The most important part of the train's energy is consumed by the traction system. The tractiv... more The most important part of the train's energy is consumed by the traction system. The tractive energy depends mainly on the driving behaviour. Improving driving strategies has great potential to enhance the energy efficiency. This paper presents a speed profile optimization approach based on a genetic algorithm. The objective of the genetic algorithm is to find, for each interstation, the best speed profile which minimizes the energy consumption. The optimized profile takes into account both the physical and the operational constraints such as the maximum allowed travel time, the speed limitations per section and the maximum allowed acceleration and jerk. The fitness function is based on a Random Forest model which is built using on-board measurements. The aim of the model is to estimate accurately the energy consumption corresponding to each speed profile. The initial population of genetic algorithm is mainly composed of the best realistic speed profiles extracted from the coll...
2021 International Symposium on Computer Science and Intelligent Controls (ISCSIC), 2021
Demand estimation in public transport is critical for transport stakeholders. Thanks to the emerg... more Demand estimation in public transport is critical for transport stakeholders. Thanks to the emerging technologies in recent years, many sources of mobility data are available to model passengers flow in public transport network. One of the most added-value mobility data is smart card Origin-Destination (OD) data. These data could inform us on when, where and how flows transit within the network. The OD matrix used in this work is obtained from smart card data collected by Automated Fare Collection (AFC) system in the Greater Paris Area which is called Navigo Pass. Despite its immense value, this matrix doesn't cover the entire passenger flow. This is due to fraud, other types of tickets (e.g. the standard paper ticket) and uncertainties in destination estimates. In this paper we propose a twostep approach for correcting and scaling smart card OD matrix based on adapted Growth Factor model considering the complexity caused by temporal variation of the OD matrix. In the first step we map all the OD pairs in the OD matrix over our area of study to infer their departure and arrival stations and time. In the second step we exploit passengers' counting data and use growth factor model to scale the OD matrix to obtain a new corrected matrix which can present the real flow in the transit network. We apply our proposed methodology to scale an OD matrix constructed only from smart card validation data which presents between 40% to 65% of the overall flow. For this purpose, passengers' counting data are exploited.
Massive data are surrounding us in our daily lives. Urban mobility generates a very high number o... more Massive data are surrounding us in our daily lives. Urban mobility generates a very high number of complex data reflecting the mobility of people, vehicles and objects. Transport operators are primary users who strive to discover the meaning of phenomena behind traffic data, aiming at regulation and transport planning. This paper tackles the question "How to design a supportive tool for visual exploration of digital mobility data to help a transport analyst in decision making?” The objective is to support an analyst to conduct an ex post analysis of train circulation and passenger flows, notably in disrupted situations. We propose a problem-solution process combined with data visualisation. It relies on the observation of operational agents, creativity sessions and the development of user scenarios. The process is illustrated for a case study on one of the commuter line of the Paris metropolitan area. Results encompass three different layers and multiple interlinked views to ex...
2020 17th International Conference on the European Energy Market (EEM), 2020
The need for deploying fast-charging stations for electric vehicles (EVs) is becoming essential i... more The need for deploying fast-charging stations for electric vehicles (EVs) is becoming essential in recent years. This need is justified by the increasing recharging demand, and supported by new recharging technologies making EV chargers more efficient. For this purpose, we introduce an optimization model for finding the optimal deployment of EV fast-charging stations through a network of highways, while taking into account their recharging demands and the restrictions imposed by the electric grid. We also consider the availability of local photovoltaic (PV) power station and integrate its energy to the proposed recharging network. In addition, we present a case study on Paris-Saclay area where the actual mobility flows and electric network specifications are considered. We thus provide the optimal locations for EV charging stations at the studied area, and we indicate the number of chargers to be deployed at each station. Finally, we study the potential benefits of integrating local PV energy and analyze its different prices and production costs.
2020 Forum on Integrated and Sustainable Transportation Systems (FISTS), 2020
Route planning in public transport receives an increasing interest in smart cities and particular... more Route planning in public transport receives an increasing interest in smart cities and particularly in metropolitan cities where crowded and jammed traffic is daily recorded in the transportation network. The availability of digital footprints, such as ticketing logs, or load on board the trains, provides a relevant opportunity to develop innovative decision-making tools for urban routing of passengers in order to assist them to better plan their journeys. In this paper, we propose to enrich existing journey planners with predictive travel information to enhance the passenger travel experience during his journey. For that purpose, we augment the planned trips with predictive passenger flow indicators such as the load on board trains, and passenger attendees at the station. These indicators are forecasted along the journey with the help of the developed machine learning models. The experiments are conducted on a real historical dataset covering the Paris Region with a focus on a railway transit network that serves mainly the suburb of Paris.
2020 23rd Euromicro Conference on Digital System Design (DSD), 2020
Management of a multimodal transport network is a challenging task and operators regularly have t... more Management of a multimodal transport network is a challenging task and operators regularly have to deal with different types of disturbances that affect the quality of service of the transport system. To support them in their decision-making, we present the architecture of a multimodal supervision system that implements monitoring, prediction of affluence, disturbances, impacts and evaluation of functionalities focusing on one or multiples lines of the network. Our system uses the real-time data of transport operators to compute several key performance indicators (KPI) in order to monitor the network status and to detect disturbances. The prediction function is used to predict passengers’ attendance at stations, incident duration and the evolution of the whole network, in particular, phenomena emerging from the interconnections linked to the affected line. Finally, we present a use case of the supervision system applied to a real-time control of a bus line belonging to a very dense and busy segment of the multimodal transport network in Ile-de-France area.
Route planning in public transport receive an increasing interest in smart cities and particularl... more Route planning in public transport receive an increasing interest in smart cities and particularly in metropolitan cities where crowded and jammed traffic is daily recorded in transportation network. The availability of digital footprints such as ticketing logs, or load on board the trains provide a relevant opportunity to develop innovative decision-making tools for urban routing of passengers in order to assist them to better planning their journeys. This planning must consider the forthcoming evolution of the traffic in order to adapt its response to the next state of the network and thus avoid unpleasant situations for the passengers. In this paper, we present a system for individual trip planning that incorporates short and long terms forecasting of different indicators related to the station attendance, occupancy of the trains, and delays in train schedule.
Le travail que nous presentons ici traite de la problematique de l'etiquetage morpho-syntaxiq... more Le travail que nous presentons ici traite de la problematique de l'etiquetage morpho-syntaxique des corpus de specialite non annotes. Les etiqueteurs existants sont entraines sur divers corpus, et produisent des erreurs d'etiquetage dans les textes de specialite. La methode la plus triviale pour apprendre des automates adaptes a un domaine specialise est d'etiqueter manuellement de grands corpus du domaine, ce qui necessite un travail tres important. Pour eviter cela, nous proposons une methode semi-automatique. ETIQ, le nouvel etiqueteur que nous avons developpe, permet de corriger la base de regles obtenue par l‘etiqueteur de BRILL et de l‘adapter a un corpus de specialite. L'expert du domaine visualise l‘etiquetage de base et le corrige par l‘ecriture et l‘application de regles lexicales et contextuelles specialisees. Les regles inserees sont plus expressives que celles de Brill. Pour aider l'expert dans sa tâche, nous avons concu un algorithme inductif biaise...
2021 IEEE International Smart Cities Conference (ISC2), 2021
Public transport route planning is of growing interest in smart cities and especially in metropol... more Public transport route planning is of growing interest in smart cities and especially in metropolitan areas where congestions and traffic jams are frequently recorded. The availability of multiple data sources, such as passenger load in trains or ticketing logs, provides an interesting opportunity to develop decision support tools to help passengers better plan their trips around the city and to enhance their travel experience. We present, in this paper, a multi-criteria journey planner that incorporates train load predictions as criteria. To this end, on the one hand, we enrich the proposed routes with predictive indicators of passenger flow such as the load on board the trains. These indicators are computed for each section of the itinerary using machine learning algorithms. On the other hand, we design a journey planner that incorporates the predicted load in its search criteria.
The need for deploying fast-charging stations for electric vehicles (EVs) is becoming essential i... more The need for deploying fast-charging stations for electric vehicles (EVs) is becoming essential in recent years. This need is justified by the increasing charging demand and supported by new charging technologies making EV chargers more efficient. In this paper, we provide a survey on EV fast-charging models and introduce a data-driven approach with an optimization model for deploying EV fast-chargers for both electric vehicles and heavy trucks traveling through a network of suburban highways. This deployment aims at satisfying EV charging demands while respecting the limits imposed by the electric grid. We also consider the availability of local photovoltaic (PV) farm and integrate its produced power to the proposed charging network. Finally, through a case study on Paris-Saclay area, we provide locations for EV charging stations and analyze the benefits of integrating PV power at different prices, production costs and charging capacities. The obtained results also suggest potentia...
Résumé. Le DÉ Fouille de Textes (DEFT) a consisté à supprimer les phrases non pertinentes dans un... more Résumé. Le DÉ Fouille de Textes (DEFT) a consisté à supprimer les phrases non pertinentes dans un corpus de discours politiques en français. Il a eu lieu en 2005 et réuni onze équipes, totalisant une trentaine de participants. Cet article décrit les prétraitements eectués sur les corpus de F. Mitterrand et de J. Chirac dans le cadre de ce dé. Notamment, la conversion au format texte, le découpage en phrases, le classement des discours, l'introduction de phrases de F. Mitterrand dans les discours de J. Chirac et l'identication des dates et noms de personnes. Les résultats obtenus par les onze équipes participantes sont aussi présentés.
Le DE Fouille de Textes (DEFT) a consiste a supprimer les phrases non pertinentes dans un corpus ... more Le DE Fouille de Textes (DEFT) a consiste a supprimer les phrases non pertinentes dans un corpus de discours politiques en francais. Il a eu lieu en 2005 et reuni onze equipes, totalisant une trentaine de participants. Cet article decrit les pretraitements eectues sur les corpus de F. Mitterrand et de J. Chirac dans le cadre de ce de. Notamment, la conversion au format texte, le decoupage en phrases, le classement des discours, l'introduction de phrases de F. Mitterrand dans les discours de J. Chirac et l'identication des dates et noms de personnes. Les resultats obtenus par les onze equipes participantes sont aussi presentes.
ABSTRACT In this paper, we treat the problem of the grammatical tagging of non-annotated corpora ... more ABSTRACT In this paper, we treat the problem of the grammatical tagging of non-annotated corpora of specialty. The existing taggers are trained on general language corpora, and give inconsistent results on the specialized texts, as technical and scientific ones. In order to learn rules adapted to a specialized field, the usual approach labels manually a large corpus of this field. This is extremely time-consuming. We propose here a semi-automatic approach for tagging corpora of specialty. ETIQ, the new tagger we are building, make it possible to correct the base of rules obtained by Brill’s tagger and to adapt it to a corpus of specialty. The user visualizes an initial and basic tagging and corrects it either by extending Brill’s lexicon or by the insertion of specialized lexical and contextual rules. The inserted rules are richer and more flexible than Brill’s ones. To help the expert in this task, we designed an inductive algorithm biased by the “correct” knowledge he acquired beforehand. By using techniques of machine learning and enabling the expert to incorporate knowledge of the field in an interactive and friendly way, we improve the tagging of specialized corpora. Our approach has been applied to a corpus of molecular biology.
In the context of Part-of-Speech (PoS)-tagging of specialized corpora, we proposed an inductive a... more In the context of Part-of-Speech (PoS)-tagging of specialized corpora, we proposed an inductive approach focusing on the most 'important' PoStags because mistaking them can lead to a total misunderstanding of the text. After a standard tagging of a biological corpus by Brill's tagger, we noted persistent errors that are very hard to deal with. As an application, we studied two cases of different nature: first, confusion between past participle, adjective and preterit for verbs that end with 'ed'; second, confusion between plural nouns and verbs, 3rd person singular present. With a friendly user interface, the expert corrected the examples. Then, from these well-annotated examples, we induced rules using a propositional rule induction algorithm. Experimental validation showed improvement in tagging precision. The relevance of the terminology of the considered field, here molecular biology, is greatly improved when the number of these tagging errors decreases.
2008 3rd International Conference on Information and Communication Technologies: From Theory to Applications, 2008
In this paper, we describe a Text-Mining system used to extract archaeological knowledge from spe... more In this paper, we describe a Text-Mining system used to extract archaeological knowledge from specialized texts. Our approach relies on two basic principals. First, we use userfriendly tools enabling experts to transfer easily their knowledge. Second, our tools include inductive algorithms to reduce the experts' workload.
La compréhension de textes de spécialité nécessite unétiquetage morpho-syntaxique de bonne qualit... more La compréhension de textes de spécialité nécessite unétiquetage morpho-syntaxique de bonne qualité. Or, lorsque les textesétudiés sont issus de domaines spécifiques et peu usités, il est rare de disposer de dictionnaires et autres ressources lexicales fiables. Le logiciel que nous proposons permet d'utiliser unétiquetage réalisé par unétiqueteur généraliste, puis d'améliorer cetétiquetage en intégrant des connaissances d'experts du domaineétudié. Grâce au logiciel développé, il est relativement aisé pour un expert du domaine de détecter des erreurs d'étiquetage et de mettre en place des règles de ré-étiquetage. Ces règles peuventêtre obtenues de deux manières différentes : (1) soit en utilisant un langage de programmation permettant d'exprimer des règles complexes de ré-étiquetage, (2) soit par apprentissage automatique des règlesà partir d'exemples corrigés au moyen d'une interface dédiée. Cet apprentissage propose de nouvelles règlesà l'expert, acquises automatiquement.
The text-mining system we are building deals with the specific problem of identifying the instanc... more The text-mining system we are building deals with the specific problem of identifying the instances of relevant concepts present in the texts. Our system relies therefore on interactions between a field expert and the various linguistic modules we use, often adapted from existing ones, such as Brill's tagger or CMU's Link parser. We have developed learning procedures adapted to various steps of the linguistic treatment, mainly for grammatical tagging, terminology, and concept learning. Our interaction with the expert differs from classical supervised learning, in that the expert is not simply a resource who is only able to provide examples, and unable to provide the formalized knowledge underlying these examples. We are developing specific programming languages which enable the field expert to intervene directly in some of the linguistic tasks. Our approach is thus devoted to help one expert in one field to detect the concepts relevant for his/her field, using a large amount of texts. Our approach is made of two steps. The first one is an automatic approach that finds relevant and novel sentences in the texts. The second one is based on the expert's knowledge and finds more specific relevant sentences. Working on 50 different domains without an expert has been a challenge in itself, and explains our relatively poor results for the first Novelty task.
The most important part of the train's energy is consumed by the traction system. The tractiv... more The most important part of the train's energy is consumed by the traction system. The tractive energy depends mainly on the driving behaviour. Improving driving strategies has great potential to enhance the energy efficiency. This paper presents a speed profile optimization approach based on a genetic algorithm. The objective of the genetic algorithm is to find, for each interstation, the best speed profile which minimizes the energy consumption. The optimized profile takes into account both the physical and the operational constraints such as the maximum allowed travel time, the speed limitations per section and the maximum allowed acceleration and jerk. The fitness function is based on a Random Forest model which is built using on-board measurements. The aim of the model is to estimate accurately the energy consumption corresponding to each speed profile. The initial population of genetic algorithm is mainly composed of the best realistic speed profiles extracted from the coll...
2021 International Symposium on Computer Science and Intelligent Controls (ISCSIC), 2021
Demand estimation in public transport is critical for transport stakeholders. Thanks to the emerg... more Demand estimation in public transport is critical for transport stakeholders. Thanks to the emerging technologies in recent years, many sources of mobility data are available to model passengers flow in public transport network. One of the most added-value mobility data is smart card Origin-Destination (OD) data. These data could inform us on when, where and how flows transit within the network. The OD matrix used in this work is obtained from smart card data collected by Automated Fare Collection (AFC) system in the Greater Paris Area which is called Navigo Pass. Despite its immense value, this matrix doesn't cover the entire passenger flow. This is due to fraud, other types of tickets (e.g. the standard paper ticket) and uncertainties in destination estimates. In this paper we propose a twostep approach for correcting and scaling smart card OD matrix based on adapted Growth Factor model considering the complexity caused by temporal variation of the OD matrix. In the first step we map all the OD pairs in the OD matrix over our area of study to infer their departure and arrival stations and time. In the second step we exploit passengers' counting data and use growth factor model to scale the OD matrix to obtain a new corrected matrix which can present the real flow in the transit network. We apply our proposed methodology to scale an OD matrix constructed only from smart card validation data which presents between 40% to 65% of the overall flow. For this purpose, passengers' counting data are exploited.
Massive data are surrounding us in our daily lives. Urban mobility generates a very high number o... more Massive data are surrounding us in our daily lives. Urban mobility generates a very high number of complex data reflecting the mobility of people, vehicles and objects. Transport operators are primary users who strive to discover the meaning of phenomena behind traffic data, aiming at regulation and transport planning. This paper tackles the question "How to design a supportive tool for visual exploration of digital mobility data to help a transport analyst in decision making?” The objective is to support an analyst to conduct an ex post analysis of train circulation and passenger flows, notably in disrupted situations. We propose a problem-solution process combined with data visualisation. It relies on the observation of operational agents, creativity sessions and the development of user scenarios. The process is illustrated for a case study on one of the commuter line of the Paris metropolitan area. Results encompass three different layers and multiple interlinked views to ex...
2020 17th International Conference on the European Energy Market (EEM), 2020
The need for deploying fast-charging stations for electric vehicles (EVs) is becoming essential i... more The need for deploying fast-charging stations for electric vehicles (EVs) is becoming essential in recent years. This need is justified by the increasing recharging demand, and supported by new recharging technologies making EV chargers more efficient. For this purpose, we introduce an optimization model for finding the optimal deployment of EV fast-charging stations through a network of highways, while taking into account their recharging demands and the restrictions imposed by the electric grid. We also consider the availability of local photovoltaic (PV) power station and integrate its energy to the proposed recharging network. In addition, we present a case study on Paris-Saclay area where the actual mobility flows and electric network specifications are considered. We thus provide the optimal locations for EV charging stations at the studied area, and we indicate the number of chargers to be deployed at each station. Finally, we study the potential benefits of integrating local PV energy and analyze its different prices and production costs.
2020 Forum on Integrated and Sustainable Transportation Systems (FISTS), 2020
Route planning in public transport receives an increasing interest in smart cities and particular... more Route planning in public transport receives an increasing interest in smart cities and particularly in metropolitan cities where crowded and jammed traffic is daily recorded in the transportation network. The availability of digital footprints, such as ticketing logs, or load on board the trains, provides a relevant opportunity to develop innovative decision-making tools for urban routing of passengers in order to assist them to better plan their journeys. In this paper, we propose to enrich existing journey planners with predictive travel information to enhance the passenger travel experience during his journey. For that purpose, we augment the planned trips with predictive passenger flow indicators such as the load on board trains, and passenger attendees at the station. These indicators are forecasted along the journey with the help of the developed machine learning models. The experiments are conducted on a real historical dataset covering the Paris Region with a focus on a railway transit network that serves mainly the suburb of Paris.
2020 23rd Euromicro Conference on Digital System Design (DSD), 2020
Management of a multimodal transport network is a challenging task and operators regularly have t... more Management of a multimodal transport network is a challenging task and operators regularly have to deal with different types of disturbances that affect the quality of service of the transport system. To support them in their decision-making, we present the architecture of a multimodal supervision system that implements monitoring, prediction of affluence, disturbances, impacts and evaluation of functionalities focusing on one or multiples lines of the network. Our system uses the real-time data of transport operators to compute several key performance indicators (KPI) in order to monitor the network status and to detect disturbances. The prediction function is used to predict passengers’ attendance at stations, incident duration and the evolution of the whole network, in particular, phenomena emerging from the interconnections linked to the affected line. Finally, we present a use case of the supervision system applied to a real-time control of a bus line belonging to a very dense and busy segment of the multimodal transport network in Ile-de-France area.
Route planning in public transport receive an increasing interest in smart cities and particularl... more Route planning in public transport receive an increasing interest in smart cities and particularly in metropolitan cities where crowded and jammed traffic is daily recorded in transportation network. The availability of digital footprints such as ticketing logs, or load on board the trains provide a relevant opportunity to develop innovative decision-making tools for urban routing of passengers in order to assist them to better planning their journeys. This planning must consider the forthcoming evolution of the traffic in order to adapt its response to the next state of the network and thus avoid unpleasant situations for the passengers. In this paper, we present a system for individual trip planning that incorporates short and long terms forecasting of different indicators related to the station attendance, occupancy of the trains, and delays in train schedule.
Le travail que nous presentons ici traite de la problematique de l'etiquetage morpho-syntaxiq... more Le travail que nous presentons ici traite de la problematique de l'etiquetage morpho-syntaxique des corpus de specialite non annotes. Les etiqueteurs existants sont entraines sur divers corpus, et produisent des erreurs d'etiquetage dans les textes de specialite. La methode la plus triviale pour apprendre des automates adaptes a un domaine specialise est d'etiqueter manuellement de grands corpus du domaine, ce qui necessite un travail tres important. Pour eviter cela, nous proposons une methode semi-automatique. ETIQ, le nouvel etiqueteur que nous avons developpe, permet de corriger la base de regles obtenue par l‘etiqueteur de BRILL et de l‘adapter a un corpus de specialite. L'expert du domaine visualise l‘etiquetage de base et le corrige par l‘ecriture et l‘application de regles lexicales et contextuelles specialisees. Les regles inserees sont plus expressives que celles de Brill. Pour aider l'expert dans sa tâche, nous avons concu un algorithme inductif biaise...
2021 IEEE International Smart Cities Conference (ISC2), 2021
Public transport route planning is of growing interest in smart cities and especially in metropol... more Public transport route planning is of growing interest in smart cities and especially in metropolitan areas where congestions and traffic jams are frequently recorded. The availability of multiple data sources, such as passenger load in trains or ticketing logs, provides an interesting opportunity to develop decision support tools to help passengers better plan their trips around the city and to enhance their travel experience. We present, in this paper, a multi-criteria journey planner that incorporates train load predictions as criteria. To this end, on the one hand, we enrich the proposed routes with predictive indicators of passenger flow such as the load on board the trains. These indicators are computed for each section of the itinerary using machine learning algorithms. On the other hand, we design a journey planner that incorporates the predicted load in its search criteria.
The need for deploying fast-charging stations for electric vehicles (EVs) is becoming essential i... more The need for deploying fast-charging stations for electric vehicles (EVs) is becoming essential in recent years. This need is justified by the increasing charging demand and supported by new charging technologies making EV chargers more efficient. In this paper, we provide a survey on EV fast-charging models and introduce a data-driven approach with an optimization model for deploying EV fast-chargers for both electric vehicles and heavy trucks traveling through a network of suburban highways. This deployment aims at satisfying EV charging demands while respecting the limits imposed by the electric grid. We also consider the availability of local photovoltaic (PV) farm and integrate its produced power to the proposed charging network. Finally, through a case study on Paris-Saclay area, we provide locations for EV charging stations and analyze the benefits of integrating PV power at different prices, production costs and charging capacities. The obtained results also suggest potentia...
Résumé. Le DÉ Fouille de Textes (DEFT) a consisté à supprimer les phrases non pertinentes dans un... more Résumé. Le DÉ Fouille de Textes (DEFT) a consisté à supprimer les phrases non pertinentes dans un corpus de discours politiques en français. Il a eu lieu en 2005 et réuni onze équipes, totalisant une trentaine de participants. Cet article décrit les prétraitements eectués sur les corpus de F. Mitterrand et de J. Chirac dans le cadre de ce dé. Notamment, la conversion au format texte, le découpage en phrases, le classement des discours, l'introduction de phrases de F. Mitterrand dans les discours de J. Chirac et l'identication des dates et noms de personnes. Les résultats obtenus par les onze équipes participantes sont aussi présentés.
Le DE Fouille de Textes (DEFT) a consiste a supprimer les phrases non pertinentes dans un corpus ... more Le DE Fouille de Textes (DEFT) a consiste a supprimer les phrases non pertinentes dans un corpus de discours politiques en francais. Il a eu lieu en 2005 et reuni onze equipes, totalisant une trentaine de participants. Cet article decrit les pretraitements eectues sur les corpus de F. Mitterrand et de J. Chirac dans le cadre de ce de. Notamment, la conversion au format texte, le decoupage en phrases, le classement des discours, l'introduction de phrases de F. Mitterrand dans les discours de J. Chirac et l'identication des dates et noms de personnes. Les resultats obtenus par les onze equipes participantes sont aussi presentes.
ABSTRACT In this paper, we treat the problem of the grammatical tagging of non-annotated corpora ... more ABSTRACT In this paper, we treat the problem of the grammatical tagging of non-annotated corpora of specialty. The existing taggers are trained on general language corpora, and give inconsistent results on the specialized texts, as technical and scientific ones. In order to learn rules adapted to a specialized field, the usual approach labels manually a large corpus of this field. This is extremely time-consuming. We propose here a semi-automatic approach for tagging corpora of specialty. ETIQ, the new tagger we are building, make it possible to correct the base of rules obtained by Brill’s tagger and to adapt it to a corpus of specialty. The user visualizes an initial and basic tagging and corrects it either by extending Brill’s lexicon or by the insertion of specialized lexical and contextual rules. The inserted rules are richer and more flexible than Brill’s ones. To help the expert in this task, we designed an inductive algorithm biased by the “correct” knowledge he acquired beforehand. By using techniques of machine learning and enabling the expert to incorporate knowledge of the field in an interactive and friendly way, we improve the tagging of specialized corpora. Our approach has been applied to a corpus of molecular biology.
In the context of Part-of-Speech (PoS)-tagging of specialized corpora, we proposed an inductive a... more In the context of Part-of-Speech (PoS)-tagging of specialized corpora, we proposed an inductive approach focusing on the most 'important' PoStags because mistaking them can lead to a total misunderstanding of the text. After a standard tagging of a biological corpus by Brill's tagger, we noted persistent errors that are very hard to deal with. As an application, we studied two cases of different nature: first, confusion between past participle, adjective and preterit for verbs that end with 'ed'; second, confusion between plural nouns and verbs, 3rd person singular present. With a friendly user interface, the expert corrected the examples. Then, from these well-annotated examples, we induced rules using a propositional rule induction algorithm. Experimental validation showed improvement in tagging precision. The relevance of the terminology of the considered field, here molecular biology, is greatly improved when the number of these tagging errors decreases.
2008 3rd International Conference on Information and Communication Technologies: From Theory to Applications, 2008
In this paper, we describe a Text-Mining system used to extract archaeological knowledge from spe... more In this paper, we describe a Text-Mining system used to extract archaeological knowledge from specialized texts. Our approach relies on two basic principals. First, we use userfriendly tools enabling experts to transfer easily their knowledge. Second, our tools include inductive algorithms to reduce the experts' workload.
La compréhension de textes de spécialité nécessite unétiquetage morpho-syntaxique de bonne qualit... more La compréhension de textes de spécialité nécessite unétiquetage morpho-syntaxique de bonne qualité. Or, lorsque les textesétudiés sont issus de domaines spécifiques et peu usités, il est rare de disposer de dictionnaires et autres ressources lexicales fiables. Le logiciel que nous proposons permet d'utiliser unétiquetage réalisé par unétiqueteur généraliste, puis d'améliorer cetétiquetage en intégrant des connaissances d'experts du domaineétudié. Grâce au logiciel développé, il est relativement aisé pour un expert du domaine de détecter des erreurs d'étiquetage et de mettre en place des règles de ré-étiquetage. Ces règles peuventêtre obtenues de deux manières différentes : (1) soit en utilisant un langage de programmation permettant d'exprimer des règles complexes de ré-étiquetage, (2) soit par apprentissage automatique des règlesà partir d'exemples corrigés au moyen d'une interface dédiée. Cet apprentissage propose de nouvelles règlesà l'expert, acquises automatiquement.
The text-mining system we are building deals with the specific problem of identifying the instanc... more The text-mining system we are building deals with the specific problem of identifying the instances of relevant concepts present in the texts. Our system relies therefore on interactions between a field expert and the various linguistic modules we use, often adapted from existing ones, such as Brill's tagger or CMU's Link parser. We have developed learning procedures adapted to various steps of the linguistic treatment, mainly for grammatical tagging, terminology, and concept learning. Our interaction with the expert differs from classical supervised learning, in that the expert is not simply a resource who is only able to provide examples, and unable to provide the formalized knowledge underlying these examples. We are developing specific programming languages which enable the field expert to intervene directly in some of the linguistic tasks. Our approach is thus devoted to help one expert in one field to detect the concepts relevant for his/her field, using a large amount of texts. Our approach is made of two steps. The first one is an automatic approach that finds relevant and novel sentences in the texts. The second one is based on the expert's knowledge and finds more specific relevant sentences. Working on 50 different domains without an expert has been a challenge in itself, and explains our relatively poor results for the first Novelty task.
Uploads
Papers by Ahmed Amrani