Text Item Research Articles

Using generative adversarial network to improve the accuracy of detecting AI-generated tweets.

This paper provides a novel approach using state-of-the-art generative Artificial Intelligence (AI) models to enhance the accuracy of machine learning methods in detecting AI-generated texts; the underlying generative capabilities are used along with ensemble-based learning methods for the exact characterization of created text attributes. Four basic steps are involved in the proposed methodology. The first step of the text process is the preprocessing stage itself consisting of several steps for the purification of irrelevant data. These stages include noise removal, text tokenization, removal of stop-words, word normalization, and handling uncommon words. In the next step, feature engineering and text representations are done whereby every preprocessed text is represented by a square matrix. This matrix encapsulates data about word correlations, cooccurrence, and word weights. The third step is Generative Adversarial Network (GAN)-based feature extraction, using a GAN model to extract efficient features in classifying the texts based on their creator type. After that, it turns the discriminator part into a strong feature extraction model. The fourth step is weighted Random Forest (RF)-based detection, with the features extracted by the discriminator of GAN serving as input to the RF-based detection model. This approach has covered the differences between texts generated by a human and that generated by Artificial Intelligence, with a significant improvement of 99.60% average accuracy, representing a 1.5% improvement against comparative methods.

Slovak morphological tokenizer using the Byte-Pair Encoding algorithm

This study introduces a new approach to text tokenization, SlovaK Morphological Tokenizer (SKMT), which integrates the morphology of the Slovak language into the training process using the Byte-Pair Encoding (BPE) algorithm. Unlike conventional tokenizers, SKMT focuses on preserving the integrity of word roots in individual tokens, crucial for maintaining lexical meaning. The methodology involves segmenting and extracting word roots from morphological dictionaries and databases, followed by corpus preprocessing and training SKMT alongside a traditional BPE tokenizer. Comparative evaluation against existing tokenizers demonstrates SKMT’s outstanding ability to maintain root integrity, achieving 99.7% root integrity compared to SlovakBERT (90.5%) and a pureBPE tokenizer (93.1%). Further validation involved fine-tuning models on a sentiment classification NLP task, where models trained with SKMT achieved an F1-score improvement of 3.5% over those trained with conventional BPE tokenization, followed by a focus on the Semantic Textual Similarity (STS) task. These findings suggest that training language models on the SKMT tokenizer significantly enhances model performance and quality.

Fine-grained semantic oriented embedding set alignment for text-based person search

Text-based person search aims to retrieve images of a person that are highly semantically relevant to a given textual description. The difficulty of this retrieval task is modality heterogeneity and fine-grained matching. Most existing methods only consider the alignment using global features, ignoring the fine-grained matching problem. The cross-modal attention interactions are popularly used for image patches and text markers for direct alignment. However, cross-modal attention may cause a huge overhead in the reasoning stage and cannot be applied in actual scenarios. In addition, it is unreasonable to perform patch-token alignment, since image patches and text tokens do not have complete semantic information. This paper proposes an Embedding Set Alignment (ESA) module for fine-grained alignment. The module can preserve fine-grained semantic information by merging token-level features into embedding sets. The ESA module benefits from pre-trained cross-modal large models, and it can be combined with the backbone non-intrusively and trained in an end-to-end manner. In addition, an Adaptive Semantic Margin (ASM) loss is designed to describe the alignment of embedding sets, instead of adapting a loss function with a fixed margin. Extensive experiments demonstrate that our proposed fine-grained semantic embedding set alignment method achieves state-of-the-art performance on three popular benchmark datasets, surpassing the previous best methods.

Semantic-aware frame-event fusion based pattern recognition via large vision–language models

Pattern recognition through the fusion of RGB frames and Event streams has emerged as a novel research area in recent years. Current methods typically employ backbone networks to individually extract the features of RGB frames and event streams, and subsequently fuse these features for pattern recognition. However, we posit that these methods may suffer from two key issues: (1). They attempt to directly learn a mapping from the input vision modality to the semantic labels. This approach often leads to sub-optimal results due to the disparity between the input and semantic labels; (2). They utilize small-scale backbone networks for the extraction of RGB and Event input features, thus these models fail to harness the recent performance advancements of large-scale visual-language models. In this study, we introduce a novel pattern recognition framework that consolidates the semantic labels, RGB frames, and event streams, leveraging pre-trained large-scale vision–language models. Specifically, given the input RGB frames, event streams, and all the predefined semantic labels, we employ a pre-trained large-scale vision model (CLIP vision encoder) to extract the RGB and event features. To handle the semantic labels, we initially convert them into language descriptions through prompt engineering and polish using ChatGPT, and then obtain the semantic features using the pre-trained large-scale language model (CLIP text encoder). Subsequently, we integrate the RGB/Event features and semantic features using multimodal Transformer networks. The resulting frame and event tokens are further amplified using self-attention layers. Concurrently, we propose to enhance the interactions between text tokens and RGB/Event tokens via cross-attention. Finally, we consolidate all three modalities using self-attention and feed-forward layers for recognition. Comprehensive experiments on the HARDVS and PokerEvent datasets fully substantiate the efficacy of our proposed SAFE model. The source code has been released at https://github.com/Event-AHU/SAFE_LargeVLM.

Field-Testing Multiple-Choice Questions With AI Examinees: English Grammar Items.

Field-testing is an essential yet often resource-intensive step in the development of high-quality educational assessments. I introduce an innovative method for field-testing newly written exam items by substituting human examinees with artificially intelligent (AI) examinees. The proposed approach is demonstrated using 466 four-option multiple-choice English grammar questions. Pre-trained transformer language models are fine-tuned based on the 2-parameter logistic (2PL) item response model to respond like human test-takers. Each AI examinee is associated with a latent ability θ, and the item text is used to predict response selection probabilities for each of the four response options. For the best modeling approach identified, the overall correlation between the true and predicted 2PL correct response probabilities was .82 (bias = 0.00, root mean squared error = 0.18). The study results were promising, showing that item response data generated from AI can be used to calculate item proportion correct, item discrimination, conduct item calibration with anchors, distractor analysis, dimensionality analysis, and latent trait scoring. However, the proposed approach did not achieve the level of accuracy obtainable with human examinee response data. If further refined, potential resource savings in transitioning from human to AI field-testing could be enormous. AI could shorten the field-testing timeline, prevent examinees from seeing low-quality field-test items in real exams, shorten test lengths, eliminate test security, item exposure, and sample size concerns, reduce overall cost, and help expand the item bank. Example Python code from this study is available on Github: https://github.com/hotakamaeda/ai_field_testing1.

Discrete diffusion models with Refined Language-Image Pre-trained representations for remote sensing image captioning

RS image captioning (RSIC) utilizes natural language to provide a description of image content, assisting in the comprehension of object properties and relationships. Nonetheless, RS images are characterized by variations in object scales, distributions, and quantities, which make it challenging to obtain global semantic information and object connections. To enhance the accuracy of captions produced from RS images, this paper proposes a novel method referred to as Discrete Diffusion Models with Refined Language-Image Pre-trained representations (DDM-RLIP), leveraging an advanced discrete diffusion model (DDM) for nosing and denoising text tokens. DDM-RLIP is based on an advanced DDM-based method designed for natural pictures. The primary approach for refining image representations involves fine-tuning a CLIP image encoder on RS images, followed by adapting the transformer with an additional attention module to focus on crucial image regions and relevant words. Furthermore, experiments were conducted on three datasets, Sydney-Captions, UCM-Captions, and NWPU-Captions, and the results demonstrated the superior performance of the proposed method compared to conventional autoregressive models. On the NWPU-Captions dataset, the CIDEr score improved from 116.4 to 197.7, further validating the efficacy and potential of DDM-RLIP. The implementation codes for our approach DDM-RLIP are available at https://github.com/Leng-bingo/DDM-RLIP.

A Web Semantic Mining Method for Fake Cybersecurity Threat Intelligence in Open Source Communities

In order to overcome the challenges of inadequate classification accuracy in existing fake cybersecurity threat intelligence mining methods and the lack of high-quality public datasets for training classification models, we propose a novel approach that significantly advances the field. We improved the attention mechanism and designed a generative adversarial network based on the improved attention mechanism to generate fake cybersecurity threat intelligence. Additionally, we refine text tokenization techniques and design a detection model to detect fake cybersecurity threats intelligence. Using our STIX-CTIs dataset, our method achieves a remarkable accuracy of 96.1%, outperforming current text classification models. Through the utilization of our generated fake cybersecurity threat intelligence, we successfully mimic data poisoning attacks within open-source communities. When paired with our detection model, this research not only improves detection accuracy but also provides a powerful tool for enhancing the security and integrity of open-source ecosystems.

Characteristics of opinions in the societal and non-societal domains

With the increasing availability of user opinions on the web, understanding the distinct nature of opinions in societal and non-societal contexts becomes crucial for opinion mining and sentiment analysis tasks. Societal topics, encompassing social unrest, terrorist acts, and government policies, differ significantly from non-societal topics like product reviews, movie reviews, and restaurant reviews. Given the regional specificity of societal issues and the lack of sentiment-annotated resources for them, this paper highlights the need to comprehend the differences in opinions between these domains for effective sentiment analysis. Through statistical text and network analysis, it investigates word usage, sentiment word association, and homogeneity in societal versus non-societal contexts. The study also explores graph-based analysis as a novel approach to sentiment analysis, considering its advantage in easily expanding context through the addition of nodes, as opposed to the complexity of inserting relevant tokens in text. The findings suggest that while non-societal sentiment resources might not be directly applicable to societal domains, graph-based analysis offers promising avenues for sentiment analysis in diverse societal topics.

Multi-Level Attention with 2D Table-Filling for Joint Entity-Relation Extraction

Joint entity-relation extraction is a fundamental task in the construction of large-scale knowledge graphs. This task relies not only on the semantics of the text span but also on its intricate connections, including classification and structural details that most previous models overlook. In this paper, we propose the incorporation of this information into the learning process. Specifically, we design a novel two-dimensional word-pair tagging method to define the task of entity and relation extraction. This allows type markers to focus on text tokens, gathering information for their corresponding spans. Additionally, we introduce a multi-level attention neural network to enhance its capacity to perceive structure-aware features. Our experiments show that our approach can overcome the limitations of earlier tagging methods and yield more accurate results. We evaluate our model using three different datasets: SciERC, ADE, and CoNLL04. Our model demonstrates competitive performance compared to the state-of-the-art, surpassing other approaches across the majority of evaluated metrics.

MedPodGPT: A multilingual audio-augmented large language model for medical research and education.

The proliferation of medical podcasts has generated an extensive repository of audio content, rich in specialized terminology, diverse medical topics, and expert dialogues. Here we introduce a computational framework designed to enhance large language models (LLMs) by leveraging the informational content of publicly accessible medical podcast data. This dataset, comprising over 4, 300 hours of audio content, was transcribed to generate over 39 million text tokens. Our model, MedPodGPT, integrates the varied di-alogue found in medical podcasts to improve understanding of natural language nuances, cultural contexts, and medical knowledge. Evaluated across multiple benchmarks, MedPodGPT demonstrated an average improvement of 2.31% over standard open-source benchmarks and showcased an improvement of 2.58% in its zero-shot multilingual transfer ability, effectively generalizing to different linguistic contexts. By harnessing the untapped potential of podcast content, MedPodGPT advances natural language processing, offering enhanced capabilities for various applications in medical research and education.

Team talk and the evaluation of medical guidance documentation

This article looks at team talk in a validation committee meeting assessment of a guidance document text item. The item assessment was not evidence-based in terms of Evidence Based Medicine (EBM) criteria; instead, the item was assessed via the committee members present drawing on their clinical practitioner members’ knowledge and professional experience. Analysis of the meeting reveals such apparently ‘mere opinion’ to be a systematic evaluation of professional knowledge and personal experiences, in ways ‘compatible’ with thought experiments. Thought experiments are argued to be a members’ resource as well as an analyst’s one, although their detailed occasionedness is not reducible to a constructivist formalisation. The article’s approach is informed by ethnomethodology and conversation analysis, and while the use of thought experiments as a heuristic device in the analysis is controversial, a warrant for this is attempted. The research was undertaken to locate ways of understanding and supporting team members’ work of robust and useful guidance content production. ‘Validating’ guidance is shown in-and-as the emergent collaborative work of the committee members themselves.

Gender differences in literacy in PIAAC: do assessment features matter?

BackgroundPrevious research based on large-scale studies consistently suggests that on average, male students tend to have lower literacy compared to their female students during secondary schooling. However, this gender gap in literacy seems to “disappear” during adulthood. Up until today, only a few studies investigated the role of assessment features in gender differences in literacy performance in adulthood. This study aims to understand the relationship between assessment features and gender differences in literacy skills.MethodsUsing the German 2012 PIAAC data (N = 4,512), we applied item-level analyses using linear probability models to examine gender differences in the probability of solving a literacy item correctly with six assessment features including (1) text format, (2) text topics, (3) text length, (4) cognitive strategies, (5) numerical content of the text/questions, and (6) gender typicality of content.ResultsWe found that men had a 13.4% higher probability of solving items with a noncontinuous text format correctly than women. Men also had a 9.4% higher probability of solving short text items correctly and a 4.6% higher probability of solving items with a medium/high numerical content in the question correctly than women. There were small to negligible gender differences in literacy performance in terms of text topics, cognitive strategies, and gender typicality of content.ConclusionsOur findings highlight the role of text format, text length, and numerical content in gender differences in literacy skills, suggesting that further refining these practices can enhance the fairness and accuracy of literacy assessments. Specifically, we advocate for ongoing research aimed at understanding and minimizing the potential bias introduced by these assessment features. Such efforts are not only crucial for developing instruments that accurately measure literacy skills, but they also yield insights that hold significant implications for educational researchers and practitioners dedicated to creating more equitable assessment environments.

The Revised Sexual Experiences Survey Victimization Version (SES-V): Conceptualization, Modifications, Items and Scoring

ABSTRACT The Sexual Experiences Survey [SES] is considered the gold standard measure of non-consensual sexual experiences. This article introduces a new victimization version [SES-V] developed by a multidisciplinary collaboration, the first revision since 2007. The 2024 SES-V is designed to measure the construct of sexual exploitation since the 14th birthday. Notable revisions are adoption of a freely given permission standard for non-consent, introduction of new tactics and acts, including made to perform or to penetrate another person’s body, tactics-first wording order, and emphasis on gender and sexual orientation inclusivity. The SES-V is modularized to allow whole or partial administration. Modules include Non-contact, Technology-facilitated, Illegal (largely penetrative), and Verbally pressured sexual exploitation. Tables provide item text, multiple scoring approaches, module follow-up, specific incident description and demographics. Future plans include developing a scoring algorithm based on weighting our hypothesized dimensions of sexual exploitation severity: invasiveness, pressure, and norm violation combined with frequency. This article is the first in a special issue on the SES-V. Subsequent articles focus on the taxonomies and literature that informed each module. The issue concludes with two empirical papers demonstrating the feasibility and validity of the SES-V: (1) psychometric comparison with the 2007 SES-SFV; and (2) prevalence data from a census-matched adult community sample.

Sharing knowledge without borders: Collaboration between U.S. business and law students and Spanish translation students

Collaboration between technical and professional (TP) writers in English as a second language is important for successful communication. This is evident in legal documents in both English and Spanish, where accuracy and clarity are paramount. However, achieving this is challenging because law students often lack effective writing tools, and translation students may lack the required encyclopedic knowledge. The present paper describes a co-designed exchange project between U.S. students in business and law and students of legal translation at a university in Spain. This paper aims to determine potential improvements in students’ encyclopedic and conceptual knowledge of this specialized language and their translation performance after the collaboration intervention. To evaluate these aspects, three textual items were targeted: terminological, nominal and, what we call, substitutional density. Our results indicate that peer collaboration helped Spanish students acquire encyclopedic knowledge in the field and avoid repetitive terms in their translations, leading to greater textual cohesion through the use of general language, synonyms and hypernyms. Other aspects of this peer collaboration as well as follow-up approaches to consolidate the observed effects are needed.

The Students’ Difficulties in Identifying the Five Verb Forms on the News Item Text at NHU Pematangsiantar

The way people communicate is communication. Communication needed meaning. The meaning which contained of the sentence has the differences of meaning form. The problem of this research is the difficulties of the students’ in identifying the five verb forms in English. The aimed of this research is identifying the findings of the students’ errors in stating which formed contained of it. It can be Base (Regular): (To) laugh, Past tense: Laughed, Past Participle: Laughed, Present Participle: Laughing and Third-Person Singular: Laughs. The method applied qualitative research. The subject is the students in the third semester. The instrument of the data is a written test. The researcher made steps how to analyse the data namely: Counted the vocabulary which contained of the verb forms, Classified it by making them a code into their groups, Founded the variant of the difficulties in identifying verb forms that written by the students mostly, and described it. The student’s difficulties in identifying the five verb forms on the text entitled “Hundreds of thousands affected by floods, landslides as heaviest rain in 60 years hits southern China” got the highest one is very good category with 48 students with 81,3%, then followed by the good category with 5 students on the 8,4%, 2 students with low category namely 3,3%, and the last fair category with 4 students on the 6,7%. It found the dominantly applied by the students in identifying the difficulties is stated to the Nouns became Verbs, The Preposition became Verbs, and the most of the students’ difficulties in differentiate the correct verb forms, if it’s a V2 (Preterit) or V3 (Past Participle) form

Improving EFL Students’ Writing Skills Through News Video-Based Learning Strategy In News Item Text

The present research aimed to discover improvements in EFL students’ writing skills ability through a news video-based learning strategy in news item text. The researcher applied a quantitative method with a one-group pretest-posttest classroom research design (pre-experimental). The data obtained in this research were calculated using the manual statistical method and SPSS statistics version 27. The data were analyzed using the statistical formula of paired sample t-test computation. The population of the research was all of the students in the Acceleration class of MAN 2 Kota Makassar, where the participants consisted of 14 students as well as the sample of the research. Based on the mean score of the students’ pretest was 78 in average to good categories while the mean score of the students’ posttest was 84 in good to excellent categories. Moreover, the result of the data was supported by the t-test analysis, in which the t-test value was higher than the t-table value (-3,113>2,160). So, it can be summarized that H0 was rejected and Ha was accepted, which means a News Video–based Learning Strategy through YouTube of VOA Learning English effectively increases EFL students’ writing skills while composing a news item text paragraph.

An Analysis of Grammatical Equivalence Used by Students in Translating News Item Text

ABSTRACT Spreading news in this era seems to be easier due to the rapid development of technology. However, some readers from different language background seem to have difficulties in understanding the news because of language barrier. Translation is needed to overcome this problem and to make news understandable for readers in greater community. This research examined the types of grammatical equivalence used by the students in translating news item text from Indonesian to English. Besides, this research also aimed to figure out the problems found in the students’ grammatical equivalence. Descriptive qualitative approach was used in this research since the data were presented in the form of description instead of number. The primary data were taken from news item text translation works done by the seventh-semester students of English Department of Pamulang University in academic year of 2022/2023. The data were collected using Google Form and then analysed using the grammatical equivalence theory proposed by Baker (2001). There were five steps of data analysis used in this research: observing the data, identifying the data, classifying the data, and evaluating the data. The results showed that the students used five types of grammatical equivalence namely number, gender, person, tenses & aspects, and voice. The results also revealed that the students faced problems related to number and tenses & aspects.

AN ANALYSIS OF LEXICAL DENSITY ON READING MATERIALS IN ENGLISH INTERLANGUAGE TEXTBOOK FOR THE TENTH-GRADE SENIOR HIGH SCHOOL

Reading is one of the important language skills students have to master. Some forms of reading materials are written in English, to mention some are books, articles, and textbooks. English Interlanguage Textbook for the Tenth-Grade Senior High School was one of the reading materials chosen as the focus of this research. In terms of the content of the textbook, the teacher should be able to decide whether it is informative or not. The objectives of this study were (1) to find out the lexical density of the reading materials in the English Interlanguage textbook, and (2) to describe the lexical density of the reading materials in the English Interlanguage Textbook. In this study, the researcher chose qualitative research as the methodology. The data collected from the textbook showed that there were 9 reading materials that were categorized into different kinds of text. Those were narrative text, recount text, descriptive text, and news item text. The lexical density of reading material in the English Interlanguage Textbook for the tenth-grade of Senior High School was around 40% - 60%. It infers that the lexical density was sufficient. Also, the reading materials were clear, understandable, and suitable for students.Keywords: Descriptive research, lexical density, textbook analysis

PENGEMBANGAN MULTIMEDIA PEMBELAJARAN BAHASA INGGRIS UNTUK SISWA KELAS X MENGENAI TEKS NEWS ITEM

This article is based on the development research that aims to develop multimedia of learning English which is attractive and effective for grade X Students about news item text. The development of the multimedia for learning English uses Lee and Owens model. The procedure of developing the multimedia covers four steps. They are: 1) need analysis and assessment, 2) designing multimedia learning, 3) developing and implementing the multimedia, and 4) evaluating the multimedia. The data of this research are compiled any sources i.e. validation result of leaning design expert, subject matter expert and learning media expert. It is also from students as treatment subjects and English lesson teacher. The treatment subjects are 39 students consisting of 3 students for personal treatment, 5 students for small group treatment and 31 students for field treatment. The results of the research show that: 1) developing the Macromedia Flash based multimedia of English learning for Class X was assessed by validators as learning media which is in good category and it is suitable to use. 2) the treatment subjects, students and English teacher gave response and conclusion that the multimedia is attractive for learning English, shown from the result of personal treatment subjects’ questionnaire of 88.7%, the result of small group treatment subjects’ questionnaire of 64.4% and the result of field treatment subjects’ questionnaire of 84.2%, therefore, the average of total treatment subjects’ questionnaire is 79.7%, then 3) this multimedia can be used effectively to achieve learning competencies, proven by the result of increasing score gained by the students in pre-test and post-test which is 72.03%. Keyword: development research, multimedia of learning English, news item text

SENTIMENTAL ANALYSIS ON TOURISM REVIEWS

Sentiment analysis plays a pivotal role in understanding the sentiments and opinions expressed in textual data, offering valuable insights into various domains, including tourism. In this study, we present a comprehensive review of sentiment analysis techniques applied to tourism reviews using machine learning algorithms. The abundance of user-generated content on tourism platforms has made sentiment analysis an indispensable tool for businesses and researchers alike. By leveraging machine learning algorithms, researchers can extract sentiments from vast amounts of textual data efficiently and accurately. This review outlines the key methodologies and approaches utilized in sentiment analysis of tourism reviews. It discusses preprocessing techniques such as text tokenization, stop-word removal, and stemming, which are crucial for preparing textual data for analysis. Furthermore, it examines various machine learning algorithms employed for sentiment classification, including Naive Bayes, Support Vector Machines, and Recurrent Neural Networks. Additionally, the review delves into feature extraction methods such as bag-of-words, TF-IDF, and word embeddings, highlighting their impact on sentiment analysis accuracy. Moreover, it explores the challenges and limitations associated with sentiment analysis in the tourism domain, such as sarcasm detection and language nuances.