Previous studies have reported that deficits in attention are often a common complaint in individuals suffering from pain and attentional impairment in patients with pain has been demonstrated on a variety of neuropsychological measures. Much of the research to-date, however, has not taken into account extraneous factors that may contribute to observed cognitive deficits. Using the California Verbal Learning Test - II, attention and memory performance was examined in two clinical populations (pain and mild traumatic brain injury) while controlling for effort using the Word Memory Test. Controlling for effort led to different explanations of poor performance on attention variables. While mild deficits were expected, and could be accounted for by psychological factors (i.e. somatization), extremely poor performance was more likely related to poor effort. The findings of this study strongly support the necessity of measuring effort during neuropsychological and pain psychological evaluations.

Universities have increasingly sought to provide accommodative services to students with learning disorders and Attention-Deficit/Hyperactivity Disorder (ADHD) in recent decades thereby creating a need for diagnostic batteries designed to evaluate cognitive abilities relevant to academic performance. Given that accommodative services (extended time on tests, alternate test forms, etc.) provide incentive to distort impairment steps should be taken to estimate the rate at which students distort impairment and to evaluate the accuracy with which symptom distortion is identified. In order to address these concerns, the Word-Memory Test, Test of Memory Malingering, and Fake Bad Scale (of the MMPI-2) were compared in terms of their clinical utility in a university sample within a two-part study. In the first portion of the study, an analogue design (which included a control group (n = 29) and an experimental group (n = 30) that was asked to simulate an academic disability) was used to calculate the sensitivity and specificity of each measure. In the second portion of this study, scores were collected for 121 consecutively presenting students who were evaluated for academic difficulty at a large private university. Failure rates on measures of malingering placed the base rate of malingering within this population between 10 and 25 percent. The Word-Memory Test (WMT) demonstrated the most robust sensitivity and specificity. The modest sensitivity of the Test of Memory Malingering (TOMM) can be partially explained by the ease with which the measure is completed by university students as well as the format of its presentation. Although the scores on Fake Bad Scale (FBS) are modestly correlated with group membership (between controls and simulators), its use should be discouraged in this context due to poor sensitivity and to high rates of false positives.

Detection of sub-optimal effort is a critical element of all psychological assessment procedures. Failure to consider the validity of the client's performance and symptom reporting may result in inaccurate conclusions about the degree of impairment. Because the American with Disabilities Act requires colleges to provide accommodations for students with documented disabilities, providing resources for students feigning impairment may ultimately drain university resources intended to help those students with disabilities. This study sought to examine the relationship between two different types of measures of effort and variables related to academic ability. De-identified archival data was gathered from the University Accessibility Center (UAC) at Brigham Young University (BYU) which provided psychological assessments for accommodation seeking students (N = 602) for a reduced fee. Measures used to detect sub-optimal effort included the Test of Memory Malingering (TOMM), Word Memory Test (WMT), Validity Indicator Profile (VIP), California Verbal Learning Test-Second Edition (CVLT-II), Reliable Digit Span (RDS), and the Integrated Visual and Auditory Continuous Performance Advanced Edition (IVA-AE). Measures indicating academic ability included select subtests from the Woodcock Johnson Test of Achievement Third Edition (WJ-III). Additionally, Matrix Reasoning of the Wechsler Adult Intelligence Scale Fourth Edition (WAIS-IV) was included as a cognitive measure of nonverbal IQ. Two point biserial correlations were conducted. Results indicated that the nonverbal portion of the VIP had a significant relationship with writing fluency. The TOMM also had a significant relationship with writing fluency. Additionally, results demonstrated that Reliable Digit Span had a significant relationship with Academic Fluency, Writing Fluency, Letter Word Identification, and Math Fluency. Data suggests that university disability service offices may wish to include the RDS, TOMM, and VIP in their considerations of effort.

The Word Memory Test (WMT) is a popular effort test that requires participants to memorize lists of paired words and repeat them back in a variety of different memory tasks. Four brain injured patients participated in two trials of the delayed recall (DR) portion of the WMT while undergoing fMRI scanning. In the first trial subjects put forth full effort, and during the second trial subjects were instructed to simulate increased memory impairment in order to represent poor effort. fMRI activation from both trials were compared in order to contrast full and simulated poor effort activation patterns during the WMT. Raw scores from full effort and simulated poor effort trials were compared to a control group to test the hypothesis that a brain injured population will score lower than a healthy population on the WMT while putting forth full effort. Raw score results showed lower WMT scores for TBI group. fMRI results showed larger between-group differences than between-condition differences, suggesting that the WMT is sensitive to TBI.

According to the Americans with Disabilities Act, universities and colleges are required to provide accommodative services for students with disabilities. Many studies have examined the role of malingering mental health symptoms in order to obtain psychotropic medications, but very little research has been done on the role of accommodations as secondary gain in students who may malinger learning disabilities. This study sought to examine both the usefulness of implementing specific malingering detection measures in psychological evaluations with university students and the agreement of those measures within the population. Archival data was gathered from a university accommodation clinic that provided free psychological evaluations for consecutively presenting students (N=121). Four malingering detection measures were used: the Test of Memory and Malingering (TOMM), the Word Memory Test (WMT), the WAIS Digit Span (DS) and two cut scores for the MMPI-2 F Scale (F Scale 80 and F Scale 95). Scores for these four malingering detection measures were compared in terms of their agreement rates, their classification rates (at a 10% malingering base rate recommendation), and their sensitivity, specificity, positive and negative predictive powers using both the TOMM and WMT independently as diagnostic criterion. A qualitative examination of the data revealed that different combinations of measures did classify some of the same respondents as malingering. Results indicated that each of these four measures share the ability to detect malingering in its different forms and have similar classification rates. Although the TOMM and WMT likely provide overlapping information, the pragmatic implementation of one of these measures may assist in the evaluation of suspected malingering with accommodation-seeking students.

Dans ce mémoire, nous examinons certaines propriétés des représentations distribuées de mots et nous proposons une technique pour élargir le vocabulaire des systèmes de traduction automatique neurale. En premier lieu, nous considérons un problème de résolution d'analogies bien connu et examinons l'effet de poids adaptés à la position, le choix de la fonction de combinaison et l'impact de l'apprentissage supervisé. Nous enchaînons en montrant que des représentations distribuées simples basées sur la traduction peuvent atteindre ou dépasser l'état de l'art sur le test de détection de synonymes TOEFL et sur le récent étalon-or SimLex-999. Finalament, motivé par d'impressionnants résultats obtenus avec des représentations distribuées issues de systèmes de traduction neurale à petit vocabulaire (30 000 mots), nous présentons une approche compatible à l'utilisation de cartes graphiques pour augmenter la taille du vocabulaire par plus d'un ordre de magnitude. Bien qu'originalement développée seulement pour obtenir les représentations distribuées, nous montrons que cette technique fonctionne plutôt bien sur des tâches de traduction, en particulier de l'anglais vers le français (WMT'14). / In this thesis, we examine some properties of word embeddings and propose a technique to handle large vocabularies in neural machine translation. We first look at a well-known analogy task and examine the effect of position-dependent weights, the choice of combination function and the impact of supervised learning. We then show that simple embeddings learnt with translational contexts can match or surpass the state of the art on the TOEFL synonym detection task and on the recently introduced SimLex-999 word similarity gold standard. Finally, motivated by impressive results obtained by small-vocabulary (30,000 words) neural machine translation embeddings on some word similarity tasks, we present a GPU-friendly approach to increase the vocabulary size by more than an order of magnitude. Despite originally being developed for obtaining the embeddings only, we show that this technique actually works quite well on actual translation tasks, especially for English to French (WMT'14).

Tato diplomová práce popisuje principy statistického strojového překladu a demonstruje, jak sestavit systém pro statistický strojový překlad Moses. V přípravné fázi jsou prozkoumány volně dostupné bilingvní česko-anglické korpusy. Empirická analýza časové náročnosti vícevláknových nástrojů pro zarovnání slov demonstruje, že MGIZA++ může dosáhnout až pětinásobného zrychlení, zatímco PGIZA++ až osminásobného zrychlení (v porovnání s GIZA++). Jsou otestovány tři způsoby morfologického pre-processingu českých trénovacích dat za použití jednoduchých nefaktorových modelů. Zatímco jednoduchá lemmatizace může snížit BLEU, sofistikovanější přístupy většinou BLEU zvyšují. Positivní efekty morfologického pre-processingu se vytrácejí s růstem velikosti korpusu. Vztah mezi dalšími charakteristikami korpusu (velikost, žánr, další data) a výsledným BLEU je empiricky měřen. Koncový systém je natrénován na korpusu CzEng 0.9 a vyhodnocen na testovacím vzorku z workshopu WMT 2010.