Current models of long-term memory and semantic memory assume that abstract representations (concepts, schemas, knowledge) are stored and developed from singular experiments according to an "abstract" principle of generalization of information that can produce semantic spaces from large corpus texts. We show how Latent Semantic Analysis (LSA) simulates verbal associations, lexical meanings, recall and summary of narratives, Judgment of Relative Importance and vocabulary acquisition in younger children.
A computational model of the construction of word meaning through exposure to texts is built in order to simulate the effects of co-occurrence values on word semantic similarities, paragraph by paragraph. Semantic similarity is here viewed as association. It turns out that the similarity between two words W1 and W2 strongly increases with a co-occurrence, decreases with the occurrence of W1 without W2 or W2 without W1, and slightly increases with high-order cooccurrences. Therefore, operationalizing similarity as a frequency of co-occurrence probably introduces a bias: first, there are cases in which there is similarity without co-occurrence and, second, the frequency of co-occurrence overestimates similarity. Nous présentons un modèle informatique de la construction de la signification des mots par l'exposition aux textes, dans le but de simuler, paragraphe après paragraphe, les effets des valeurs de cooccurrence sur les similarités sémantiques intermots. La similarité est ici considérée comme une association sémantique. Les résultats montrent que la similarité entre deux mots M1 et M2 augmente fortement avec leur cooccurrence, diminue avec l'occurrence de M1 sans M2 ou de M2 sans M1, et augmente légèrement avec des cooccurrences d'ordre supérieur. Opérationnaliser la similarité par la fréquence de cooccurrence introduit donc probablement un Effects of High-Order Co-occurrences on Word Semantic Similarity Current psychology letters, 18, Vol. 1, 2006 | 2006 11
Les modèles actuels de la mémoire à long terme et de la mémoire sémantique supposent que des représentations abstraites (concept, schémas, connaissances) sont stockées et élaborées à partir d’expériences singulières selon un principe « abstractif » de généralisation des informations pouvant produire des espaces sémantiques à partir de larges corpus de textes. Nous montrons comment l’Analyse de la Sémantique Latente (LSA) permet de modéliser et simuler chez les enfants : les associations verbales, les significations lexicales, le rappel et le résumé de récits, les Jugements d’Importance Relative et l’acquisition du vocabulaire.
User-centered design not only requires designers to analyse and anticipate how users are likely to use a Web application, but also to validate their assumptions with regard to user behaviour in real environments. Cognitive neuroscience, for its part, addresses the questions of how psychological functions are produced by neural circuitry. The emergence of powerful new measurement techniques allows neuroscientists and psychologists to address abstract questions such as how human cognition and emotion are mapped to specific neural substrates. This paper focus on the validation of user-centered designs and requirements of Web applications by neuroscience techniques and suggest the use of these techniques to achieve efficient and effectiveness validated designs by real behavior of potential users.
Controversy still persists on whether emotional valence and arousal influence cognitive activities. Our study sought to compare how these two factors foster the spread of activation within the semantic network. In a lexical decision task, prime words were varied depending on the valence (pleasant or unpleasant) or on the level of emotional arousal (high or low). Target words were carefully selected to avoid semantic priming effects, as well as to avoid arousing specific emotions (neutral). Three SOA durations (220, 420 and 720 ms) were applied across three independent groups. Results indicate that at 220 ms, the effect of arousal is significantly higher than the effect of valence in facilitating spreading activation while at 420 ms, the effect of valence is significantly higher than the effect of arousal in facilitating spreading activation. These findings suggest that affect is a sequential process involving the successive intervention of arousal and valence.
In this article, we present a set of 12 norms that characterize emotional terms in French, English, German, Spanish, Italian, and Finnish. The high correlation between the norm values in the two emotional dimensions of valence and arousal suggests an interlingual homogeneity of emotional representations and allows a significant metanorm-EMONORM-to be established with 6,383 terms characterized in valence and 4,345 terms characterized in arousal. This metanorm is a resource for creating experimental materials in studies on language and emotions. Furthermore, we perform three tests using EMONORM, with the objectives of (1) identifying basic emotions from their valence and arousal values, (2) determining the orientation of texts referring to positive and negative emotions, and (3) evaluating the intensity of emotions expressed in texts. The results are highly similar to those for human judgments. Finally, we present EMOVAL/SEMOTEX, a Web application for static and dynamic valence and arousal emotional analysis of texts using EMONORM ( http://www.semotex.fr ).
EMOVAL: automatic evaluation the emotional valence and arousal of texts using a 5656 root-words metanorm. EMOVAL is an emotional valence and arousal analysis model of texts. EMOVAL draws from linguistic tradition the hypothesis that every word has a denotative aspect ("meaning") and a connotative aspect ("affective halo"). It uses a meta-analysis of seven French norms and one English norm with the objective to characterize the emotional valence of texts, paragraphs, or sentences in a pleasant or unpleasant way. The meta-analysis indicates that the seven French norms data are highly correlated in between (0.82 to 0.99), and highly correlated with the Affectiv Norm for English Words (Bradley et Lang, 1999) (0.81 to 0.97). Arousal values taken from Affective Norm for English Words (ANEW) (Bradley et Lang, 1999) and the Leleu (1987) norm are significately correlated (0.55). The metanorm has 5656 words (nouns, verbs, adjectives, adverbs) characterized in valence (-1 to +1), and 3265 words characterized in arousal. These items are used by EMOVAL for valence judgments of texts. Two types of texts are proposed: the evaluation of the whole (702) or of extracts (110) of a corpus of sentences judged in a seven-point scale (-3 very unpleasant to +3 very pleasant) (Bestgen et al., 2004), and of 12 texts positively valenced (happiness and good surprise) and negatively valenced (fear, anger, disgust, sadness, and bad surprise). These two types of tests confirm the psychological pertinence of EMOVAL. Limits regarding the arousal dimension are discussed. The metanorm presented in this article can be obtained from the authors. (C) 2011 Societe francaise de psychologie. Publie par Elsevier Masson SAS. Tous droits reserves.
Emotional lexicon organization in memory Nicolas Leveau CHArt - LUTIN - UMS CNRS 2809 Sandra Jhean-Larose Universit´e d’Orl´eans CHArt - LUTIN - UMS CNRS 2809 Guy Denhi` ere CHArt - LUTIN - UMS CNRS 2809 Abstract: In this article, we study the organization of emotions in memory supported by strategic processes. Three hypotheses are tested. First, if emotions are structured in a n-dimension space, to which emotional characteristics do these dimensions return. Second, is it possible to give an account of basic emotions through their organization in memory. Third, can semantic relations give an account of emotional relations. Two word sorting experiments are performed. In the first experiment, 128 pleasant and unpleasant words are divided into 8 lists. In the second experiment, one positive and one negative 128 words list are each divided into 8 groups of 16 words. The results support the hypothesis of an organization around a core affect (Russell, 2003), reinforce the existence of discrete emotions (Johnson-Laird & Oatley, 1989), and confirm the predictive function of semantic relations accounted by the Latent Semantic Analysis (Landauer & Dumais, 1997) on the organization of emotions.
This study looks at how combinations of two French nouns are interpreted. The order of occurrence of the constituents of two types of conceptual combinations, relation and property, was manipulated in view of determining how property-based and relation-based interpretations evolve with age. Three groups of French-speaking children (ages 6, 8, and 10) and a group of adults performed an interpretation-selection task. The results for the children indicated that while property-based interpretations increased with age, relation-based interpretations were in the majority for both combination types, whereas for the adults, relation-based interpretations were in the minority for property combinations. For the children and adults alike, the most frequent interpretations were ones in which the head noun came first and was followed by the modifier (the opposite of the order observed for English).
In this paper, we present an application using the SUMMA-LSA platform developed by Baier, Lehnard, Hoffmann & Schneider (this volume). SUMMA-LSA was used to evaluate biology knowledge of 7th and 8th grades students dealing with The human body energy requirements . Student knowledge has been measured by means of classical and “evidential” multiple choices questions (MCQs) as well as open questions. The highly significant correlations between student answers and LSA cosine values opened encouraging perspectives in the development of e-learning systems that allows self-evaluations in real time and adaptation to learners’ knowledge.
Latent semantic analysis is a computation method to demonstrate a major component of language learning and use. Thus, in this sense, it is a theory of meaning, such that it applies to and offers an explanation of phenomena of meaning in words and passages of words. This enables LSA to hold a strong position in the automated document classification, document analysis, etc. Though the experiments show that LSA can reach a very high accuracy in document classification, it also depends on the various factors such as quality and amount of training documents, characteristics of representative vector and composition of the to be classified documents, etc. On the other hand, pretopology is showing its strength in the fields of data classification and modeling. Besides, some applications, which are to strengthen the pretopology with visualization in the domain of classification, have shown promising results. In this paper two document classification algorithms based on pretopology and LSA are proposed, which are suitable for different situations, and their results with deft07 contest data are discussed. This work also shows future possibility of visualization integration, which could help human intervention in the classification process. RESUME. L’Analyse de la Semantique Latente (LSA) est une methode de calcul qui permet de rendre compte de l’apprentissage du langage et de son utilisation. Dans ce sens, LSA est une theorie de la signification des mots et groupes de mots (paragraphes, passages, textes) et de leur emploi. Cette propriete permet a LSA d’occuper une position enviable dans la classification automatique de documents, l’analyse de documents, etc. Bien que de nombreuses experiences indiquent que LSA peut atteindre une grande precision dans la classification de documents, ses Studia Informatica Universalis. performances sont tributaires de facteurs tels que la qualite et la quantite de documents utilises pour l’entrainement, les caracteristiques des vecteurs representatifs et la composition des documents a classer. De son cote, la pretopologie a montre son efficacite dans les domaines de la classification des donnees et de la modelisation. De plus, certaines applications ont renforce la pretopologie en ajoutant la visualisation au domaine de la classification et ont donne des resultats prometteurs. Dans cet article, nous proposons deux algorithmes de classification des documents bases sur LSA et la pretopologie, algorithmes qui sont adaptes a des situations differentes et dont nous discutons les resultats obtenus quand ils sont appliques aux donnees du defi DEFT07. Ce travail dessine egalement les possibilites futures d’integration de la visualisation, integration qui pourra contribuer a l’intervention humaine dans les processus de
This study investigates whether figurative comprehension in schizophrenia is influenced by the salience of idiomatic meaning, and whether it is affected by clinical and demographic factors and IQ. Twenty-seven schizophrenic patients and 25 healthy participants performed a semantic relatedness judgement task which required the comprehension of idioms with two plausible meanings (literal and figurative). The study also used literal expressions. The figurative meaning of the idioms was less salient (ILS), more salient (IFS), or equally salient (IES) compared to the literal meaning. The results showed "a salience effect" (i.e., all participants understood the salient meanings better than the less salient meanings). There was also a "figurativeness effect" (i.e., healthy individuals understood the figurative meaning of IES better than the literal meaning but not schizophrenic patients). In patients, their thought disorder influenced the figurative comprehension of IFS. The verbal IQ influenced the figurative comprehension of ILS. The thought disorder, the verbal IQ, and the educational level influenced the figurative comprehension of IES. The patients' clinically evaluated concretism was associated with a reduced figurative comprehension of IFS and IES evaluated at a cognitive level. The results are discussed in relation to cognitive mechanisms which underscore figurative comprehension in schizophrenia.
This article presents the current state of a work in progress, whose objective is to better understand the effects of factors that significantly influence the performance of latent semantic analysis (LSA). A difficult task, which consisted of answering (French) biology multiple choice questions, was used to test the semantic properties of the truncated singular space and to study the relative influence of the main parameters. A dedicated software was designed to fine-tune the LSA semantic space for the multiple choice questions task. With optimal parameters, the performances of our simple model were quite surprisingly equal or superior to those of seventh- and eighthgrade students. This indicates that semantic spaces were quite good despite their low dimensions and the small sizes of the training data sets. In addition, we present an original entropy global weighting of the answers’ terms for each of the multiple choice questions, which was necessary to achieve the model’s success.
The present research addresses flow people interpret novel noun-noun conceptual combinations. First, we focused oil two types of conceptual combinations: property and relational combinations. Secondly, we manipulated the order of the constituents. Finally, we Studied if the interpretation in terms of "Property" or "Relation" changes along with age. So, four groups of 6-, 8- and 10-year-old children and adults participated in a production task. Our results indicated that the interpretations in terms of relation were more frequent for the "Relation" combinations compared to the "Property" ones. Property-transferring interpretations increased with age when Property combinations are presented. The most frequent interpretations followed the order Head noun-Modifier, which is opposite to the order observed in English. (C) 2009 Societe francaise de psychologie. Published by Elsevier Masson SAS. All rights reserved.