This research is concerned with the identification of sentiment in multimodal content. This is of particular interest given the increasing presence of subjective multimodal content on the web and other sources, which contains a rich and vast source of people's opinions, feelings, and experiences. Despite the need for tools that can identify opinions in the presence of diverse modalities, most of current methods for sentiment analysis are designed for textual data only, and few attempts have been made to address this problem. The dissertation investigates techniques for augmenting linguistic representations with acoustic, visual, and physiological features. The potential benefits of using these modalities include linguistic disambiguation, visual grounding, and the integration of information about people's internal states. The main goal of this work is to build computational resources and tools that allow sentiment analysis to be applied to multimodal data. This thesis makes three important contributions. First, it shows that modalities such as audio, video, and physiological data can be successfully used to improve existing linguistic representations for sentiment analysis. We present a method that integrates linguistic features with features extracted from these modalities. Features are derived from verbal statements, audiovisual recordings, thermal recordings, and physiological sensors signals. The resulting multimodal sentiment analysis system is shown to significantly outperform the use of language alone. Using this system, we were able to predict the sentiment expressed in video reviews and also the sentiment experienced by viewers while exposed to emotionally loaded content. Second, the thesis provides evidence of the portability of the developed strategies to other affect recognition problems. We provided support for this by studying the deception detection problem. Third, this thesis contributes several multimodal datasets that will enable further research in sentiment and deception detection.

L’hexabromocyclododécane (HBCD) a été utilisé comme additif dans des polystyrènes (PS). En 2013, l’HBCD a été inclus à la Convention de Stockholm comme Polluant Organique Persistant. Les denrées alimentaires d’origine animale (DAOA) constituent la source principale d’exposition de la population générale à ce polluant. Le niveau moyen d’exposition du consommateur ne représente pas un risque majeur pour la sécurité sanitaire des aliments (EFSA, 2011). Malgré des concentrations moyennes généralement proches d’un ng g-1 lipides dans les DAOA, des concentrations maximales supérieures à 2000 ng g-1 lipides dans certains échantillons d’œuf ont été rapportées. Plusieurs sources potentielles d’exposition orale des volailles existent. Les poules pondeuses pourraient être exposées majoritairement à l’isomère α-HBCD. Au vu de l’absence d’informations dans la littérature sur le devenir de l’α-HBCD chez la poule pondeuse, une expérimentation in vivo a été réalisée et a permis d’estimer la plupart des paramètres cinétiques du devenir de l’α-HBCD ingéré chez la poule pondeuse. Un modèle pharmacocinétique basé sur la physiologie (PBPK) de transfert de l’α-HBCD chez la poule pondeuse a été développé, calibré par des données de la littérature, soumis à analyse de sensibilité et plusieurs paramètres ont été ajustés sur des résultats in vivo. Une validation du modèle a été testée. Ce modèle a permis de déterminer l’influence des caractéristiques physiologiques de la poule pondeuse sur le devenir de l’α-HBCD en cas d’exposition à la molécule. Parmi les caractéristiques d’intérêt, le poids vif (PV) est celle qui influence le plus la concentration d’α-HBCD dans les tissus et les œufs. Un PV bas et des performances de ponte réduites prédisposent à une concentration plus élevée dans ces matrices. La mue produit une augmentation de la charge corporelle de la molécule. Ce modèle peut être extrapolé à d’autres animaux et molécules lipophiles après modification de certains paramètres / Hexabromocyclododecane (HBCD) has been used as an additive in polystyrene (PS). In 2013, HBCD was included in the Stockholm Convention as a Persistent Organic Pollutant. The foodstuffs of animal origin are the main source of exposure of the general population to this pollutant. The mean exposure level of the consumers to HBCD doesn’t represent a major risk to the food safety (EFSA, 2011). In spite of mean concentrations generally close to 1 ng g-1 lipids in the foodstuffs of animal origin, maximal concentrations higher than 2000 ng g-1 lipids have been reported in some egg samples. There are several potential sources of oral exposure of the hens to this pollutant. The hens could be exposed mainly to the α-HBCD isomer. Since the bibliographic synthesis highlights the absence of information about the α-HBCD kinetics in the laying hen, an in vivo experience has been realized and has allowed the estimation of the majority of the α-HBCD kinetic parameters in the laying hen. A physiologically based pharmacokinetic (PBPK) model has been developed, calibrated by literature data, submit to a sensitivity analysis and some parameters have been adjusted on the basis of experimental data. A validation of the model has been attempted. This model has allowed to determine the influence of the main physiological characteristics of the laying hen exposed to the molecule on the α-HBCD accumulation in tissues and transfer to eggs. Among the physiological characteristics of concern, the body weight (BW) is the main feature influencing the α-HBCD kinetics. If the hen is exposed to the α-HBCD, a low BW and low laying performances predispose the hen to have higher concentrations of the molecule in all the matrices. The molting stage raises the body charge of the molecule. This model can be extrapolated to other animals and lipophilic molecules after modification of some parameters