Clinical named entity recognition is an essential task for humans to analyze large-scale electron... more Clinical named entity recognition is an essential task for humans to analyze large-scale electronic medical records efficiently. Traditional rule-based solutions need considerable human effort to build rules and dictionaries; machine learning-based solutions need laborious feature engineering. For the moment, deep learning solutions like Long Short-term Memory with Conditional Random Field (LSTM–CRF) achieved considerable performance in many datasets. In this paper, we developed a multitask attention-based bidirectional LSTM–CRF (Att-biLSTM–CRF) model with pretrained Embeddings from Language Models (ELMo) in order to achieve better performance. In the multitask system, an additional task named entity discovery was designed to enhance the model’s perception of unknown entities. Experiments were conducted on the 2010 Informatics for Integrating Biology & the Bedside/Veterans Affairs (I2B2/VA) dataset. Experimental results show that our model outperforms the state-of-the-art solution b...
2018 14th International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery (ICNC-FSKD), 2018
Through the enormous impact of social media on our daily life, an important information database ... more Through the enormous impact of social media on our daily life, an important information database can be obtained from theses medias. But we have a lack of information about those who create the contents. The aim of author profiling is to analyze authors published texts in order to determine theirs profile category. To solve and optimize the multi-class categorization problem we propose in this paper a profile categorization approach. Our approach divided into processing module and the classifying module. Firstly we are going to adopt TF-IDF with threshold filtering method during the feature extraction step. Then we incorporate the fuzzy set theory into Twin SVM (TSVM), then OAA-TSVM (One-Against-All TSVM) OAO-TSVM(One-Against-One TSVM) are used in the classifying module. Our proposed classifier are tested using the RepLab 2014 Data set and performance measures: precision, recall and F-measure are used for the evaluation. Result obtained show that our proposed profile categorization approach perform very well.
Nowadays, social medias are very popular among their users. One of the most well-known social net... more Nowadays, social medias are very popular among their users. One of the most well-known social networks is Twitter. It is a micro-blog that enables its users to send short messages called tweets. A tweet is a 140 characters long message that is rarely self-cont, hence additional information are necessary to allow better readability of the tweet. This new task has attracted a great deal of attention recently. Given a tweet, the aim of tweet contextualization is to produce an informative and coherent paragraph, called a context, from a set of documents in response to topics treated by the tweet. In this paper, we propose a new approach of Tweet Contextualization based on combining automatic summarization techniques and sentence aggregation. The main idea of our proposed method is to select relevant, informative and semantically related sentences that best describe themes expressed by the tweet, and then build a concise context.
The need for efficient content-based image retrieval has increased hugely. Two meth- ods are reco... more The need for efficient content-based image retrieval has increased hugely. Two meth- ods are recognized for describing the content of images: using global features and using local features. In this paper, we propose two methods for image retrieving based on visual similarity. The first one characterizes images by global features, when the second is based on local fea- tures. In the global descriptor attributes are computed on the whole image, whereas in the local descriptor attributes are computed on regions of the image. The aim of this paper is to compare global features versus local features for Web images retrieval. RESUME. On reconnait actuellement, dans les systemes de recherche d'image par contenu, deux methodes pour la description du contenu des images : a travers des attributs locaux ou a travers des attributs globaux. Dans ce papier, nous proposons deux methodes pour la recherche d'image qui sont basees sur la similitude visuelle. La premiere caracterise les images...
Communications in Computer and Information Science, 2020
The popularity and the great success of social networks are due to their ability to offer Interne... more The popularity and the great success of social networks are due to their ability to offer Internet users a free space for expression where they can produce a large amount of information. Thus the new challenges of information research and data mining are to extract and analyze this mass of information which can then be used in different applications. This information is characterized mainly by incompleteness, imprecision, and heterogeneity. Indeed the task of analysis using models based on statistics and word frequencies is crucial. To solve the problem of uncertainty, the possibility theory turns out to be the most adequate. In this article, we propose a new approach to find relevant short texts such as tweets using the dual possibility and necessity. Our goal is to translate the fact that a tweet can only be relevant if there is not only a semantic relationship between the tweet and the query but also a synergy between the terms of the tweet. We have modeled the problem through a possibility network to measure the possibility of the relevance of terms in relation to a concept of a given query and a necessity network to measure the representativeness of terms in a tweet. The evaluation shows that using the theory of possibilities with a set of concepts relevant to an initial query gives the best precision rate compared to other approaches.
International Journal of Information Retrieval Research, 2017
This paper proposes a new relevance feedback approach to collaborative information retrieval base... more This paper proposes a new relevance feedback approach to collaborative information retrieval based on a confidence's network, which performs propagation relevance between annotations terms. The main contribution of our approach is to extract relevant terms to reformulate the initial user query considering the annotations as an information source. The proposed model introduces the concept of necessity that allows determining the terms that have strong association relationships. The authors estimated the association relationship to a measure of a confidence. Another contribution consists on determining the relevant annotations for a given evidence source. Since the user is over whelmed by a variety of contradictory annotations on even one which are far from the original subject, the authors' model proceed filtering these annotations to determine the relevant one and then it classify them by grouping those related semantically. The experimental study conducted on different quer...
Proceedings of the 2006 ACM symposium on Applied computing - SAC '06, 2006
The aim of this study is to evaluate the impact of context to better identify relevant elements i... more The aim of this study is to evaluate the impact of context to better identify relevant elements in XML retrieval. Context is represented here by clues on whole document relevance. We represent context according to different points of view: by introducing document dimension while computing terms weights, by using document relevance when evaluating elements relevance or by ranking elements on
Relevance Feedback (RF) is a technique that allows to enrich an initial query according to the us... more Relevance Feedback (RF) is a technique that allows to enrich an initial query according to the user feedback. The goal is to express more precisely the user's needs. Some open issues arise when considering semi-structured documents like XML documents. They are mainly related to the form of XML documents which mix content and structure information and to the new granularity
Relevance Feedback (RF) is a technique allowing to enrich an initial query according to the user ... more Relevance Feedback (RF) is a technique allowing to enrich an initial query according to the user feedback. The goal is to express more precisily the user’s needs. Some open issues appear when considering semi-structured documents like XML documents. Most of the RF approaches proposed in XML retrieval are simple adaptations of traditional RF to the new granularity of information. They enrich queries by adding terms extracted from relevant elements instead of terms extracted from whole documents. In this paper we show how structural constraints can also be used in RF. We propose a new approach that is able to extend the initial query by adding one or more generative structures. This approach is applied to unstructured queries. Experiments are carried out on INEX collection and results show the interest of our method. Categories and Subject Descriptors H.3.3 [Information storage and retrieval]: Information search and retrieval—retrieval models; H.3.4 [Information storage and retrieval]...
ABSTRACT. The need for efficient content-based image retrieval has increased hugely. Two methods ... more ABSTRACT. The need for efficient content-based image retrieval has increased hugely. Two methods are recognized for describing the content of images: using global features and using local features. In this paper, we propose two methods for image retrieving based on visual similarity. The first one characterizes images by global features, when the second is based on local features. In the global descriptor attributes are computed on the whole image, whereas in the local descriptor attributes are computed on regions of the image. The aim of this paper is to compare global features versus local features for Web images retrieval. RÉSUMÉ. On reconnait actuellement, dans les systèmes de recherche d’image par contenu, deux méthodes pour la description du contenu des images: à travers des attributs locaux ou à travers des attributs globaux. Dans ce papier, nous proposons deux méthodes pour la recherche d’image qui sont basées sur la similitude visuelle. La première caractérise les images pa...
2020 15th International Workshop on Semantic and Social Media Adaptation and Personalization (SMA, 2020
A Social Information Retrieval (SIR) result re-ranking system offers several techniques and metho... more A Social Information Retrieval (SIR) result re-ranking system offers several techniques and methods that meet the need of the user to express through a given query and return a re-ranked list of returned results. This list should contains the information relevant to the top ranking since the majority of these users are only interested in the first results displayed. However, optimizing the returned results ranking becomes a fundamental problem when searching for tweets. In this regard, this paper presents a re-ranking model of social retrieval results based on automatic learning. Indeed, this model can re-rank the list of results obtained by the SIR system based on adaptive features given query category. Also, our model devises queries into categories and learn the representative features of each. More precisely, in order to reorder the result obtained from a given query, our model allows recalculating a new relevance score for each corresponding tweet. To validate this proposed approach, the experiments were carried out based on queries from different categories. The results obtained from his experiments show the effectiveness of our model compared to classic ranking approaches.
Relevance Feedback (RF) is a technique allowing to enrich an initial query according to the user ... more Relevance Feedback (RF) is a technique allowing to enrich an initial query according to the user feedback. The goal is to express more precisily the user’s needs. Some open issues appear when considering semi-structured documents like XML. Most of the existing RF approaches are applied in the content of elements. We prpose a new approach that is able to extend the initial query by adding a generative structure and keywords. This approach is applied to both structured and un-structured queries. Experiments are carried out with INEX campaign and results show the interest of our method for some query types. MOTS-CLÉS :reformulation, document XML, structure générique, contenu, RI.
As opposed to traditional Information Retrieval (IR) which views whole documents as atomic units ... more As opposed to traditional Information Retrieval (IR) which views whole documents as atomic units of retrieval, XML IR processes XML elements as possible units of retrieval. Many open issues appear when considering Relevance Feedback (RF) in XML documents. They are mainly related to the form of XML documents that mix content and structure and to the new granularity of information processed by the Information Retrieval Systems (IRS). Most of the RF approaches proposed in XML retrieval are simple adaptations of traditional RF to the new granularity of information. They enrich queries by adding terms extracted from relevant elements instead of terms extracted from whole documents. In this paper, we propose to extend the initial query by adding both content and structural constraints. Experiments are carried out with the INEX evaluation campaign and results show the interest of our method.
Though the enormous impact of social media on our daily life, we observe a lack of information ab... more Though the enormous impact of social media on our daily life, we observe a lack of information about those who create the contents. In this regard, author profiling tries to determine the profile category of authors by analysing their published texts. This paper presents a profile categorization system to solve the multi-class categorization problem. The system consists of two modules: the processing module and the classifying module. In the first module TF-IDF with threshold filtering method are used to extract the relevant terms. While support vector machine (SVM) is used in the classifying module. Our proposed feature reduction method is tested using the RepLab 2014 Data set and evaluated using performance measures: precision, recall and F-measure. Result obtained show that proposed method improves system performance when solving a multi-class profile categorization problem.
Collaborative retrieval allows increasing the amount of relevant information found and sharing hi... more Collaborative retrieval allows increasing the amount of relevant information found and sharing history with others. The collaborative retrieval can reduce the retrieval time performed by the users of the same profile. This chapter proposes a new relevance feedback algorithm to collaborative information retrieval based on a confidence network, which performs propagation relevance between annotations terms. The main contribution in this work is the extraction of relevant terms to reformulate the initial user query considering the annotations as an information source. The proposed model introduces the concept of necessity that allows determining the terms that have strong association relationships estimated to the measure of a confidence. Since the user is overwhelmed by a variety of contradictory annotations, another contribution consists of determining the relevant annotations for a given evidence source. The experimental study gives very encouraging results. Uncertain Confidence Net...
Nowadays, social media like Twitter, Facebook, blogs, and LinkedIn are considered the most used s... more Nowadays, social media like Twitter, Facebook, blogs, and LinkedIn are considered the most used sources of information, while at the same time being the most visited and most used sources of disinformation. These can have a negative impact on several areas and on our minds, hence on our behavior. It is obvious that this disinformation is closely related to the profiles of the authors of this information. The purpose of author profiling is to analyze the texts published by the authors in order to determine their profile category. A wide range of methods for selecting statistical characteristics and machine learning has been studied in recent years in order to automatically classify this information. However, these main methods of selecting statistical characteristics and machine learning used for this purpose have not proven their great performance in the processing of data from social networks. The main contribution of this article consists in integrating the semantic component, which has not been taken into account in the main approaches studied in the literature, as additional functionalities enabling the identification of relevant information. Our hypothesis is that the concepts and the relationships between these concepts tend to have a more coherent correlation with relevant and irrelevant information, and can therefore increase the discriminating power of classifiers. The semantic approach proposed revolves around an ontology combined with the linear SVM classifier and then with the fuzzy SVM classifier. The experimental study carried out, on the different collections of Twitter profiles. On our approach and on the main approaches to the literature that we have studied, as well as the analysis of the results obtained. The results we have clearly show the limits of these studied approaches and confirm the performance of our approach, as well as the efficiency of the integration of the semantic component in the categorization of Twitter profiles.
Clinical named entity recognition is an essential task for humans to analyze large-scale electron... more Clinical named entity recognition is an essential task for humans to analyze large-scale electronic medical records efficiently. Traditional rule-based solutions need considerable human effort to build rules and dictionaries; machine learning-based solutions need laborious feature engineering. For the moment, deep learning solutions like Long Short-term Memory with Conditional Random Field (LSTM–CRF) achieved considerable performance in many datasets. In this paper, we developed a multitask attention-based bidirectional LSTM–CRF (Att-biLSTM–CRF) model with pretrained Embeddings from Language Models (ELMo) in order to achieve better performance. In the multitask system, an additional task named entity discovery was designed to enhance the model’s perception of unknown entities. Experiments were conducted on the 2010 Informatics for Integrating Biology & the Bedside/Veterans Affairs (I2B2/VA) dataset. Experimental results show that our model outperforms the state-of-the-art solution b...
2018 14th International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery (ICNC-FSKD), 2018
Through the enormous impact of social media on our daily life, an important information database ... more Through the enormous impact of social media on our daily life, an important information database can be obtained from theses medias. But we have a lack of information about those who create the contents. The aim of author profiling is to analyze authors published texts in order to determine theirs profile category. To solve and optimize the multi-class categorization problem we propose in this paper a profile categorization approach. Our approach divided into processing module and the classifying module. Firstly we are going to adopt TF-IDF with threshold filtering method during the feature extraction step. Then we incorporate the fuzzy set theory into Twin SVM (TSVM), then OAA-TSVM (One-Against-All TSVM) OAO-TSVM(One-Against-One TSVM) are used in the classifying module. Our proposed classifier are tested using the RepLab 2014 Data set and performance measures: precision, recall and F-measure are used for the evaluation. Result obtained show that our proposed profile categorization approach perform very well.
Nowadays, social medias are very popular among their users. One of the most well-known social net... more Nowadays, social medias are very popular among their users. One of the most well-known social networks is Twitter. It is a micro-blog that enables its users to send short messages called tweets. A tweet is a 140 characters long message that is rarely self-cont, hence additional information are necessary to allow better readability of the tweet. This new task has attracted a great deal of attention recently. Given a tweet, the aim of tweet contextualization is to produce an informative and coherent paragraph, called a context, from a set of documents in response to topics treated by the tweet. In this paper, we propose a new approach of Tweet Contextualization based on combining automatic summarization techniques and sentence aggregation. The main idea of our proposed method is to select relevant, informative and semantically related sentences that best describe themes expressed by the tweet, and then build a concise context.
The need for efficient content-based image retrieval has increased hugely. Two meth- ods are reco... more The need for efficient content-based image retrieval has increased hugely. Two meth- ods are recognized for describing the content of images: using global features and using local features. In this paper, we propose two methods for image retrieving based on visual similarity. The first one characterizes images by global features, when the second is based on local fea- tures. In the global descriptor attributes are computed on the whole image, whereas in the local descriptor attributes are computed on regions of the image. The aim of this paper is to compare global features versus local features for Web images retrieval. RESUME. On reconnait actuellement, dans les systemes de recherche d'image par contenu, deux methodes pour la description du contenu des images : a travers des attributs locaux ou a travers des attributs globaux. Dans ce papier, nous proposons deux methodes pour la recherche d'image qui sont basees sur la similitude visuelle. La premiere caracterise les images...
Communications in Computer and Information Science, 2020
The popularity and the great success of social networks are due to their ability to offer Interne... more The popularity and the great success of social networks are due to their ability to offer Internet users a free space for expression where they can produce a large amount of information. Thus the new challenges of information research and data mining are to extract and analyze this mass of information which can then be used in different applications. This information is characterized mainly by incompleteness, imprecision, and heterogeneity. Indeed the task of analysis using models based on statistics and word frequencies is crucial. To solve the problem of uncertainty, the possibility theory turns out to be the most adequate. In this article, we propose a new approach to find relevant short texts such as tweets using the dual possibility and necessity. Our goal is to translate the fact that a tweet can only be relevant if there is not only a semantic relationship between the tweet and the query but also a synergy between the terms of the tweet. We have modeled the problem through a possibility network to measure the possibility of the relevance of terms in relation to a concept of a given query and a necessity network to measure the representativeness of terms in a tweet. The evaluation shows that using the theory of possibilities with a set of concepts relevant to an initial query gives the best precision rate compared to other approaches.
International Journal of Information Retrieval Research, 2017
This paper proposes a new relevance feedback approach to collaborative information retrieval base... more This paper proposes a new relevance feedback approach to collaborative information retrieval based on a confidence's network, which performs propagation relevance between annotations terms. The main contribution of our approach is to extract relevant terms to reformulate the initial user query considering the annotations as an information source. The proposed model introduces the concept of necessity that allows determining the terms that have strong association relationships. The authors estimated the association relationship to a measure of a confidence. Another contribution consists on determining the relevant annotations for a given evidence source. Since the user is over whelmed by a variety of contradictory annotations on even one which are far from the original subject, the authors' model proceed filtering these annotations to determine the relevant one and then it classify them by grouping those related semantically. The experimental study conducted on different quer...
Proceedings of the 2006 ACM symposium on Applied computing - SAC '06, 2006
The aim of this study is to evaluate the impact of context to better identify relevant elements i... more The aim of this study is to evaluate the impact of context to better identify relevant elements in XML retrieval. Context is represented here by clues on whole document relevance. We represent context according to different points of view: by introducing document dimension while computing terms weights, by using document relevance when evaluating elements relevance or by ranking elements on
Relevance Feedback (RF) is a technique that allows to enrich an initial query according to the us... more Relevance Feedback (RF) is a technique that allows to enrich an initial query according to the user feedback. The goal is to express more precisely the user's needs. Some open issues arise when considering semi-structured documents like XML documents. They are mainly related to the form of XML documents which mix content and structure information and to the new granularity
Relevance Feedback (RF) is a technique allowing to enrich an initial query according to the user ... more Relevance Feedback (RF) is a technique allowing to enrich an initial query according to the user feedback. The goal is to express more precisily the user’s needs. Some open issues appear when considering semi-structured documents like XML documents. Most of the RF approaches proposed in XML retrieval are simple adaptations of traditional RF to the new granularity of information. They enrich queries by adding terms extracted from relevant elements instead of terms extracted from whole documents. In this paper we show how structural constraints can also be used in RF. We propose a new approach that is able to extend the initial query by adding one or more generative structures. This approach is applied to unstructured queries. Experiments are carried out on INEX collection and results show the interest of our method. Categories and Subject Descriptors H.3.3 [Information storage and retrieval]: Information search and retrieval—retrieval models; H.3.4 [Information storage and retrieval]...
ABSTRACT. The need for efficient content-based image retrieval has increased hugely. Two methods ... more ABSTRACT. The need for efficient content-based image retrieval has increased hugely. Two methods are recognized for describing the content of images: using global features and using local features. In this paper, we propose two methods for image retrieving based on visual similarity. The first one characterizes images by global features, when the second is based on local features. In the global descriptor attributes are computed on the whole image, whereas in the local descriptor attributes are computed on regions of the image. The aim of this paper is to compare global features versus local features for Web images retrieval. RÉSUMÉ. On reconnait actuellement, dans les systèmes de recherche d’image par contenu, deux méthodes pour la description du contenu des images: à travers des attributs locaux ou à travers des attributs globaux. Dans ce papier, nous proposons deux méthodes pour la recherche d’image qui sont basées sur la similitude visuelle. La première caractérise les images pa...
2020 15th International Workshop on Semantic and Social Media Adaptation and Personalization (SMA, 2020
A Social Information Retrieval (SIR) result re-ranking system offers several techniques and metho... more A Social Information Retrieval (SIR) result re-ranking system offers several techniques and methods that meet the need of the user to express through a given query and return a re-ranked list of returned results. This list should contains the information relevant to the top ranking since the majority of these users are only interested in the first results displayed. However, optimizing the returned results ranking becomes a fundamental problem when searching for tweets. In this regard, this paper presents a re-ranking model of social retrieval results based on automatic learning. Indeed, this model can re-rank the list of results obtained by the SIR system based on adaptive features given query category. Also, our model devises queries into categories and learn the representative features of each. More precisely, in order to reorder the result obtained from a given query, our model allows recalculating a new relevance score for each corresponding tweet. To validate this proposed approach, the experiments were carried out based on queries from different categories. The results obtained from his experiments show the effectiveness of our model compared to classic ranking approaches.
Relevance Feedback (RF) is a technique allowing to enrich an initial query according to the user ... more Relevance Feedback (RF) is a technique allowing to enrich an initial query according to the user feedback. The goal is to express more precisily the user’s needs. Some open issues appear when considering semi-structured documents like XML. Most of the existing RF approaches are applied in the content of elements. We prpose a new approach that is able to extend the initial query by adding a generative structure and keywords. This approach is applied to both structured and un-structured queries. Experiments are carried out with INEX campaign and results show the interest of our method for some query types. MOTS-CLÉS :reformulation, document XML, structure générique, contenu, RI.
As opposed to traditional Information Retrieval (IR) which views whole documents as atomic units ... more As opposed to traditional Information Retrieval (IR) which views whole documents as atomic units of retrieval, XML IR processes XML elements as possible units of retrieval. Many open issues appear when considering Relevance Feedback (RF) in XML documents. They are mainly related to the form of XML documents that mix content and structure and to the new granularity of information processed by the Information Retrieval Systems (IRS). Most of the RF approaches proposed in XML retrieval are simple adaptations of traditional RF to the new granularity of information. They enrich queries by adding terms extracted from relevant elements instead of terms extracted from whole documents. In this paper, we propose to extend the initial query by adding both content and structural constraints. Experiments are carried out with the INEX evaluation campaign and results show the interest of our method.
Though the enormous impact of social media on our daily life, we observe a lack of information ab... more Though the enormous impact of social media on our daily life, we observe a lack of information about those who create the contents. In this regard, author profiling tries to determine the profile category of authors by analysing their published texts. This paper presents a profile categorization system to solve the multi-class categorization problem. The system consists of two modules: the processing module and the classifying module. In the first module TF-IDF with threshold filtering method are used to extract the relevant terms. While support vector machine (SVM) is used in the classifying module. Our proposed feature reduction method is tested using the RepLab 2014 Data set and evaluated using performance measures: precision, recall and F-measure. Result obtained show that proposed method improves system performance when solving a multi-class profile categorization problem.
Collaborative retrieval allows increasing the amount of relevant information found and sharing hi... more Collaborative retrieval allows increasing the amount of relevant information found and sharing history with others. The collaborative retrieval can reduce the retrieval time performed by the users of the same profile. This chapter proposes a new relevance feedback algorithm to collaborative information retrieval based on a confidence network, which performs propagation relevance between annotations terms. The main contribution in this work is the extraction of relevant terms to reformulate the initial user query considering the annotations as an information source. The proposed model introduces the concept of necessity that allows determining the terms that have strong association relationships estimated to the measure of a confidence. Since the user is overwhelmed by a variety of contradictory annotations, another contribution consists of determining the relevant annotations for a given evidence source. The experimental study gives very encouraging results. Uncertain Confidence Net...
Nowadays, social media like Twitter, Facebook, blogs, and LinkedIn are considered the most used s... more Nowadays, social media like Twitter, Facebook, blogs, and LinkedIn are considered the most used sources of information, while at the same time being the most visited and most used sources of disinformation. These can have a negative impact on several areas and on our minds, hence on our behavior. It is obvious that this disinformation is closely related to the profiles of the authors of this information. The purpose of author profiling is to analyze the texts published by the authors in order to determine their profile category. A wide range of methods for selecting statistical characteristics and machine learning has been studied in recent years in order to automatically classify this information. However, these main methods of selecting statistical characteristics and machine learning used for this purpose have not proven their great performance in the processing of data from social networks. The main contribution of this article consists in integrating the semantic component, which has not been taken into account in the main approaches studied in the literature, as additional functionalities enabling the identification of relevant information. Our hypothesis is that the concepts and the relationships between these concepts tend to have a more coherent correlation with relevant and irrelevant information, and can therefore increase the discriminating power of classifiers. The semantic approach proposed revolves around an ontology combined with the linear SVM classifier and then with the fuzzy SVM classifier. The experimental study carried out, on the different collections of Twitter profiles. On our approach and on the main approaches to the literature that we have studied, as well as the analysis of the results obtained. The results we have clearly show the limits of these studied approaches and confirm the performance of our approach, as well as the efficiency of the integration of the semantic component in the categorization of Twitter profiles.
Uploads