Trained Words Research Articles

Definition of customer requirements in big data using word vectors and affinity propagation clustering.

Customer requirements (CRs) have a significant impact on product design. The existing methods of defining CRs, such as customer surveys and expert evaluations, are time-consuming, inaccurate and subjective. This paper proposes an automatic CRs definition method based on online customer product reviews using the big data analysis. Word vectors are defined using a continuous bag of words (CBOW) model. Online customer reviews are searched by a crawling method and filtered by the parts of speech and frequency of words. Filtered words are then clustered into groups by an affinity propagation (AP) clustering method based on trained word vectors. Exemplars in each clustering group are finally used to define CRs. The proposed method is verified by case studies of defining CRs for product design. Results show that the proposed method has better performance to determine CRs compared to existing CRs definition methods.

Maximizing Cochlear Implant Outcomes with Short-Term Aural Rehabilitation.

Increasing numbers of adults are receiving cochlear implants (CIs) and many achieve high levels of speech perception and improved quality of life. However, a proportion of implant recipients still struggle due to limited speech recognition and/or greater communication demands in their daily lives. For these individuals a program of aural rehabilitation (AR) has the potential to improve outcomes. The study investigated the effects of a short-term AR intervention on speech recognition, functional communication, and psychosocial outcomes in post lingually deafened adult CI users. The experimental design was a multisite clinical study with participants randomized to either an AR treatment or active control group. Each group completed 6 weekly 90-minute individual treatment sessions. Assessments were completed pretreatment, 1 week and 2 months post-treatment. Twenty-five post lingually deafened adult CI recipients participated. AR group: mean age 66.2 (48-80); nine females, four males; months postactivation 7.7 (3-16); mean years severe to profound deafness 18.4 (2-40). Active control group: mean age 62.8 (47-85); eight females, four males; months postactivation 7.0 (3-13); mean years severe to profound deafness 18.8 (1-55). The AR protocol consisted of auditory training (words, sentences, speech tracking), and psychosocial counseling (informational and communication strategies). Active control group participants engaged in cognitive stimulation activities (e.g., crosswords, sudoku, etc.). Repeated measures ANOVA or analysis of variance, MANOVA or multivariate analysis of variance, and planned contrasts were used to compare group performance on the following measures: CasperSent; Hearing Handicap Inventory; Nijmegen Cochlear Implant Questionnaire; Client Oriented Scale of Improvement; Glasgow Benefit Inventory. The AR group showed statistically significant improvements on speech recognition performance, psychosocial function, and communication goals with no significant improvement seen in the control group. The two groups were statistically equivalent on all outcome measures at preassessment. The robust improvements for the AR group were maintained at 2 months post-treatment. Results of this clinical study provide evidence that a short-term AR intervention protocol can maximize outcomes for adult post lingually deafened CI users. The impact of this brief multidimensional AR intervention to extend CI benefit is compelling, and may serve as a template for best practices with adult CI users.

Impact of Mixed Cognitive Intervention Training on Early Onset Dementia.

ObjectivesThe aim of this study was to investigate the impact of mixed cognitive intervention training using spaced retrieval training, and errorless learning in participants with early onset dementia. This was based on reality orientation therapy for cognitive function, depression, and occupational performance of patients.MethodsTwo early onset vascular dementia patients (> 65 years) with mild or moderate impairment were enrolled in a pre-test - post-test single-subject research design study. Prior to the study, the caregivers were interviewed about meaningful times, people, places, and areas of interest for the participant. A list of individual training words were selected based upon this information, and the participant was instructed to recall them after a 45-second, 90-second, 6-minute, and 12-minute delay. Baseline (3 sessions), intervention (20 sessions), and a second baseline period (3 sessions) were conducted. Activities of daily living were measured, and cognition was measured using the Consortium to Establish a Registry of Alzheimer’s Disease Korean version, whilst depression was measured using the Korean Form Geriatric Depression Scale, and task performance and satisfaction measured by the Canadian Occupational Performance Measure.ResultsAfter intervention, both participants showed improvements in activities of daily living (ADL), word list memory/recognition, trail making A, occupational performance, and satisfaction improvement, which was clinically significant in 1 participant who also had a reduced score in the scale of depression classifying him as not depressed.ConclusionSpaced retrieval training and errorless learning based on reality orientation therapy is an effective intervention in patients with early onset dementia and mild or moderate impairment.

Morphological Skip-Gram: Replacing FastText characters n-gram with morphological knowledge

Natural language processing systems have attracted much interest of the industry. This branch of study is composed of some applications such as machine translation, sentiment analysis, named entity recognition, question and answer, and others. Word embeddings (i.e., continuous word representations) are an essential module for those applications generally used as word representation to machine learning models. Some popular methods to train word embeddings are GloVe and Word2Vec. They achieve good word representations, despite limitations: both ignore morphological information of the words and consider only one representation vector for each word. This approach implies the word embeddings does not consider different word contexts properly and are unaware of its inner structure. To mitigate this problem, the other word embeddings method FastText represents each word as a bag of characters n-grams. Hence, a continuous vector describes each n-gram, and the final word representation is the sum of its characters n-grams vectors. Nevertheless, the use of all n-grams character of a word is a poor approach since some n-grams have no semantic relation with their words and increase the amount of potentially useless information. This approach also increase the training phase time. In this work, we propose a new method for training word embeddings, and its goal is to replace the FastText bag of character n-grams for a bag of word morphemes through the morphological analysis of the word. Thus, words with similar context and morphemes are represented by vectors close to each other. To evaluate our new approach, we performed intrinsic evaluations considering 15 different tasks, and the results show a competitive performance compared to FastText. Moreover, the proposed model is $40\%$ faster than FastText in the training phase. We also outperform the baseline approaches in extrinsic evaluations through Hate speech detection and NER tasks using different scenarios.

A hybrid of CNN and LSTM methods for securing web application against cross-site scripting attack

<span>Cross-site scripting (XSS) is today one of the biggest threatthat could targeting the Web application. Based on study published by the open web applications security project (OWASP), XSS vulnerability has been present among the TOP 10 Web application vulnerabilities.Still,an important security-related issue remains how to effectively protect web applications from XSS attacks.In first part of this paper, a method for detecting XSS attack was proposed by combining </span><span lang="EN-GB">convolutional</span><span> neural network (CNN) with long short term memories<strong> (</strong>LSTM), Initially, pre-processing was applied to XSS Data Set by decoding, generalization and tokanization, and then word2vec was applied to convert words into word vectors in XSS payloads. And then we use the combination CNN with LSTM to train and test word vectors to produce a model that can be used in a web application. Based on the obtaned results, it is observed that the proposed model achevied an excellent result with accuracy of 99.4%.</span>

The challenge of achieving greater generalization in phonological treatment of Aphasia

ABSTRACT Background Stimulus selection is important to anomia treatment because similarity between trained and untrained words in the mental lexicon may influence treatment generalization. We focused on phonological similarity between trained and untrained words from a clinical trial of Phonomotor Treatment (PMT) that showed gains in confrontation naming accuracy of untrained words post-treatment. One way to capture the amount of similarity between the trained and untrained words is to consider the phonological network path distance between words. We posited that the distance between trained and untrained words in a phonological network could account for the improvement in confrontation naming accuracy post-treatment. Aim To define the phonological network distance between trained and untrained words that influences change in confrontation naming accuracy post-treatment. Methods and procedures We retrospectively analyzed data from 28 people with aphasia who received PMT as part of a clinical trial. Participants completed confrontation naming (baseline, post-treatment, and 3-months post-treatment) of words varying in phonological distance to the treatment stimuli. We used a phonological network to calculate the average shortest path length (ASPL), defined by the number of phoneme differences, between an untrained word and all trained words. We used mixed effects regression models to predict change in confrontation naming accuracy of untrained words post-treatment from ASPL. Several post-hoc analyses were also conducted. Outcomes and results We found no effect of ASPL on change in confrontation naming accuracy of untrained words immediately post- and 3-months post-treatment. However, post-hoc analyses indicated significant subject heterogeneity and limitations in observable path distance between trained and untrained words. Conclusion Despite the clinical trial report that confrontation naming of untrained words improved after PMT, we found no overall effect of ASPL on the amount of improvement. We discuss further investigation of the entire domain of phonological sequence knowledge (the phonological sequence knowledge landscape) and its influence on treatment generalization, and the potential importance of identifying predictors of treatment response to enhance the effects of treatment generalization.

Attention-based BiLSTM fused CNN with gating mechanism model for Chinese long text classification

Neural networks have been widely used in the field of text classification, and have achieved good results on various Chinese datasets. However, for long text classification, there are a lot of redundant information in text data, and some of the redundant information may involve other topic information, which makes long text classification challenging. To solve the above problems, this paper proposes a new text classification model, called attention-based BiLSTM fused CNN with gating mechanism(ABLG-CNN). In ABLG-CNN, word2vec is used to train word vector representation. The attention mechanism is used to calculate context vector of words to derive keyword information. Bidirectional long short-term memory network(BiLSTM) captures context features. Based on this, convolutional neural network(CNN) captures topic salient features. In view of the possible existence of sentences involving other topic information in long texts, a gating mechanism is introduced to assign weights to BiLSTM and CNN output features to obtain text fusion features that are favorable for classification. ABLG-CNN can capture text context semantics and local phrase features, and perform experimental verification on two long text news datasets. The experimental results show that ABLG-CNN’s classification performance is better than other latest text classification methods.

Value perception impact and countermeasures analysis of new energy vehicle purchase behavior based on consumer level user review big data mining

The development of new energy vehicles is inseparable from the drive of consumers. Therefore, to explore the influencing factors of purchase behavior from the consumer's personal level is helpful for businesses to adopt corresponding sales strategies and the government to adopt relevant policies. Based on the individual level of consumers, this paper constructs a new energy vehicle purchase behavior prediction model from the review text, and explores the predictive effect of consumer personal factors on the purchase behavior of new energy vehicles. First of all, this paper proposes a quantitative method of consumer individual level factors, which combines word-of-mouth reviews with statistics. In this method, word2vec is used to train word vectors in word-of-mouth corpus to mine initial keywords, and core keywords are selected through statistical correlation analysis. Secondly, based on the core keywords of consumers' personal level, the gbdt model is constructed to predict the purchase behavior of new energy vehicles. The results show that the probability of correctly predicting consumers' purchase behavior is more than 72%.

White Matter Integrity Predicts Electrical Stimulation (tDCS) and Language Therapy Effects in Primary Progressive Aphasia

Background Transcranial direct current stimulation (tDCS), in conjunction with language therapy, improves language therapy outcomes in primary progressive aphasia (PPA). However, no studies show whether white matter integrity predicts language therapy or tDCS effects in PPA. Objective We aimed to determine whether white matter integrity, measured by diffusion tensor imaging (DTI), predicts written naming/spelling language therapy effects (letter accuracy on trained and untrained words) with and without tDCS over the left inferior frontal gyrus (IFG) in PPA. Methods Thirty-nine participants with PPA were randomly assigned to tDCS or sham condition, coupled with language therapy for 15 daily sessions. White matter integrity was measured by mean diffusivity (MD) and fractional anisotropy (FA) in DTI scans before therapy. Written naming outcomes were evaluated before, immediately after, 2 weeks, and 2 months posttherapy. To assess tDCS treatment effect, we used a mixed-effects model with treatment evaluation and time interaction. We considered a forward model selection approach to identify brain regions/fasciculi of which white matter integrity can predict improvement in performance of word naming. Results Both sham and tDCS groups significantly improved in trained items immediately after and at 2 months posttherapy. Improvement in the tDCS group was greater and generalized to untrained words. White matter integrity of ventral language pathways predicted tDCS effects in trained items whereas white matter integrity of dorsal language pathways predicted tDCS effects in untrained items. Conclusions White matter integrity influences both language therapy and tDCS effects. Thus, it holds promise as a biomarker for deciding which patients will benefit from language therapy and tDCS.

Effects of a Word Boxes Electronic App on Improving Letter-Sound Correspondences of Learners of English as an Additional Language

The purpose of this study was to examine the effects of Build a Word-Easy Spelling with Phonics, an electronic application of the word box intervention, on the acquisition, maintenance, and generalization of letter-sound correspondences for Learners of English as an Additional Language (LEALs). A single-subject multiple probe experimental design across five kindergarten LEALs. Students’ acquisition and maintenance of letter-sound correspondences were measured by the number of correct letter-sound correspondences identified in trained words and the number of trained words read correctly. Generalization measures involved identifying letter-sound correspondences in untrained real and nonsense words and a spelling task. Visual analysis and TAU-U calculations were used to examine intervention effects. Results demonstrate that the intervention Build a Word-Easy Spelling with Phonics substantially improved the acquisition, maintenance, and generalization of letter-sound correspondences for the five LEALs. This study extends research on the effects of systematic phonics interventions for LEALs. Implications for practice are discussed.

Evaluating a Maintenance-Based Treatment Approach to Preventing Lexical Dropout in Progressive Anomia.

Purpose Primary progressive aphasia (PPA) and the amnestic variant of Alzheimer's disease (AD) are neurodegenerative conditions characterized by a profound loss of functional communication abilities. Communicative impairment in AD and PPA is especially apparent in the domain of naming common objects and familiar faces. We evaluated the effectiveness of a language intervention targeting maintenance of an individualized core vocabulary in a longitudinal cohort of older adults experiencing either PPA or AD. Method PPA (n = 9) and AD (n = 1) patients were administered a semantically based language treatment for up to 2 years. Patients repeatedly named and generated semantic features for a personalized lexicon consisting of 100 words. We evaluated naming accuracy and off-line neuropsychological measures at four successive timepoints. Naming accuracy was assessed in patients (n = 7) who completed at least three recurrent evaluations. Off-line neuropsychological performance was assessed across timepoints in all patients. Results Patients demonstrated relative preservation of naming trained words relative to a steep decline for untrained (control) words. The greatest decrements were observed for naming people relative to objects. Conclusion These results suggest that consistent training of a finite set of words can protect a core lexicon composed of crucial target concepts (e.g., a spouse's name). We discuss potential benefits and clinical implications of maintenance-based approaches to promoting language functioning in the context of neurodegeneration.

Efficacy of spoken word comprehension therapy in patients with chronic aphasia: a cross-over randomised controlled trial with structural imaging

ObjectiveThe efficacy of spoken language comprehension therapies for persons with aphasia remains equivocal. We investigated the efficacy of a self-led therapy app, ‘Listen-In’, and examined the relation between brain structure...

Pseudoword spelling ability predicts response to word spelling treatment in acquired dysgraphia

ABSTRACT Although rehabilitation of acquired dysgraphia can be quite effective, identifying predictors of responsiveness to treatment is useful for prognosis and individualization of treatment protocols. This study examined whether various features of treatment response were predicted by the integrity of one or more of the central cognitive components of spelling: orthographic long-term memory, orthographic working memory, and phoneme-grapheme conversion. Twenty dysgraphic individuals received 12 weeks of bi-weekly, individualized, lexically-based spelling rehabilitation using a spell-study-spell paradigm. Linear multiple regression modelling examined whether the type and severity of the dysgraphic deficit, assessed before rehabilitation, predicted the magnitude and rate of improvement, generalization to untrained items and maintenance of treatment gains. The results revealed that pseudoword spelling accuracy – indexing the integrity of the phoneme-grapheme conversion system – was the only factor examined that significantly predicted the rate of accuracy gains for trained words as well as the extent of generalization to untrained words. Pre-treatment pseudoword spelling accuracy also predicted retention of gains for trained and untrained words at 3-month follow-up. These findings reveal that the integrity of the phoneme-grapheme conversion system prior to dysgraphia rehabilitation may play a key role in rehabilitation-driven recovery, even when the treatment approach targets lexical rather than pseudoword spelling processes.

Effects of high variability phonetic training on cue reweighting in non-native vowel perception by adult Chinese learners

While native English speakers have been found to primarily use vowel spectrum on /i/-/ɪ/ perception, Chinese learners of English dominantly use vowel duration. This study examined training effects on cue reweighting in perceiving /i/-/ɪ/ by Chinese learners. We modified the canonical HVPT paradigm with introducing variability along the secondary dimension for the contrast distinction (i.e., duration). Forty native Chinese-speaking adults were randomly assigned to two groups: the training group and the control group. Pre- and post- training tests used the tasks of natural word identification and synthetic phoneme identification. Synthesized phoneme stimuli were the two-dimensional (spectrum and duration) stimulus continua ranging from English /i/ to /ɪ/, and the natural word stimuli included both trained and untrained words produced by novel talkers. The results demonstrated that the training group showed more reliance on the primary spectral cue and less reliance on the secondary durational cue after training. The corroborating data from the word identification showed the training group improved significantly and generalized learning to new talkers and phonetic contexts. The control group did not show similar changes. These results indicated that high variability phonetic training is effective on helping L2 learners to retune attention to primary cues which facilitate generalization outcomes.

Generalisation in language learning can withstand total sleep deprivation

Research suggests that sleep plays a vital role in memory. We tested the impact of total sleep deprivation on adults’ memory for a newly learned writing system and on their ability to generalise this knowledge to read untrained novel words. We trained participants to read fictitious words printed in a novel artificial orthography, while depriving them of sleep the night after learning (Experiment 1) or the night before learning (Experiment 2). Following two nights of recovery sleep, and again 10 days later, participants were tested on trained words and untrained words, and performance was compared to control groups who had not undergone sleep deprivation. Participants showed a high degree of accuracy in learning the trained words and in generalising their knowledge to untrained words. There was little evidence of impact of sleep deprivation on memory or generalisation. These data support emerging theories which suggest sleep-associated memory consolidation can be accelerated or entirely bypassed under certain conditions, and that such conditions also facilitate generalisation.

Effects of Semantic Feature Type, Diversity, and Quantity on Semantic Feature Analysis Treatment Outcomes in Aphasia.

Purpose Semantic feature analysis (SFA) is a naming treatment found to improve naming performance for both treated and semantically related untreated words in aphasia. A crucial treatment component is the requirement that patients generate semantic features of treated items. This article examined the role feature generation plays in treatment response to SFA in several ways: It attempted to replicate preliminary findings from Gravier et al. (2018), which found feature generation predicted treatment-related gains for both trained and untrained words. It examined whether feature diversity or the number of features generated in specific categories differentially affected SFA treatment outcomes. Method SFA was administered to 44 participants with chronic aphasia daily for 4 weeks. Treatment was administered to multiple lists sequentially in a multiple-baseline design. Participant-generated features were captured during treatment and coded in terms of feature category, total average number of features generated per trial, and total number of unique features generated per item. Item-level naming accuracy was analyzed using logistic mixed-effects regression models. Results Producing more participant-generated features was found to improve treatment response for trained but not untrained items in SFA, in contrast to Gravier et al. (2018). There was no effect of participant-generated feature diversity or any differential effect of feature category on SFA treatment outcomes. Conclusions Patient-generated features remain a key predictor of direct training effects and overall treatment response in SFA. Aphasia severity was also a significant predictor of treatment outcomes. Future work should focus on identifying potential nonresponders to therapy and explore treatment modifications to improve treatment outcomes for these individuals. Supplemental Material https://doi.org/10.23641/asha.12462596.

Deep learning based emotion analysis of microblog texts

Traditional text emotion analysis methods are primarily devoted to studying extended texts, such as news reports and full-length documents. Microblogs are considered short texts that are often characterized by large noises, new words, and abbreviations. Previous emotion classification methods usually fail to extract significant features and achieve poor classification effect when applied to processing of short texts or micro-texts. This study proposes a microblog emotion classification model, namely, CNN_Text_Word2vec, on the basis of convolutional neural network (CNN) to solve the above-mentioned problems. CNN_Text_Word2vec introduces a word2vec neural network model to train distributed word embeddings on every single word. The trained word vectors are used as input features for the model to learn microblog text features through parallel convolution layers with multiple convolution kernels of different sizes. Experiment results show that the overall accuracy rate of CNN_Text_Word2vec is 7.0% higher than that achieved by current mainstream methods, such as SVM, LSTM and RNN. Moreover, this study explores the impact of different semantic units on the accuracy of CNN_Text_Word2vec, specifically in processing of Chinese texts. The experimental results show that comparing to using feature vectors obtained from training words, feature vector obtained from training Chinese characters yields a better performance.

Orthographic Learning In L1 And L2 Alphabets: The Impact of Phonological Inconsistency Across Cyrillic and Roman Scripts

The acquisition of new orthographic representations has been systematically found as a fast and accurate process in monolingual readers. The present study aims to extended this research to biliterate and bialphabetic population, addressing the impact of phonological inconsistencies across the native (L1) and second (L2) alphabet. Naming latencies were collected from 50 Russian-English biliterates through a reading-aloud task, in which familiar and novel words were repeatedly presented across 10 blocks. The stimuli were equally divided in three script conditions: Cyrillic, Roman (in both cases, using script-specific graphemes) and ambiguous (using graphemes common to L1 and L2 alphabets, and thus phonologically inconsistent). Linear mixed-effects modelling revealed differences in the process of orthographic learning depending on the script. Thus, although naming latencies for novel and familiar words in the ambiguous condition were matched along the training, this effect was much faster in conditions of phonological consistency. Nonetheless, post-training outcomes of learning revealed similar recall and recognition performance in familiar and trained words regardless of the script. Overall, our results indicate that phonological inconsistency interferes with the decoding of novel words but does not prevent the efficient achievement of orthographic representations in biliterates.

Intensive computer-based phonics training in the educational setting of children with Down syndrome: An explorative study.

Children with Down syndrome (DS) using intensive computer-based phonics (GraphoGame, GG) were studied. The children's independence and improvement in phonological processing, letter knowledge, word decoding, and reading strategies were investigated. Seventeen children (5-16 years) with DS participated in a crossover design through 8 weeks (one period), with three test sessions separated by 4 weeks. Children were randomly assigned to GG intervention or regular schooling (RS). All children completed one period and eight children completed two periods. A majority gradually became independent in managing GG. At the group level, very little benefit was found from working with GG. At the individual level, several children with mild to severe intellectual disabilities showed increased decoding of trained words. After one period of GG and RS, an increase in alphabetically decoded words was found. The finding suggests that when individual challenges are considered, computer-based phonics may be beneficial for children with DS in their educational setting.

Detecting malicious JavaScript code based on semantic analysis

Web development technology has undergone tremendous evolution, the creation of JavaScript has greatly enriched the interactive capabilities of the client. However, attackers use the dynamics feature of JavaScript language to embed malicious code into web pages for the purpose of drive-by-download, redirection, etc. The traditional method based on static feature detection is difficult to detect the malicious code after obfuscation, and the method based on dynamic analysis has low efficiency. To overcome these challenges, this paper proposes a static detection model based on semantic analysis. The model firstly generates an abstract syntax tree from JavaScript source codes, then automatically converts them to syntactic unit sequences. FastText algorithm is introduced to training word vectors. The syntactic unit sequences are represented as word vectors which will be input into Bi-LSTM network with attention mechanism. The detection model with Bi-LSTM network with attention mechanism is the key to detect malicious JavaScript. We experimented with the dataset using a five-fold cross-validation method. Experiments showed that the model can effectively detect obfuscated malicious JavaScript code and improve the detection speed, with a precision of 0.977 and recall of 0.974.