Sentence Level Research Articles

Improving Layman Understanding of Forensic Evidence: Can the Language of Autopsy Reports and Personal Examination Reports be Made more Lay-friendly?

In murder and attempted murder cases, the information provided by the autopsy report or personal examination report and the evidence given in court by forensic pathologists is often essential for the legal outcome. These reports written by forensic pathologists contain very specialised language; however, when used in connection with legal cases, the target audience also comprises non-experts in medicine such as the police, lawyers, judges, jury members and lay judges. Therefore, the reports must be comprehensible to this lay audience. This study investigates the language of 15 written autopsy reports and personal examination reports used in court with the aim of identifying potentially incomprehensible linguistic features or features which make the reports unnecessarily complex to laymen. Results show that many linguistic elements both at word, sentence and text level can be changed to more lay-friendly options without loss of precision. We discuss best practice recommendations as well as potential barriers to implementing these recommendations in practice.

A framework for decision making to purchase the best product using feature-based opinions

As more people express their thoughts on products on various online shopping platforms, the feelings expressed in these opinions are becoming a significant source of information for marketers and buyers. These opinions have a big impact on consumers’ decision to buy the best quality product. When there are too many features or a small number of records to analyze, the decision-making process gets difficult. A recent stream of study has used the conventional quantitative star score ratings and textual content reviews in this context. In this research, a decision-making framework is proposed that relies on feature-based opinions to analyze the textual content of reviews and classify buyer’s opinions, thereby assisting consumers in making long-term purchases. The framework is proposed in this paper for product purchase decision making based on feature-based opinions and deep learning. Framework consists of four components: i) Pre-processing, ii) Feature extraction, iii) Feature-based opinion classification, and iv) Decision-making. Web scraping is used to obtain the dataset of Smartphone reviews, which is subsequently clean and pre-processed using tokenization and POS tagging. From the tagged dataset, noun labeled words are retrieved, and then the probable product’s features are extracted. These feature-based sentences or reviews are processed using a word embedding to generate review vectors that identify contextual information. These word vectors are used to construct hidden vectors at the word and sentence levels using a hierarchical attention method. With respect to each feature, reviews are divided into five classes: extremely positive, positive, extremely negative, negative, and neutral. The proposed method may readily detect a customer’s opinion on the quality of a product based on a certain attribute, which is beneficial in making a purchase choice.

Multi-Level interaction network for temporal sentence grounding in videos

Temporal sentence grounding in videos (TSGV), which aims to retrieve video segments from an untrimmed videos that semantically match a given query. Most previous methods focused on learning either local or global query features and then performed cross-modal interaction, but ignore the complementarity between local and global features. In this paper, we propose a novel Multi-Level Interaction Network for Temporal Sentence Grounding in Videos. This network explores the semantics of queries at both phrase and sentence levels, interacting phrase-level features with video features to highlight video segments relevant to the query phrase and sentence-level features with video features to learn more about global localization information. A stacked fusion gate module is designed, which effectively captures the temporal relationships and semantic information among video segments. This module also introduces a gating mechanism to enable the model to adaptively regulate the fusion degree of video features and query features, further improving the accuracy of predicting the target segments. Extensive experiments on the ActivityNet Captions and Charades-STA benchmark datasets demonstrate that the proposed method outperforms the state-of-the-art methods.

Sentence level Classification through machine learning with effective feature extraction using deep learning

Social networking website usage has increased dramatically during the past few years. Users can read other users' views, which are categorized into several sentiment classes on this medium with an array of data. These opinions are becoming more and more important while making decisions. To address the above-mentioned issues and improve the sentence-level classification's classification rate, this work introduces a new extensive pinball loss function based twin support vector machine with Deep Learning the (EPLF-TSVM-DL) to identify the polarity (negative and positive) of sentiment sentences. There are four primary components of this technique: The first portion consists of pre-processing the data to minimize noise and improve quality; the second part utilizes word embedding techniques to transform textual data into numerical data. The third part is the CNN for an efficient automatic method of extracting the features-based feature extraction and final is EPLF-TSVM-DL is used for sentence level classification that forms two classes such as Negative and Positive. The findings demonstrated that the EPLF-TSVM-DL outperforms the other classifiers with respect to of time consumption, convergence, complexity, and stability as well as true negative, true positive, error rate, false positive, precision, false negative, and classification rate.

Bilingual oral language development among dual language immersion students: Use of a Bayesian approach with language learning progressions

AbstractThe dual language development of dual language immersion (DLI) students, although often examined at the domain level (e.g., listening or reading), remains understudied for more specific skills (e.g., word, sentence, or discourse). This study examines the eleven-month progression of oral language skills in a picture description task in two languages (French and English) for early-elementary (Transitional Kindergarten through first grade) DLI students (N = 42). Using Bayesian methods, which estimate parameters using both the data and prior information, we describe French and English growth patterns as measured by learning progressions whose focus is on language features at the word, sentence, and discourse levels. For French oral language, we found evidence of meaningful positive linear growth for all language features, whereas for English oral language, meaningful linear positive growth was only detected for sophistication of topic vocabulary. Overall, coming from a French-speaking household was associated with steeper French oral language trajectories, but coming from an English-only household did not specifically impact English oral language trajectories. In both languages, grade level influenced the trajectories of some—but not all—features. We conclude with theoretical and practical implications, advocating for a language progression approach in instruction and research on bilingualism.

Dependency distance reflects L2 processing difficulty: Evidence from the relationship between dependency distance, L2 processing speed, and L2 proficiency

Aims and objectives: This study aims to investigate whether dependency distance (DD) can reflect L2 processing difficulty by exploring the relationship between DD, L2 processing speed, and L2 proficiency. Methodology: The sentences in a maze task were annotated based on dependency syntax to indicate L2 processing difficulty. The maze task then was used to capture the word-level and sentence-level processing speed. The task was successfully performed by 62 Chinese learners of English, who were assigned into an intermediate-level group and an advanced-level group according to their scores on an elicited imitation task (EIT) and a vocabulary and grammar test. Data and analysis: Reaction time (RT) data of processing word-level and sentence-level items were collected from the maze task. The DD-based estimates of word-level and sentence-level processing difficulty were calculated. Generalized linear regression models (GLMs) were employed to inspect the relationship between the DD-based estimate of processing difficulty, L2 processing speed, and L2 proficiency. Findings: Our results showed that DD significantly predicted L2 processing speed at both word and sentence levels, irrespective of participants’ L2 proficiency. As DD increased, advanced learners’ word-level processing time grew significantly slower than that of intermediate learners. Originality: Using psycholinguistic experiments, the study is one of the first attempts to empirically validate the effectiveness of DD in capturing L2 processing difficulty. Significance: Our findings suggest that DD is useful in predicting L2 processing difficulty, and the extent to which learners are able to efficiently process dependencies with increasing distance can differentiate their L2 proficiency levels.

Speech Task Force and Quality of Life after Surgery in Children with Cleft Lip and Palate: Limitation of Professionals.

Background Shortage of speech and language therapists results in lack of speech services. The aims of this study were to find the effectiveness of a combination speech therapy model at Level IV: General speech and language pathologist (GSLP) and Level V: Specific speech and language pathologist (SSLP) in reduction of the number of articulation errors and promotion the quality of life (QoL) for children with cleft palate with or without cleft lip (CP ± L). Methods Fifteen children with CP ± L, aged 4 years 1 month to 10 years 9 months (median = 76 months; minimum:maximum = 49:129 months) were enrolled in this study. Pre- and post-assessment included oral peripheral examination; articulation tests via Articulation Screening Test, Thai Universal Parameters of Speech Outcomes for People with Cleft Palate, Hearing Evaluation, The World Health Organization Quality of Life Brief_Thai (WHOQOL-BRIEF-THAI) version questionnaire for QoL were performed. Speech therapy included a 3-day intensive speech camp by SSLP, five 30-minute speech therapy sessions by a GSLP, and five 1-day follow-up speech camps by SSLP that provided four 45-minute speech therapy sessions for each child. Results Post-articulation revealed statistically significant reduction of the numbers of articulation errors at word, sentence, and screening levels (median difference [MD] = 3, 95% confidence interval [CI] = 2-5; MD = 6, 95% CI = 4.5-8; MD = 2.25, 95% CI = 1.5-3, respectively) and improvement of QoL. Conclusion A speech task force consisting of a combination of Level IV: GSLP and Level V: SSLP could significantly reduce the number of articulation errors and promote QoL.

Finding common features in multilingual fake news: a quantitative clustering approach

Abstract Since the Internet is a breeding ground for unconfirmed fake news, its automatic detection and clustering studies have become crucial. Most current studies focus on English texts, and the common features of multilingual fake news are not sufficiently studied. Therefore, this article uses English, Russian, and Chinese as examples and focuses on identifying the common quantitative features of fake news in different languages at the word, sentence, readability, and sentiment levels. These features are then utilized in principal component analysis, K-means clustering, hierarchical clustering, and two-step clustering experiments, which achieved satisfactory results. The common features we proposed play a greater role in achieving automatic cross-lingual clustering than the features proposed in previous studies. Simultaneously, we discovered a trend toward linguistic simplification and economy in fake news. Furthermore, fake news is easier to understand and uses negative emotional expressions in ways that real news does not. Our research provides new reference features for fake news detection tasks and facilitates research into their linguistic characteristics.

Therapy program assisted with therapy dogs for children with Autism Spectrum Disorder (ASD)

IntroductionAutism spectrum disorders (ASD) is a neurodevelopmental disorder with an estimated lifetime prevalence of at least 1%. Some studies suggest that approximately 60% of children with ASD show emotional dysregulation and 44-86% sleeping disorders.Studies suggest that cognitive behavioral therapy and animal-assisted therapy AAT can be an intervention strategy to promote emotional and behavioral dysregulation and sleep disturbances.ObjectivesThe purpose of this study was to investigate the effect of a therapy program assisted with dogs (AAT) together with cognitive behavioral therapy (CBT) on behavioral and emotional regulation and sleep disorders in children diagnosed with ASD.MethodsThe sample was composed of 24 children between 7 and 10 years old diagnosed with ASD randomized into two groups (CGT group (control group) and AAT + CBT group (experimental group). Inclusion criteria: communication level of simple sentences, mild-moderate difficulties in behavioral and emotional regulation and sleep disturbances. Exclusion criteria: intellectual disability, children with specific dogs phobia. Assessment included ADOS-2, WISC-V, CGAS. dysregulation profile of Achenbach scale and Sleep Disturbance Scale for Children-Bruni. A program of 12 sessions (weekly one-hour sessions) focusing emotional and behavioral regulation and sleep disturbance was designed (Behavior Emotional Sleep Treatment Program. A pre-post evaluation was performed.ResultsParticipants were 20 boys and 2 girls (2 participants dropped out), with a mean age of 9. Regarding the socio-demographic and clinical characteristics, no significant differences has been observed between both groups in the global functioning measured with the Children’s Global Assessment scale (CGAS) (p=0.832), nor in the cognitive capacity (QI ) neither in reference to the associated comorbidities (p=0.103) nor in the variable prescription pharmacological treatment (p= 0.142). In emotional self-regulation, a significant improvement in emotional regulation difficulties was observed after treatment in both groups (experimental group: p=0.014; control group: p=0.012). However, the comparison between the pre-post intervention results between groups, regarding the emotional regulation variable, a greater improvement is observed in the experimental group (p=0.013). Significant improvements were also observed in sleep disorders (Bruni scale total score and in the sleep conciliation and maintenance difficulties scale) in both groups.ConclusionsTo conclude, although this is a pilot study with a small sample size and further research is needed, results suggest that a therapy program assisted with therapy dogs and CBT have positive effects on emotional dysregulation and sleep disturbances in children with ASD and offers a possible intervention strategy.Disclosure of InterestNone Declared

Subjective Modality and Means of its Expression in a Journalistic Style

This article is dedicated to the study of one of the categories of the sentence — modality. Many domestic linguists define this category as the attitude of the speaker to reality. Representing a complex phenomenon,it does not have a clear definition and according to many linguists requires clarification. The article presents the main provisions from the works of V. V. Vinogradov, who compiled a voluminous list of means denoting modal meanings, and G. A. Zolotova. The definition of the concepts of objective and subjective modality is given according to the «Russian grammar» (of 1980) and the «Linguistic encyclopedic dictionary». The means of expressing subjective modality are analyzed in journalistic style texts by means of an example of online newspaper «Culture». Cases of expression are identified and considered at the level of simple and complex sentences. It is noted that the subjective modality is associated with the use of lexical repetition, phrases, introductory words, separate, interjections, words that cause desire, possibility, intention or non-union complex sentences, complex sentences, unions and allied words. It was found in the article that the subjective modality is mostly expressed by using units of lexical and syntactic levels.

Effects of Audio-visual and Contextual Conditions on Auditory Word Recognition in Patients with Normal Pressure Hydrocephalus

Objectives: In this study, we compared performance on the auditory word recognition tasks between patients with normal pressure hydrocephalus (NPH) and healthy elderly adults (HE) under different visual, auditory, and contextual conditions to determine how these variables affect auditory comprehension with cognitive decline.Methods: We conducted a standardized comprehension assessment, Paradise Korean Western Aphasia Battery Revised (PK-WAB-R) and the auditory word recognition tasks on 30 patients with NPH and 30 HE. The auditory word recognition task was divided into noisy and quiet conditions (auditory variable), blurry and clean conditions (visual variable), and low- and high-context conditions (contextual variable).Results: Patients with NPH showed relatively good comprehension on the standardized comprehension test of PK-WAB-R. For the auditory word recognition tasks, there was a significant main effect of group, a significant within-group effect of auditory and contextual conditions, and significant interaction effects between auditory condition and group, auditory condition and contextual information, and auditory condition, contextual information, and group.Conclusion: This study confirmed that patients with NPH have difficulty processing auditory comprehension at the sentence level. In addition, the auditory word recognition performance in patients with NPH was significantly lower than HE, and that their difficulties may be more pronounced under certain conditions, such as noisy conditions and low contextual information.

The relationship between speech in noise perception and auditory working memory capacity in monolingual and bilingual adults

Objective Speech-in-noise perception is dependent on the interaction between sensory and cognitive processes. One factor that can relate to both processes is bilingualism. This study aimed to determine the correlation between auditory-working-memory and speech-in-noise in Persian monolinguals and Kurdish-Persian bilinguals. Design Speech-in-noise tests (sentences-in-noise and syllables-in-noise) and auditory-working-memory tests (forward and backward digit span, and n-back) were performed. Study sample Participants were 48 Kurdish-Persian bilinguals with a mean age of 24 (±4) years and 48 Persian monolinguals with a mean age of 25 (±2) years with normal hearing. Results Both language groups scored within normal limits in all memory and speech-in-noise tests. However, bilinguals performed significantly worse than monolinguals on all auditory-working-memory tests and the sentences-in-noise test. Monolinguals outperformed bilinguals in sentences-in-noise test (∼1.5 dB difference) and all auditory-working-memory tests (∼1 digit difference). The two groups did not significantly differ in syllables-in-noise test. Both groups had a significant correlation between working memory capacity and sentences-in-noise test. However, no significant correlation was found between syllables-in-noise and working memory capacity at any SNR. Conclusions Cognitive factors such as auditory working memory appear to correlate with speech-in-noise perception ability (at least at the sentence level) in monolingual and bilinguals young adults.

Navigating misinformation in voice messages: Identification of user‐centered features for digital interventions

AbstractMisinformation presents a challenge to democracies, particularly in times of crisis. One way in which misinformation is spread is through voice messages sent via messenger groups, which enable members to share information on a larger scale. Gaining user perspectives on digital misinformation interventions as countermeasure after detection is crucial. In this paper, we extract potential features of misinformation in voice messages from literature, implement them within a program that automatically processes voice messages, and evaluate their perceived usefulness and comprehensibility as user‐centered indicators. We propose 35 features extracted from audio files at the character, word, sentence, audio, and creator levels to assist (1) private individuals in conducting credibility assessments, (2) government agencies faced with data overload during crises, and (3) researchers seeking to gather features for automatic detection approaches. We conducted a think‐aloud study with laypersons () to provide initial insight into how individuals autonomously assess the credibility of voice messages, as well as which automatically extracted features they find to be clear and convincing indicators of misinformation. Our study provides qualitative and quantitative insights into valuable indicators, particularly when they relate directly to the content or its creator, and uncovers challenges in user interface design.

IcsPLMs: Exploring Pre-trained Language Models in Intelligent Customer Service (Student Abstract)

Pre-trained language models have shown their high performance of text processing in intelligent customer service platforms. However, these models do not leverage domain specific information. In this paper, we propose icsPLMs optimized for intelligent customer service on both word and sentence levels. Our experimental results represent that using targeted strategies can further improve the performance of pre-trained language models in this field.

"Approaches to sentiment analysis of Hungarian political news at the sentence level"

Automated sentiment analysis of textual data is one of the central and most challenging tasks in political communication studies. However, the toolkits available are primarily for English texts and require contextual adaptation to produce valid results—especially concerning morphologically rich languages such as Hungarian. This study introduces (1) a new sentiment and emotion annotation framework that uses inductive approaches to identify emotions in the corpus and aggregate these emotions into positive, negative, and mixed sentiment categories, (2) a manually annotated sentiment data set with 5700 political news sentences, (3) a new Hungarian sentiment dictionary for political text analysis created via word embeddings, whose performance was compared with other available sentiment dictionaries. (4) Because of the limitations of sentiment analysis using dictionaries we have also applied various machine learning algorithms to analyze our dataset, (5) Last but not least to move towards state-of-the-art approaches, we have fine-tuned the Hungarian BERT-base model for sentiment analysis. Meanwhile, we have also tested how different pre-processing steps could affect the performance of machine-learning algorithms in the case of Hungarian texts.

Document-Level Relation Extraction with Uncertainty Pseudo-Label Selection and Hard-Sample Focal Loss

Relation extraction is a fundamental task in natural language processing that aims to identify structured triple relationships from unstructured text. In recent years, research on relation extraction has gradually advanced from the sentence level to the document level. Most existing document-level relation extraction (DocRE) models are fully supervised and their performance is limited by the dataset quality. However, existing DocRE datasets suffer from annotation omission, making fully supervised models unsuitable for real-world scenarios. To address this issue, we propose the DocRE method based on uncertainty pseudo-label selection. This method first trains a teacher model to annotate pseudo-labels for a dataset with incomplete annotations, trains a student model on the dataset with annotated pseudo-labels, and uses the trained student model to predict relations on the test set. To mitigate the confirmation bias problem in pseudo-label methods, we performed adversarial training on the teacher model and calculated the uncertainty of the model output to supervise the generation of pseudo-labels. In addition, to address the hard-easy sample imbalance problem, we propose an adaptive hard-sample focal loss. This loss can guide the model to reduce attention to easy-to-classify samples and outliers and to pay more attention to hard-to-classify samples. We conducted experiments on two public datasets, and the results proved the effectiveness of our method.

Contextual Sentence Similarity from News Articles

An important topic in the field of natural language processing is the measurement of sentence similarity. It's important to precisely gauge how similar two sentences are. Existing methods for determining sentence similarity challenge two problems Because sentence level semantics are not explicitly modelled at training, labelled datasets are typically small, making them insufficient for training supervised neural models; and there is a training-test gap for unsupervised language modelling (LM) based models to compute semantic scores between sentences. As a result, this task is performed at a lower level. In this paper, we suggest a novel paradigm to handle these two concerns by robotics method framework. The suggested robotics framework is built on the essential premise that a sentence's meaning is determined by its context and that sentence similarity may be determined by comparing the probabilities of forming two phrases given the same context. In an unsupervised way, the proposed approach can create high-quality, large-scale datasets with semantic similarity scores between two sentences, bridging the train-test gap to a great extent. Extensive testing shows that the proposed framework does better than existing baselines on a wide range of datasets.

Life fits home: Exploring people's experience with a COVID-19 tracing app in Turkey through a qualitative study

Mobile apps have been developed to manage COVID-19 in many countries. However, for these apps to be truly effective, they need to be widely adopted by society. To date, there has been less qualitative research on user experiences and perspectives on these apps. The goal of this study was to explore how users perceive and use different features of a COVID-19 tracing app provided in Turkey. Semi-structured interviews ( n = 15) were conducted over the phone, audio recorded, and then transcribed verbatim. The transcriptions were analyzed through thematic analysis. The analysis began by categorizing each transcript at the sentence level through open codes, which then grouped into broader themes. In total, 8 male and 7 female participants from 10 different cities took part in the interviews. On average, participants were 32 years ( SD = 6.8) old, and their app usage experience was 9.8 months ( SD = 3.1). Thematic analysis revealed five key themes: long-term adoption and engagement, perception and desired features, privacy concerns, reliability, and emotional and behavioral impact. The main reasons for users to download and use the app were to check the COVID-19 density map in their region and to access personal health codes for security checks. The COVID-19 density map caused many users to change where they usually travel and shop. Most participants felt comfortable sharing their personal data to collectively manage the pandemic. The majority found the app useful, stating that the app allows them to take precautions against the virus and therefore helps them feel good emotionally. Future contact tracing apps need to provide indicators to enable users to evaluate whether app use is making a difference in the pandemic.

A Neural Network model for Indian Supreme Court legal judgment prediction

Industry experts are motivated to collect, collate, and analyse historical data in the legal sector in attempt to predict court case outcomes as the amount of historical data available in this field has increased over time. But using judicial data to predict and defend court judgements is no simple undertaking. Using Machine Learning (ML) models and traditional approaches for categorical feature encoding, previous research on predicting court outcomes using limited experimental datasets produced a number of unexpected predictions. The paper proposes an ensemble model combining Convolutional Neural Network (CNN), attention mechanism and eXtreme Gradient Boosting (XGB) algorithm. This model is primarily based on a self-attention network, which could simultaneously capture linguistic relationships over lengthy sequences like RNN (Recurrent Neural Network) and is nevertheless speedy to train like CNN. C-XGB can obtain accuracy that surpasses the state-of-art model on numerous classification/prediction tasks simultaneously as being twice as speedy to train. The proposed C-XGB model is designed to process the documents hierarchically and calculates the attention weights. Two convolutional layers are used to calculate the attention weights, one at the word level and another at the sentence level. And finally, at the last layer, the XGB algorithm predicts the input case file’s outcome. The experimental results shows that the proposed model outperforms the existing model with 4.67% improvement in accuracy value.

The Effect of Pitch Accent on the Perception of English Lexical Stress: Evidence from English and Mandarin Chinese Listeners

The relative weighting of f0 and vowel reduction in English spoken word recognition at the sentence level were investigated in one two-alternative forced-choice word identification experiment. In the experiment, an H* pitch-accented or a deaccented word fragment (e.g., AR- in the word archive) was presented at the end of a carrier sentence for identification. The results of the experiment revealed differences in the cue weighting of English lexical stress perception between native and non-native listeners. For native English listeners, vowel quality was a more prominent cue than f0, while native Mandarin Chinese listeners employed both vowel quality and f0 in a comparable fashion. These results suggested that (a) vowel reduction is superior to f0 in signaling initial stress in the words and (b) f0 facilitates the recognition of word initial stress, which is modulated by first language.