Hatem Ghorbel

Sentiment Analysis of French Movie Reviews

In sentiment analysis of reviews we focus on classifying the polarity (positive, negative) of conveyed opinions from the perspective of textual evidence. Most of the work in the field has been intensively applied on the English language and only few experiments have explored other languages. In this paper, we present a supervised classification of French movie reviews where sentiment analysis is based on some shallow linguistic features such as POS tagging, chunking and simple negation forms. In order to improve classification, we extracted word semantic orientation from the lexical resource SentiWordNet. Since SentiWordNet is an English resource, we apply a word-translation from French to English before polarity extraction. Our approach is evaluated using French movie reviews. Obtained results showed that shallow linguistic features has significantly improved the classification performance with respect to the bag of words baseline.

Further Experiments in Sentiment Analysis of French Movie Reviews

In sentiment analysis of reviews we focus on classifying the polarity (positive, negative) of conveyed opinions from the perspective of textual evidence. Most of the work in the field has been intensively applied on the English language and only few experiments have explored other languages. In this paper, we present a supervised classification of French movie reviews where sentiment analysis is based on some shallow linguistic features such as POS tagging, chunking and simple negation forms. In order to improve classification, we extracted word semantic orientation from the lexical resource SentiWordNet. Since SentiWordNet is an English resource, we apply a word-translation from French to English before polarity extraction. Our approach is evaluated on French movie reviews, obtained results showed that shallow linguistic features has significantly improved the classification performance with respect to the bag of words baseline.

Automatic Dialogue Act Annotation within Arabic Debates

Dialogue acts play an important role in the identification of argumentative discourse structure in human conversations. In this paper, we propose an automatic dialogue acts annotation method based on supervised learning techniques for Arabic debates programs. The choice of this kind of corpora is justified by its large content of argumentative information. To experiment annotation results, we used a specific annotation scheme relatively reliable for our task with a kappa agreement of 84%. The annotation process was yield using Weka platform algorithms experimenting Naive Bayes, SVM and Decision Trees classifiers. We obtained encouraging results with an average accuracy of 53%.

Sequential dialogue act recognition for Arabic argumentative debates

Dialogue act recognition remains a primordial task that helps user to automatically identify participants’ intentions. In this paper, we propose a sequential approach consisting of segmentation followed by annotation process to identify dialogue acts within Arabic politic debates.To perform DA recognition, we used the CARD corpus labeled using the SADA annotation schema. Segmentation and annotation tasks were then carried out using Conditional Random Fields probabilistic models as they prove high performance in segmenting and labeling sequential data. Learning results are notably important for the segmentation task (F-score=97.9%) and relatively reliable within the annotation process (fscore= 63.4%) given the complexity of identifying argumentative tags and the presence of disfluencies in spoken conversations.

Focus Definition and Extraction of Opinion Attitude Questions

In Question Answering Systems (QAS), Question Analysis is an important task that consists in general in identifying the semantic type of the question and extracting the question focus. In this context, and as part of a framework aiming to implement an Arabic opinion QAS for political debates, this paper addresses the problem of defining the focus of opinion attitude questions and proposes an approach for extracting it. The proposed approach is based on semi-automatically constructed lexico-syntactic patterns. Evaluation results are considered very encouraging with an average precision of around 87.37%.

Question focus extraction and answer passage retrieval

Question Analysis is an important task in Question Answering Systems (QAS). It consists generally in identifying the semantic type of the question and extracting the main focus of the question. The goal is to better specify the required information by the question. In this context and as part of a framework aiming to implement an Arabic opinion QAS for political debates, this paper addresses the problem of defining the focus of opinion questions and proposes particularly an approach for extracting the focus of attitude questions. The proposed approach is based on semi-automatically constructed lexico-syntactic patterns. Furthermore, the paper presents an adapted Vector Space Model (VSM) based method to retrieve candidate answer passages from a transcribed TV political show. Several experiments were carried out and showed that the focus extraction approach has achieved over 72% as F1 score for holder and target extraction, and has improved the baseline passage retrieval task by over than 25%.

L'alignement des documents médiévaux

Le but de l’alignement des textes est la mise en correspondance des sous-parties similaires de deux ou plusieurs traductions ou versions d’un meme ecrit. La plupart des methodes utilisees dans la technique d’alignement reposent sur l’analyse statistique des frequences de mots ou de caracteres, ou sur la cooccurrence des chaines que ceux-ci constituent. Afin d’en ameliorer l’efficacite, d’autres approches incluent certaines proprietes linguistiques (morpho-syntaxiques et lexico-semantiques) et structurelles (marques de chapitres, de sections, etc.) des documents. Cet article expose les resultats d’une adaptation de la technique d’alignement aux etats paralleles des anciens textes a partir d’une approche multicritere qui tient compte de la similitude au niveau lexical, morpho-syntaxique et lexico-semantique du francais de la periode medievale.

MEDIEVAL: A Navigation Tool Through Medieval Documents

The integration of databases and natural language is becoming an important research field. Most of th database systems are taking advantage of th large progress of research in natural language to improve user interaction. The other way around is also a plausible approach :the use of databases in the services of NLP.This usage must not b restrict d to th static storage of natural language utilities (dictionaries,thesaurus ..),but can be extended to support dynamically the results of linguistic analysis and to improv the information extraction task. In this framework, the aim of MEDIEVAL or “Model d’EDition Informatisee d’Ecrits medievaux, Visualisee par ALignement” is to build a model for medieval text alignment and dev lop a tool for the navigation through these manuscripts.