based on reformulations of two game theoretical models: Cournot and Bertrand. The standard oligopoly models are reformulated in terms of radio resource access in unlicensed bands. In order to capture the large variety of CR interaction situations, several GT equilibrium concepts are considered: Nash, Pareto, Berge-Zhukovskii, and Lorenz. In order to capture the heterogeneity of CR interactions, the standard GT model is enriched allowing players to be biased toward different types of equilibrium (or rationality). An evolutionary game-equilibrium detection method is used. Numerical simulations bring relevant insights on the problem of autonomy vs. regulation in emerging CR environments. Relying on extensive GT simulations, some rules of behaviour - to be expanded into techno-social norms - may be derived. △ Less

Submitted 13 July, 2012; originally announced July 2012.

Comments: ("This work has been submitted to the IEEE for possible publication")

Overcoming Hierarchical Difficulty by Hill-Climbing the Building Block Structure

Authors: David Iclanzan, Dan Dumitrescu

Abstract: The Building Block Hypothesis suggests that Genetic Algorithms (GAs) are well-suited for hierarchical problems, where efficient solving requires proper problem decomposition and assembly of solution from sub-solution with strong non-linear interdependencies. The paper proposes a hill-climber operating over the building block (BB) space that can efficiently address hierarchical problems. The new… ▽ More The Building Block Hypothesis suggests that Genetic Algorithms (GAs) are well-suited for hierarchical problems, where efficient solving requires proper problem decomposition and assembly of solution from sub-solution with strong non-linear interdependencies. The paper proposes a hill-climber operating over the building block (BB) space that can efficiently address hierarchical problems. The new Building Block Hill-Climber (BBHC) uses past hill-climb experience to extract BB information and adapts its neighborhood structure accordingly. The perpetual adaptation of the neighborhood structure allows the method to climb the hierarchical structure solving successively the hierarchical levels. It is expected that for fully non deceptive hierarchical BB structures the BBHC can solve hierarchical problems in linearithmic time. Empirical results confirm that the proposed method scales almost linearly with the problem size thus clearly outperforms population based recombinative methods. △ Less

Submitted 16 February, 2007; originally announced February 2007.

Comments: submited to GECCO 2007 (jan 31)

ACM Class: G.1.6; I.2.8