Short Text Research Articles

Lost in translation: error analysis of texts translated from Arabic into English by Saudi translators

Biographic translation is crucial in today’s globalized world to preserve histories or shed light on the lives of people who influenced their nations or the world. The literature on biographic translation and the factors hindering such translation in an English as a foreign language (EFL) setting, especially in the Saudi context, is relatively limited. In the present study, the author analyzed the errors made by Saudi EFL learners when they translated a biography from Arabic into English. The data were drawn from a short text translated by female final-year university students as part of their course evaluations. The author conducted a qualitative analysis and classified errors according to Pospescu’s taxonomy. The results showed that the students committed all the three error types: linguistic, comprehensive, and translation. Within these three types, the students made diverse errors, ranging from misusing tenses to misspelling words. Consequently, their overall translations were unsatisfactory. This finding is helpful for both learners and educators, as the identified translation errors can be used as teaching aids in translation classes. The findings put emphasis on the importance of practice and prompt feedback as essential aids educators can use to support their students during their translation learning.

Limitations of Feature Attribution in Long Text Classification of Standards

Managing complex AI systems requires insight into a model's decision-making processes. Understanding how these systems arrive at their conclusions is essential for ensuring reliability. In the field of explainable natural language processing, many approaches have been developed and evaluated. However, experimental analysis of explainability for text classification has been largely constrained to short text and binary classification. In this applied work, we study explainability for a real-world task where the goal is to assess the technological suitability of standards. This prototypical use case is characterized by large documents, technical language, and a multi-label setting, making it a complex modeling challenge. We provide an analysis of approx. 1000 documents with human-annotated evidence. We then present experimental results with two explanation methods evaluating plausibility and runtime of explanations. We find that the average runtime for explanation generation is at least 5 minutes and that the model explanations do not overlap with the ground truth. These findings reveal limitations of current explanation methods. In a detailed discussion, we identify possible reasons and how to address them on three different dimensions: task, model and explanation method. We conclude with risks and recommendations for the use of feature attribution methods in similar settings.

Anomaly-aware symmetric non-negative matrix factorization for short text clustering

Anomaly-aware symmetric non-negative matrix factorization for short text clustering

Contribution of Metacognitive Questions to Accuracy of Judgment of Learning in a Digital Environment

Primary school pupils have difficulty making accurate judgments of learning (JOL) in a digital environment. In two studies, we examined the contribution of metacognitive self-questioning to JOL accuracy in the course of reading literacy (Study 1) and mathematical literacy (Study 2) digital learning programs. Both studies comprised a six-session intervention. In each session fifth-grade students studied a short text (Study 1, N=65) or story containing mathematical components (Study 2, N = 72), then judged their comprehension by predicting their performance before completing a test. For the experimental groups, metacognitive questioning was incorporated into the study materials. In both studies the experimental groups improved over the six sessions in both performance and calibration (i.e., reducing the gap between judgment and performance). The findings highlight the contribution of metacognitive support to improved performance and judgment accuracy and strengthen the case for incorporating metacognitive practices when teaching literacy in a digital environment.

Pengaruh Model Project Based Learning (PjBL) Berbantuan Teknik 3M dalam Pembelajaran Menulis Teks Cerpen Siswa Kelas X SMA Negeri 1 Batang Anai Kabupaten Padang Pariaman

The aim of this research is to describe the following things. First, to describe the learning to write short story texts for class X SMA Negeri 1 Batang Anai, Padang Pariaman Regency before using the Project Based Learning (PjBL) model assisted by 3M techniques. Second, to describe the learning to write short story texts for class X students of SMA Negeri 1 Batang Anai, Padang Pariaman Regency after using the Project Based Learning (PjBL) model assisted by 3M techniques. Third, to describe the influence of the Project Based Learning (PjBL) model assisted by 3M techniques in learning to write short story texts for class X students at SMA Negeri 1 Batang Anai, Padang Pariaman Regency. There are three results of this research. First, the learning to write short story texts for class Second, the learning to write short story texts for class Third, based on the results of the t-test, it was concluded that there was a significant influence on the use of the Project Based Learning (PjBL) model assisted by 3M techniques in learning to write short story texts for class ) at the α significance level of 0.05 and a confidence level of 95%. This shows that there is a significant influence in the use of Project Based Learning (PjBL) assisted by 3M techniques in learning to write short story texts for class X SMA Negeri 1 Batang Anai Padang Pariaman Regency.

Effects of Varying Text Message Length and Driving Speed on the Disruptive Effects of Texting on Driving Simulator Performance: Differential Effects on Eye Glance Measures

Eye glance analysis and driving performance during texting while driving: Differential effects of varying driving speed versus text message length. Background and Objective. Texting while driving continues to be a significant public health concern. Eye glances off the roadway are a measure of the visual distraction associated with texting while driving. In the present study, we examined the effects of two ‘real-world’ factors relating to the adverse effects of texting on driving performance and eye glances off the roadway: (1) text message length and (2) driving speed. Methods. Subjects ‘drove’ a fixed-base simulator and read, typed and sent text messages while driving. In study #1, the driving speed was 60 mph and the effects of short (1 word) versus longer (8–10 words) texts were compared. In study #2, the text messages were short only and driving speed was 60 or 80 mph. Driving performance was assessed using the Standard Deviation of Lane Position (SDLP). Video recordings of the drivers’ faces were used to assess eye glances from the road to the phone—and back—during texting. Results. Texting while driving impaired driving performance as measured by SDLP, and both longer text messages and faster drive speeds made driving performance even worse. Analysis of the eye glance data, however, revealed different effects of these two manipulations. Specifically, longer text messages were associated with an increase in the number of eye glances to the phone during a text message episode, an increase in the total time spent with the eyes off the road, and an increase in the single longest eye glance from the road. Moreover, with longer text messages the longest single eye glance away from the road typically occurred at or near the end of the text message episode. In contrast, increasing driving speed to 80 mph did not affect any of these eye glance measures relative to driving at 60 mph. Conclusion and Application. Both text message length and driving speed while texting adversely affect driving performance, but they do so via different mechanisms. These results have implications for how to tailor “don’t text and drive” messaging to better serve the public health.

Development of Digital Teaching Materials for Short Story Texts Based on Local Wisdom in Middle Schools (Phase D)

Education must integrate local wisdom to prevent cultural values ​​from fading in the era of globalization. The Merdeka Curriculum supports this with the flexibility to emphasize character education and independence. Digital teaching materials based on local wisdom, such as short story texts, can improve students' reading, writing and cultural understanding skills. The aim of this research is to develop and test digital teaching materials for short story texts based on local wisdom for junior high school students. This research used the ADDIE method with the help of 2 media and language validators as well as 6 students and 3 teachers in testing teaching materials. Data collection techniques were carried out using questionnaires, interviews and documentation. The teaching material design process begins with an analysis of student needs and the curriculum, followed by creating a storyboard and validation by two validators who state that the teaching material is very worthy of being tested. The implementation of local wisdom-based teaching materials shows high effectiveness, with student responses reaching 90% and teacher responses 87%, both in the very good category.

Learning-based short text compression using BERT models

Learning-based data compression methods have gained significant attention in recent years. Although these methods achieve higher compression ratios compared to traditional techniques, their slow processing times make them less suitable for compressing large datasets, and they are generally more effective for short texts rather than longer ones. In this study, MLMCompress, a word-based text compression method that can utilize any BERT masked language model is introduced. The performance of MLMCompress is evaluated using four BERT models: two large models and two smaller models referred to as “tiny”. The large models are used without training, while the smaller models are fine-tuned. The results indicate that MLMCompress, when using the best-performing model, achieved 3838% higher compression ratios for English text and 42% higher compression ratios for multilingual text compared to NNCP, another learning-based method. Although the method does not yield better results than GPTZip, which has been developed in recent years, it achieves comparable outcomes while being up to 35 times faster in the worst-case scenario. Additionally, it demonstrated a 20% improvement in compression speed and a 180% improvement in decompression speed in the best case. Furthermore, MLMCompress outperforms traditional compression methods like Gzip and specialized short text compression methods such as Smaz and Shoco, particularly in compressing short texts, even when using smaller models.

1QIsaa, OG, and MT Isaiah 63:3

Abstract The substantial textual variation among three major witnesses to Isa 63:3 (1QIsaa, OG, and MT) has received relatively little attention in the commentary literature and in the scholarly journals and monographs. These witnesses attest to a short text (1QIsaa), a long text (OG), and an even longer text (MT). The present article proposes that the OG Vorlage is an expansion of a pre-1QIsaa text, while the MT is an expansion of 1QIsaa.

Application of the Naïve Bayes Classifier Algorithm to Analyze Sentiment for the Covid-19 Vaccine on Twitter in Jakarta

The epidemic of a new disease caused by the coronavirus (2019-nCoV), commonly referred to asCOVID-19, has been declared a global virus epidemic by the World Health Organization (WHO).President Joko Widodo has officially ratified Presidential Decree No. 99 of 2020 concerning theprovision of vaccines and the implementation of vaccination activities. Twitter is a social mediaplatform that allows users to share information and opinions directly with fellow users. Tweets givencan be in any form, either positively or negatively, so one of the methods used is sentiment analysis.Sentiment analysis helps determine an opinion or comment on an issue, whether the response ispositive or negative. The Naïve Bayes algorithm is used in sentiment analysis because it is suitablefor tweets or text data that is not too long or short text. The initial stage of sentiment analysis is textpre-processing which consists of Cleaning, case folding, tokenizing, and stopword removal. Then thedata is labeled manually. The analysis results are visualized as bar charts, pie charts, and word clouds.Then the word weighting is carried out using the term frequency-inverse document (TF-IDF), andclassification is carried out using the Naïve Bayes classifier. From the test results, the accuracy valueof the confusion matrix is 82% from 2600 tweet data with 80% training data composition and 20%test data.

Implementation of Think Pair and Share (TPS) Technique in Listening to Short Story Texts Learning

This research aims to determine whether there are differences in students' abilities in listening to short story texts using the Think-Pair-Share (TTPS) cooperative learning method. Conducted as a quasi-experimental study, the sample comprised 11th-grade students from SMA DDI Maros. Data collection involved text analysis techniques, and the obtained data were analyzed through stages including the creation of raw score lists, frequency distributions, descriptive statistical analysis, and inferential statistical analysis. The results indicate significant differences in learning outcomes between students using the TTPS method and those not using it. The use of TTPS improved students' listening comprehension of short story texts, as evidenced by the mean score of 74.36 (high category) in the experimental class compared to 58.46 (low category) in the control class. This demonstrates that the TTPS method effectively enhances students' listening skills.

Application of Short Text Classification Model Based on GPT-3.5 in E-Commerce

In the realm of e-commerce, the crucial role of short text classification in enhancing user experience and platform operational efficiency has led to the development of effective. approaches encompassing feature-based algorithms and early deep learning models. However, challenges remain when dealing with complex semantics and vast amounts of unstructured data. To bridge this gap, This paper proposes a novel model of short text classification based on GPT-3.5, which significantly improves classification accuracy by combining conceptualization and character-level information. Our model's superior performence across all datasets was validated through experimentsal trials on five widely used datasets (TREC, AG News, Bing, and Movie Review).

Hazard Analysis for Massive Civil Aviation Safety Oversight Reports Using Text Classification and Topic Modeling

There are massive amounts of civil aviation safety oversight reports collected each year in the civil aviation of China. The narrative texts of these reports are typically short texts, recording the abnormal events detected during the safety oversight process. In the construction of an intelligent civil aviation safety oversight system, the automatic classification of safety oversight texts is a key and fundamental task. However, all safety oversight reports are currently analyzed and classified into categories by manual work, which is time consuming and labor intensive. In recent years, pre-trained language models have been applied to various text mining tasks and have proven to be effective. The aim of this paper is to apply text classification to the mining of these narrative texts and to show that text classification technology can be a critical element of the aviation safety oversight report analysis. In this paper, we propose a novel method for the classification of narrative texts in safety oversight reports. Through extensive experiments, we validated the effectiveness of all the proposed components. The experimental results demonstrate that our method outperforms existing methods on the self-built civil aviation safety oversight dataset. This study undertakes a thorough examination of the precision and associated outcomes of the dataset, thereby establishing a solid basis for furnishing valuable insights to enhance data quality and optimize information.

Authorship Attribution for English Short Texts

Authorship Attribution for English Short Texts

Heterogeneous graph contrastive learning with adaptive data augmentation for semi‐supervised short text classification

AbstractShort text classification has been widely used in many fields. Due to the scarcity of labelled data, implementing short text classification under semi‐supervised learning setting has become increasingly popular. Semi‐supervised short text classification methods based on graph neural networks can achieve state‐of‐the‐art classification performance by utilizing the expressive power of graph neural networks. However, these methods usually fail to mine the hidden patterns of a large amount of short text node data in the graph to optimize the short text node embeddings, which limits the semantic representation power of the short texts, thus leading to suboptimal classification performance. To overcome the limitation, this paper proposes a novel semi‐supervised short text classification method called the Heterogeneous Graph Contrastive Learning with Adaptive Data Augmentation (HGCLADA). In the knowledge bases guided soft prompt‐based data augmentation component, the related words of the tag words are used to optimize the soft prompts for generating diverse augmented samples. In the heterogeneous graph contrastive learning framework component, a heterogeneous graph that is constructed using short texts and keywords and an effective edge augmentation scheme based on a short text clustering algorithm are proposed. The optimized short text embeddings can be obtained to achieve the effective semi‐supervised short text classification. Extensive experiments on six benchmark datasets show that our HGCLADA method outperforms four classes of state‐of‐the‐art methods in terms of classification accuracy, especially with significant performance improvements of 8.74% on the TagMyNews dataset when each class only contains 20 labelled data.

Impact of an SMS intervention to support type 2 diabetes self-management: DiabeText clinical trial.

Complications arising from uncontrolled Type 2 Diabetes Mellitus (T2DM) pose a significant burden on individuals' well-being and healthcare resources. Digital interventions may play a key role in mitigating such complications by supporting patients to adequately self-manage their condition. To assess the impact of DiabeText, a new theory-based, patient-centered, mobile health intervention integrated with electronic health records to send tailored short text messages to support T2DM self-management. Pragmatic, Phase III, 12-month, two-arm randomized clinical trial with T2DM primary care patients in Spain. 742 participants with suboptimal glycemic control (HbA1c>7.5) were randomly allocated to a control (usual care) or intervention (DiabeText) group. The DiabeText group received, in addition to usual care, 165 messages focused on healthy lifestyle and medication adherence. glycated hemoglobin (HbA1c). medication possession ratio, quality of life (EQ-5D-5L), diabetes self-efficacy (DSES); and self-reported adherence to medication, Mediterranean diet (MEDAS-14), and physical activity (IPAQ). Over the 12-month period, we observed no significant differences in HbA1c between the intervention and the control groups (Beta=-0.025 (-0.198 to 0.147; p=0.772)). In comparison with the control group, the DiabeText group showed significant (p<0.05) improvements in self-reported medication adherence (OR=1.4; 95%CI: 1.0 to 1.9), DSES (Cohen's d=0.35), and EQ5D-5L (Cohen's d=0.18) scores; but not in the rest of secondary outcomes. DiabeText successfully improved quality of life, diabetes self-management, and self-reported medication adherence in primary care patients with T2DM. Further research is needed to enhance its effects on physiological outcomes.

Extracting Sentence Embeddings from Pretrained Transformer Models

Pre-trained transformer models shine in many natural language processing tasks and therefore are expected to bear the representation of the input sentence or text meaning. These sentence-level embeddings are also important in retrieval-augmented generation. But do commonly used plain averaging or prompt templates sufficiently capture and represent the underlying meaning? After providing a comprehensive review of existing sentence embedding extraction and refinement methods, we thoroughly test different combinations and our original extensions of the most promising ones on pretrained models. Namely, given 110 M parameters, BERT’s hidden representations from multiple layers, and many tokens, we try diverse ways to extract optimal sentence embeddings. We test various token aggregation and representation post-processing techniques. We also test multiple ways of using a general Wikitext dataset to complement BERT’s sentence embeddings. All methods are tested on eight Semantic Textual Similarity (STS), six short text clustering, and twelve classification tasks. We also evaluate our representation-shaping techniques on other static models, including random token representations. Proposed representation extraction methods improve the performance on STS and clustering tasks for all models considered. Very high improvements for static token-based models, especially random embeddings for STS tasks, almost reach the performance of BERT-derived representations. Our work shows that the representation-shaping techniques significantly improve sentence embeddings extracted from BERT-based and simple baseline models.

An Emotional Dialogue System Using Conditional Generative Adversarial Networks with a Sequence-to-Sequence Transformer Encoder

Understanding the expression of emotion and generating appropriate responses are key steps toward constructing emotional, conversational agents. In this paper, we propose a framework for single-turn emotional conversation generation, and there are three main components in our model, namely, a sequence-to-sequence model with stacked encoders, a conditional variational autoencoder, and conditional generative adversarial networks. For the sequence-to-sequence model with stacked encoders, we designed a two-layer encoder by combining Transformer with gated recurrent units-based neural networks. Because of the flexibility of the sequence-to-sequence model, we adopted a conditional variational autoencoder in our framework, which uses latent variables to learn a distribution over potential responses and generates diverse responses. Furthermore, we regard a conditional variational autoencoder-based, sequence-to-sequence model as the generative model, and the training of the generative model is assisted by both a content discriminator and an emotion classifier, which assists our model in promoting content information and emotion expression. We use automated evaluation and human evaluation to evaluate our model and baselines on the NTCIR short text conversation task (STC-3) Chinese emotional conversation generation (CECG) Subtask dataset [44], and the experimental results demonstrate that our proposed framework can generate semantically reasonable and emotionally appropriate responses.

Manasheer al-Intifada

Abstract Communiqués were central to the coordination of the mass popular uprising that challenged Israeli rule over Palestinians from 1987 until the early 1990s. These short political texts were called manasheer or bayanat al-Intifada, in Arabic. This essay demonstrates a methodology for engaging with the body of intifada bayanat as a source by focusing on the first of these serialized communiqués, which was distributed early in the first Intifada on January 8, 1988.

Medical Data NER and Classification Using Hybridized BERT Model

The extraction of important information from medical texts by Named Entity Recognition (NER) is a key component of advanced medical text processing. Medical practitioners rely heavily on NER's assistance with disease surveillance, clinical resolution building, and substantiation-based treatment. As the foundation of text information processing in the medical field, it guarantees precise location of data required for knowledgeable medical decisions and attentive disease surveillance. Additionally, a core goal in medical Natural Language Processing (NLP) is medical text categorization, which tries to classify short medical texts into distinct groups. Most recent work has concentrated on using pre-trained linguistic processes for text cataloging in medicine. The present work presents a novel clinical neural network architecture (NER) method that was created with a customized Rule Based BiLSTM-BERT (Bidirectional Encoder Representations from Transformers) architecture that incorporates Retrieval Augment Generation. Across several fields, including medicine, deep learning has demonstrated noteworthy advancements. These results show that, when applied to our test dataset, the BiLSTM-BERT-RAG model produced results that were almost human-like. The system proficiently recognized pertinent vocabulary representative of the intended protocol.