Scopus EXPORT DATE: 01 August 2024 Joshi R., Deshpande B. AUTHOR FULL NAMES: Joshi, Ramprasad (24828982600); Deshpande, Bharat (24829204000) 24828982600; 24829204000 Empirical and analytical study of many-objective optimization problems: Analysing distribution of nondominated solutions and population size for scalability of randomized heuristics (2014) Memetic Computing, 6 (2), pp. 133 - 145 DOI: 10.1007/s12293-014-0133-y https://www.scopus.com/inward/record.uri?eid=2-s2.0-84901593722&doi=10.1007%2fs12293-014-0133-y&partnerID=40&md5=a5bfb7bb943927ec35287a36d94b9eed ABSTRACT: Nature inspired randomized heuristics have been used successfully for single-objective and multi-objective optimization problems. However, with increasing number of objectives, what are called as "dominance resistant solutions" present a challenge to heuristics because they make it harder to locate and converge to the Pareto-optimal front. In the present work, the scalability of population-based heuristics for many-objective problems is studied using techniques from probability theory. Work in this domain tends to be more problem-specific and is largely empirical. Here a more general theoretical framework to study the problem arising from escalation of objectives is developed. This framework allows application of probability concentration inequalities to complicated multiobjective optimization heuristics. It also helps isolate the effects of escalation of objective space dimension from those of problem structure and of design space dimension. It opens up the possibility of combining the framework with more problem-specific models and with empirical work, to tune algorithms and to make problems amenable to heuristic search. © 2014 Springer-Verlag Berlin Heidelberg. AUTHOR KEYWORDS: Parameter tuning; Population-based heuristics INDEX KEYWORDS: Heuristic algorithms; Pareto principle; Population statistics; Scalability; Concentration inequality; Many-objective optimizations; Multi-objective optimization problem; Nondominated solutions; Parameter-tuning; Population-based heuristics; Randomized heuristics; Theoretical framework; Multiobjective optimization DOCUMENT TYPE: Article PUBLICATION STAGE: Final SOURCE: Scopus Bedekar M., Deshpande B., Joshi R. AUTHOR FULL NAMES: Bedekar, Mangesh (24828950900); Deshpande, Bharat (24829204000); Joshi, Ramprasad (24828982600) 24828950900; 24829204000; 24828982600 Web search personalization by user profiling (2008) Proceedings - 1st International Conference on Emerging Trends in Engineering and Technology, ICETET 2008, art. no. 4580067, pp. 1099 - 1103 DOI: 10.1109/ICETET.2008.70 https://www.scopus.com/inward/record.uri?eid=2-s2.0-51949106985&doi=10.1109%2fICETET.2008.70&partnerID=40&md5=9c21f86afe8728f27076042128d796e7 ABSTRACT: The World Wide Web is growing at a rate of about a million pages per day, making it tougher for search engines to extract relevant information for its users. Earlier Search Engines used simple indexing techniques to search for keywords in websites and gave more weightage to pages with higher frequency of keyword occurrences. This technique was easy to trick by using meta-tags liberally, claiming that their page used popular search terms, thereby, made meta-tags useless for search engines. Another technique widely used was to repeatedly use popular search terms in invisible text (white text on a white background) to fool engines. These fallacies called for a set of algorithms which would sort the results using an unbiased parameter. The currently employed Link Analysis Algorithms make use of the structure present in 'hyperlinks', sorted and displayed depending on a 'popularity index' decided to pages linking to it. In this work, we have analyzed the mathematics behind these 'link analysis algorithms' and their effective use in ecommerce applications where they could be used for displaying 'personalized information'. © 2008 IEEE. AUTHOR KEYWORDS: Apriori algorithm PageRank; Personalization; Web data mining INDEX KEYWORDS: Computer software; Hypertext systems; Industrial economics; Information retrieval; Information services; Internet; Search engines; Technology; Websites; Apriori algorithm PageRank; E-Commerce applications; Emerging trends; Hyper links; Indexing techniques; International conferences; Link analysis algorithms; Meta-tags; Personalization; Personalized information; Relevant information; Search terms; User profiling; Web data mining; Web searches; World Wide Web CONFERENCE NAME: 1st International Conference on Emerging Trends in Engineering and Technology, ICETET 2008 CONFERENCE DATE: 16 July 2008 through 18 July 2008 CONFERENCE LOCATION: Nagpur, Maharashtra CONFERENCE CODE: 73554 DOCUMENT TYPE: Conference paper PUBLICATION STAGE: Final SOURCE: Scopus Rajan A., Salgaonkar A., Joshi R. AUTHOR FULL NAMES: Rajan, Annie (57215688946); Salgaonkar, Ambuja (35485507600); Joshi, Ramprasad (24828982600) 57215688946; 35485507600; 24828982600 A survey of Konkani NLP resources (2020) Computer Science Review, 38, art. no. 100299 DOI: 10.1016/j.cosrev.2020.100299 https://www.scopus.com/inward/record.uri?eid=2-s2.0-85097499704&doi=10.1016%2fj.cosrev.2020.100299&partnerID=40&md5=397ff6812e380c241a1da354d6047dd3 ABSTRACT: The first comprehensive survey is presented of natural language processing (NLP) research in Konkani, a low-resource regional Indian language with a small population of native speakers. The combined challenges of complex linguistic phenomena, paucity of documented resources and the presence of neighboring dominant languages are elaborated. The possibilities of crowdsourcing as a means for creating linguistic resources are explored. © 2020 Elsevier Inc. AUTHOR KEYWORDS: Konkani language; Morphological analysis; Natural language processing; Part-of-speech; Text to speech INDEX KEYWORDS: Linguistics; Surveys; Indian languages; Linguistic phenomena; Linguistic resources; NAtural language processing; Small population; Natural language processing systems DOCUMENT TYPE: Review PUBLICATION STAGE: Final SOURCE: Scopus Chakrabarty S., Joshi R.S. AUTHOR FULL NAMES: Chakrabarty, Sangeeta (57214910776); Joshi, Ramprasad S. (24828982600) 57214910776; 24828982600 Dark Data: People to People Recovery (2020) Lecture Notes in Networks and Systems, 93, pp. 247 - 254 DOI: 10.1007/978-981-15-0630-7_24 https://www.scopus.com/inward/record.uri?eid=2-s2.0-85079480513&doi=10.1007%2f978-981-15-0630-7_24&partnerID=40&md5=5b3e2b11c9c06a8f720833567bdf5fb4 ABSTRACT: The Internet of Things (IoT) is exploding disruptively. The IoT is making life easier for ordinary consumers and workers, but it is also generating zettabytes of dark data. The real time analytics involving close interaction between humans and instruments on the Internet is the main commercial motivation behind the IoT revolution. This means the data that are consumed instantly and their interpretation that is filed in an indexed and structured form are the main productive outcomes, while dark data and haphazardly stored interpretations add to the tare, bringing down efficiency and increasing costs. Since out of all data stored in the world today, almost all are generated in the recent three years, and the phenomenal growth will soon lead to a crisis, we need to put in place a global framework that never lets dark data clog the information highways but actually harnesses the real time analytics for a better planned future. We propose here a plan to build a “Data Waste Management” or “Data Sewerage” and local “Data Reservoir” system. © Springer Nature Singapore Pte Ltd 2020. DOCUMENT TYPE: Book chapter PUBLICATION STAGE: Final SOURCE: Scopus Joshi R., Deshpande B. AUTHOR FULL NAMES: Joshi, Ramprasad (24828982600); Deshpande, Bharat (24829204000) 24828982600; 24829204000 Scalability of population-based search heuristics for many-objective optimization (2013) Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 7835 LNCS, pp. 479 - 488 DOI: 10.1007/978-3-642-37192-9_48 https://www.scopus.com/inward/record.uri?eid=2-s2.0-84875676539&doi=10.1007%2f978-3-642-37192-9_48&partnerID=40&md5=cb91c565f5aa25c64a5462b9d32854bd ABSTRACT: Beginning with Talagrand [16]'s seminal work, isoperimetric inequalities have been used extensively in analysing randomized algorithms. We develop similar inequalities and apply them to analysing population-based randomized search heuristics for multiobjective optimization in ân space. We demonstrate the utility of the framework in explaining an empirical observation so far not explained analytically: the curse of dimensionality, for many-objective problems. The framework makes use of the black-box model now popular in EC research. © Springer-Verlag Berlin Heidelberg 2013. AUTHOR KEYWORDS: MOEA; Multiobjective optimization; Probability Measure Theory; Talagrand-type Inequalities INDEX KEYWORDS: Computation theory; Heuristic algorithms; Curse of dimensionality; Isoperimetric inequalities; Many-objective optimizations; MOEA; Probability measures; Randomized Algorithms; Randomized search heuristics; Talagrand-type Inequalities; Multiobjective optimization SPONSORS: Algorithms, Algorithms Data Struct. Group, Inst. Informatics Digit. Innov. Edinburgh Napier Univ., Vienna Univ. Technol., Inst. Comput. Graph. CONFERENCE NAME: 16th European Conference on Applications of Evolutionary Computation, EvoApplications 2013 CONFERENCE DATE: 3 April 2013 through 5 April 2013 CONFERENCE LOCATION: Vienna CONFERENCE CODE: 96330 DOCUMENT TYPE: Conference paper PUBLICATION STAGE: Final SOURCE: Scopus Safdari M., Joshi R. AUTHOR FULL NAMES: Safdari, Mustafa (26665068400); Joshi, Ramprasad (24828982600) 26665068400; 24828982600 Evolving universal hash functions using genetic algorithms (2009) Proceedings - 2009 International Conference on Future Computer and Communication, ICFCC 2009, art. no. 5189747, pp. 84 - 87 DOI: 10.1109/ICFCC.2009.66 https://www.scopus.com/inward/record.uri?eid=2-s2.0-70449450526&doi=10.1109%2fICFCC.2009.66&partnerID=40&md5=705591f017b5ddf6c3ac09089806700c ABSTRACT: In this paper we explore using Genetic Algorithms to construct Universal Hash Functions to efficiently hash a given set of keys. The Hash Function generated in this way should give minimum number of collisions. The algorithm has less computational complexity and can be used in scenarios where the input distribution of keys is changing and the hash function needs to be modified often to rehash the values. © 2009 IEEE. AUTHOR KEYWORDS: Genetic algorithms; Key distribution; Universal hash functions INDEX KEYWORDS: Computational complexity; Computational efficiency; Genetic algorithms; Hash functions; Key distribution; Universal Hash Function; Universal hash functions; Distribution functions CONFERENCE NAME: 2009 International Conference on Future Computer and Communication, ICFCC 2009 CONFERENCE DATE: 3 April 2009 through 5 April 2009 CONFERENCE LOCATION: Kuala Lumpar CONFERENCE CODE: 78291 DOCUMENT TYPE: Conference paper PUBLICATION STAGE: Final SOURCE: Scopus Asgaonkar A., Palande P., Joshi R.S. AUTHOR FULL NAMES: Asgaonkar, Aditya (57201333736); Palande, Pranav (57201334944); Joshi, Ramprasad S. (24828982600) 57201333736; 57201334944; 24828982600 Is the cost of proof-of-work consensus quasilinear? (2018) ACM International Conference Proceeding Series, pp. 314 - 317 DOI: 10.1145/3152494.3167978 https://www.scopus.com/inward/record.uri?eid=2-s2.0-85044345859&doi=10.1145%2f3152494.3167978&partnerID=40&md5=13063ec1412ac605d6d001123989a2ab ABSTRACT: The increasing popularity of Bitcoin, Ethereum and other cryp-tocurrencies has led to a rising interest in its underlying blockchain technology. Blockchains serve as distributed ledgers, and are fundamentally different from traditional distributed databases. In view of this large scale adoption of blockchain technology, it is of interest to analyze the performance of the underlying mechanisms in peer-to-peer blockchain networks. In this work, we simulate a model for a peer-to-peer blockchain network relying on the Proof-of-Work (PoW) consensus mechanism, and present an analysis of the (overall, peer-to-peer) cost and throughput of the PoW consensus mechanism. We find compelling empirical evidence that the cost of the PoW consensus is superlinearithmic and subquadratic in the size of the network. Indeed, the number of necessary syncing calls, to maintain throughput scalable with the size, grow in that manner. © 2018 Association for Computing Machinery. AUTHOR KEYWORDS: Blockchain; Consensus Protocol; Performance Analysis; Proof-of-Work INDEX KEYWORDS: Blockchain; Costs; Peer to peer networks; Bitcoin; Consensus protocols; Distributed database; Peer to peer; Performance analysis; Proof of work; Quasi-linear; Cost benefit analysis CONFERENCE NAME: ACM India Joint 5th International Conference on Data Science and 23rd Conference on Management of Data, CoDS-COMAD 2018 CONFERENCE DATE: 11 January 2018 through 13 January 2018 CONFERENCE LOCATION: Goa CONFERENCE CODE: 134482 DOCUMENT TYPE: Conference paper PUBLICATION STAGE: Final SOURCE: Scopus Joshi R., Deshpande B., Gote P. AUTHOR FULL NAMES: Joshi, Ramprasad (24828982600); Deshpande, Bharat (24829204000); Gote, Paritosh (56440282700) 24828982600; 24829204000; 56440282700 Objective dimension and problem structure in multiobjective optimization problems (2014) Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 8602, pp. 639 - 650 DOI: 10.1007/978-3-662-45523-4_52 https://www.scopus.com/inward/record.uri?eid=2-s2.0-84915750161&doi=10.1007%2f978-3-662-45523-4_52&partnerID=40&md5=81ce6181a94378d7770d2de90a4684d0 ABSTRACT: Multiobjective optimization seeks simultaneous minimization of multiple scalar functions on ℝn. Unless weighted sums are made to replace the vector functions arising thus, such an optimization requires some partial-or quasi-ordering of points in the search space based on comparisons between the values attained by the functions to be optimized at those points. Many such orders can be defined, and searchbased (mainly heuristic) optimization algorithms make use of such orders implicitly or explicitly for refining and accelerating search. In this work, such relations are studied by modeling them as graphs. Information apparent in the structure of such graphs is studied in the form of degree distribution. It is found that when the objective dimension grows, the degree distribution tends to follow a power-law. This can be a new beginning in the study of escalation of hardness of problems with dimension, as also a basis for designing new heuristics. © Springer-Verlag Berlin Heidelberg 2014. INDEX KEYWORDS: Heuristic algorithms; Structural optimization; Vector spaces; Degree distributions; Multi-objective optimization problem; Optimization algorithms; Problem structure; Quasi-ordering; Scalar function; Search spaces; Vector functions; Multiobjective optimization SPONSORS: Free Software Office (OSL) of the University of Granada, Granada Excellence Network of Innovation Laboratories (GENIL), Institute for Informatics and Digital Innovation at Edinburgh Napier University, UK CONFERENCE NAME: 17th European Conference on Applications of Evolutionary Computation, EvoApplications 2014 CONFERENCE DATE: 23 April 2014 through 25 April 2014 CONFERENCE LOCATION: Granada CONFERENCE CODE: 111599 DOCUMENT TYPE: Conference paper PUBLICATION STAGE: Final SOURCE: Scopus Dash T., Srinivasan A., Joshi R.S., Baskar A. AUTHOR FULL NAMES: Dash, Tirtharaj (56464021900); Srinivasan, Ashwin (7202314451); Joshi, Ramprasad S. (24828982600); Baskar, A. (35434348700) 56464021900; 7202314451; 24828982600; 35434348700 Discrete Stochastic Search and Its Application to Feature-Selection for Deep Relational Machines (2019) Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 11728 LNCS, pp. 29 - 45 DOI: 10.1007/978-3-030-30484-3_3 https://www.scopus.com/inward/record.uri?eid=2-s2.0-85072868661&doi=10.1007%2f978-3-030-30484-3_3&partnerID=40&md5=7b44db8ec9ce394a6d484b55669b6996 ABSTRACT: We use a model for discrete stochastic search in which one or more objects (“targets”) are to be found by a search over n locations (“boxes”), where n is infinitely large. Each box has some probability that it contains a target, resulting in a distribution H over boxes. We model the search for the targets as a stochastic procedure that draws boxes using some distribution S. We derive first a general expression on the expected number of misses E [ Z] made by the search procedure in terms of H and S. We then obtain an expression for an optimal distribution S∗ to minimise E [ Z]. This results in a relation between: the entropy of H and the KL-divergence between H and S∗. This result induces a 2-partitions over the boxes consisting of those boxes with H probability greater than 1n and the rest. We use this result to devise a stochastic search procedure for the practical situation when H is unknown. We present results from simulations that agree with theoretical predictions; and demonstrate that the expected misses by the optimal seeker decreases as the entropy of H decreases, with a maximum obtained for uniform H. Finally, we demonstrate applications of this stochastic search procedure with a coarse assumption about H. The theoretical results and the procedure are applicable to stochastic search over any aspect of machine learning that involves a discrete search-space: for example, choice over features, structures or discretized parameter-selection. In this work, the procedure is used to select features for Deep Relational Machines (DRMs) which are Deep Neural Networks (DNNs) defined in terms of domain-specific knowledge and built with features selected from large, potentially infinite-attribute space. Empirical results obtained across over 70 real-world datasets show that using the stochastic search procedure results in significantly better performances than the state-of-the-art. © 2019, Springer Nature Switzerland AG. AUTHOR KEYWORDS: Deep Neural Network; Inductive Logic Programming; Infinite-attribute space; Relational learning; Stochastic search INDEX KEYWORDS: Deep neural networks; Entropy; Feature extraction; Inductive logic programming (ILP); Machine learning; Neural networks; Probability distributions; Stochastic models; Domain-specific knowledge; Infinite-attribute space; Optimal distributions; Parameter selection; Real-world datasets; Relational learning; Stochastic procedure; Stochastic search; Stochastic systems CONFERENCE NAME: 28th International Conference on Artificial Neural Networks, ICANN 2019 CONFERENCE DATE: 17 September 2019 through 19 September 2019 CONFERENCE LOCATION: Munich CONFERENCE CODE: 231689 DOCUMENT TYPE: Conference paper PUBLICATION STAGE: Final SOURCE: Scopus Verma D., Joshi R., Shivani A., Gupta R. AUTHOR FULL NAMES: Verma, Devika (57193139003); Joshi, Ramprasad (24828982600); Shivani, Aiman (58754633400); Gupta, Rohan (58754633500) 57193139003; 24828982600; 58754633400; 58754633500 Karaka-Based Answer Retrieval for Question Answering in Indic Languages (2023) International Conference Recent Advances in Natural Language Processing, RANLP, pp. 1216 - 1224 DOI: 10.26615/978-954-452-092-2_129 https://www.scopus.com/inward/record.uri?eid=2-s2.0-85179176889&doi=10.26615%2f978-954-452-092-2_129&partnerID=40&md5=6ec58e837b2331fc8c4a740506e60c55 ABSTRACT: Karakas from ancient Paninian grammar form a concise set of semantic roles that capture crucial aspect of sentence meaning pivoted on the action verb. In this paper, we propose employing a karaka-based approach for retrieving answers in Indic question-answering systems. To study and evaluate this novel approach, empirical experiments are conducted over large benchmark corpora in Hindi and Marathi. The results obtained demonstrate the effectiveness of the proposed method. Additionally, we explore the varying impact of two approaches for extracting karakas. The literature surveyed and experiments conducted encourage hope that karaka annotation can improve communication with machines using natural languages, particularly in low-resource languages. © 2023 Incoma Ltd. All rights reserved. INDEX KEYWORDS: Computational methods; Empirical experiments; Low resource languages; Natural languages; Question Answering; Question answering systems; Semantic roles; Semantics SPONSORS: Bulgarian National Research Fund, Cambridge University Press, ELDA, Iris.AI, Ontotext, Senso CONFERENCE NAME: 2023 International Conference Recent Advances in Natural Language Processing: Large Language Models for Natural Language Processing, RANLP 2023 CONFERENCE DATE: 4 September 2023 through 6 September 2023 CONFERENCE LOCATION: Varna CONFERENCE CODE: 194755 DOCUMENT TYPE: Conference paper PUBLICATION STAGE: Final OPEN ACCESS: All Open Access; Bronze Open Access SOURCE: Scopus Singh S., Joshi R.P., Kohli H. AUTHOR FULL NAMES: Singh, Sunit (57191038651); Joshi, Ram Prasad (24828982600); Kohli, Harsh (7005237397) 57191038651; 24828982600; 7005237397 Optimal Route Searching in Networks with Dynamic Weights Using Flow Algorithms (2016) Proceedings - 2015 International Conference on Computational Intelligence and Communication Networks, CICN 2015, art. no. 7546072, pp. 146 - 155 DOI: 10.1109/CICN.2015.37 https://www.scopus.com/inward/record.uri?eid=2-s2.0-84985911245&doi=10.1109%2fCICN.2015.37&partnerID=40&md5=48bae63d51280fdfa857d0b1394b0b6c ABSTRACT: A network with dynamic weights implies a set of vertices interconnected by a set of edges, each of which bears a weight that changes with time. One example of such networks is a traffic network, wherein, the structure of the graph remains constant but the weight on the edges, signifying the amount of traffic (traffic density) changes over time. We have dealt with scenarios where flow algorithm needs to run repeatedly to establish flows in a network with timely changing capacities and we have sought to obtain some form of computational intelligence on that subject. We have aligned our work to the context of traffic networks in order to explore practical inspection of the same. However, this study applies equally for any generic network with continuously changing capacities which requires flow re-setting time and again. © 2015 IEEE. AUTHOR KEYWORDS: Edmonds-Karp Algorithm; Ford Fulkerson Approach; Minimum Cut-Maximum Flow; Traffic Flow INDEX KEYWORDS: Algorithms; Transportation; Flow algorithm; Ford Fulkerson Approach; Generic networks; Maximum flows; Optimal routes; Traffic densities; Traffic flow; Traffic networks; Artificial intelligence CONFERENCE NAME: 7th International Conference on Computational Intelligence and Communication Networks, CICN 2015 CONFERENCE DATE: 12 December 2015 through 14 December 2015 CONFERENCE LOCATION: Jabalpur CONFERENCE CODE: 123405 DOCUMENT TYPE: Conference paper PUBLICATION STAGE: Final SOURCE: Scopus Verma D.A., Joshi R.S., Joshi S.A., Susladkar O.K. AUTHOR FULL NAMES: Verma, Devika A. (57193139003); Joshi, Ramprasad S. (24828982600); Joshi, Shubhamkar A. (57560365000); Susladkar, Onkar K. (57559924600) 57193139003; 24828982600; 57560365000; 57559924600 Study of Similarity Measures as Features in Classification for Answer Sentence Selection Task in Hindi Question Answering: Language-Specific v/s Other Measures (2021) Proceedings of the 35th Pacific Asia Conference on Language, Information and Computation, PACLIC 2021, pp. 715 - 724 https://www.scopus.com/inward/record.uri?eid=2-s2.0-85127494337&partnerID=40&md5=38723a2452fbfbfeaf4e1e9e7ccb0dcb ABSTRACT: Answer sentence selection is an important sub-task in Question Answering (QA) that determines the correct answer sentence from a passage. This task can naturally be reduced to the semantic text similarity problem between question and answer candidate. In this work, we investigate the significance of various similarity measures for the answer sentence selection task in Hindi an Indo-Aryan language. Karaka relations is the core of dependency annotation scheme used for Hindi and are crucial to syntactico-semantic analysis of the sentence. We investigate this, and compare them to other, hitherto known measures. To investigate and compare the utility of various measures, we develop a test-bench over a benchmark Hindi and English multilingual QA corpus for comparison, making two tool-chains and designing empirical experiments across combinations of similarity measures, sentence embedding schemes, and supervised machine learning models for classification. Combining Karaka relations with different similarity measures shows significant performance improvement for sentence selection task, suggesting them as potentially a semantic similarity measure. Moreover, our results give us confidence that refinement of Karaka relations extraction to optimal quality will reduce the need for availability of large pre-trained language models. © Proceedings of the 35th Pacific Asia Conference on Language, Information and Computation, PACLIC 2021. INDEX KEYWORDS: Classification (of information); Natural language processing systems; Supervised learning; Annotation scheme; Embeddings; Empirical experiments; Question Answering; Semantic analysis; Sentence selection; Similarity measure; Subtask; Test-bench; Text similarity; Semantics CONFERENCE NAME: 35th Pacific Asia Conference on Language, Information and Computation, PACLIC 2021 CONFERENCE DATE: 5 November 2021 through 7 November 2021 CONFERENCE LOCATION: Shanghai CONFERENCE CODE: 177552 DOCUMENT TYPE: Conference paper PUBLICATION STAGE: Final SOURCE: Scopus