Aspect-Term Sentiment Analysis (ATSA) in multi-aspect sentences faces a fundamental tradeoff between efficiency and expressiveness.Existing models either re-encode the sentence for each aspect or rely on static use of deep representations, leading to redundant computation and limited adaptivity.We argue that Transformer depth is a costly, queryable resource, and propose DABS, a single-pass inference framework that encodes each sentence once to construct a reusable, depth-ordered substrate.Each aspect then queries this shared representation to selectively read relevant tokens and abstraction levels, without re-encoding.This decouples shared sentence encoding from lightweight, aspect-conditioned readout.Experiments on four ATSA benchmarks show that DABS achieves competitive performance while reducing end-to-end computation by up to 60% in multi-aspect settings (M ≥ 2).Further analyses indicate that adaptive depth querying is most beneficial for linguistically complex cases such as negation and contrast.

This study investigates how visual design elements—specifically color, lighting, and spatial composition—affect emotional and empathetic engagement in Virtual Reality (VR) environments. It focuses on adult users across diverse domains such as education, psychotherapy, and humanitarian advocacy. A systematic review of literature from 2018 to 2023 was conducted using Scopus, Web of Science, and JSTOR, adhering to PRISMA guidelines. A total of 21 peer-reviewed studies were selected based on their focus on visual modalities in VR and empathy-related outcomes for adult users. The review identifies a consistent impact of visual elements on users’ empathetic and emotional responses in immersive VR contexts. However, it also uncovers key challenges: (1) a lack of standardized, validated metrics for measuring empathy in VR; (2) a fragmented approach to visual emotion analysis that isolates design elements rather than considering their interplay; and (3) limited investigation into cultural variability in interpreting visual stimuli. This study offers a novel contribution by synthesizing visual design practices in VR to foster empathy, a focus not previously addressed in our research. Unlike existing studies, including any prior work by our team, this review integrates cross-cultural perspectives, proposes standardized empathy metrics, and examines the interplay of visual elements (color, lighting, spatial composition) across diverse domains, providing a cohesive framework for empathy-driven VR design.

Background : A hadith refers to sayings, actions, and characteristics of the Prophet Muhammad peace be upon him. The authenticity of hadiths is crucial, because they constitute the source of legislation for Muslims with the Holy Quran. Classifying hadiths into groups is a matter of importance as well, to make them easy to search and recognize. Objective : To report the results of a systematic review concerning hadith authentication and classification methods. Data sources : Original articles found in ACM, IEEE Xplore, ScienceDirect, Scopus, Web of Science, Springer Link, and Wiley Online Library. Study selection criteria : Only original articles written in English and dealing with hadith authentication and classification. Reviews, editorial, letters, grey literature, and restricted or incomplete articles are excluded. Data extraction : Two authors were assigned to extract data using a predefined data extraction form to answer research questions and assess studies quality. Results : A total of 27 studies were included in this review. There are 14 studies in authentication and 13 studies in classification. Most of the selected studies (17 of 27) were published in conferences, while the others (10 of 27) were published in scientific journals. Research in the area of hadith authentication and classification has received more attention in recent years (2016–2019). Conclusions : Hadith authentication methods are classified into machine learning, rule-based, and a hybrid of rule-based and machine learning and rule-based and statistical methods. Hadith classification methods are classified into machine learning and rule-based. All classification studies used Matn, while the majority of authentication studies used isnad. As a dataset source, Sahih Al-Bukhari was used by most studies. None of the used datasets is publicly available as a benchmark dataset, either in hadith authentication or classification. Recall and Precision are the most frequent evaluation metrics used by the selected studies.