Search Results (19)

(1) Background and objectives: Large language models (LLMs) such as GPT, Mistral, and LLaMA exhibit strong capabilities in text generation, yet assessing the quality of their reasoning—particularly in open-ended and argumentative contexts—remains a persistent challenge. This study introduces Dialectical Agent, an internally developed modular framework designed to evaluate reasoning through a structured three-stage process: opinion, counterargument, and synthesis. The framework enables transparent and comparative analysis of how different LLMs handle dialectical reasoning. (2) Methods: Each stage is executed by a single model, and final syntheses are scored via two independent LLM evaluators (LLaMA 3.1 and GPT-4o) based on a rubric with four dimensions: clarity, coherence, originality, and dialecticality. In parallel, a rule-based semantic analyzer detects rhetorical anomalies and ethical values. All outputs and metadata are stored in a Neo4j graph database for structured exploration. (3) Results: The system was applied to four open-weight models (Gemma 7B, Mistral 7B, Dolphin-Mistral, Zephyr 7B) across ten open-ended prompts on ethical, political, and technological topics. The results show consistent stylistic and semantic variation across models, with moderate inter-rater agreement. Semantic diagnostics revealed differences in value expression and rhetorical flaws not captured by rubric scores. (4) Originality: The framework is, to our knowledge, the first to integrate multi-stage reasoning, rubric-based and semantic evaluation, and graph-based storage into a single system. It enables replicable, interpretable, and multidimensional assessment of generative reasoning—supporting researchers, developers, and educators working with LLMs in high-stakes contexts. Full article

Mental reconstruction (MRC) and Free Recall (FR) have been recognized for enhancing the quality of witness statements. However, the mechanisms underlying this association remain insufficiently understood. This study explores how the time allocated to MRC and FR and variations in educational level influence the quality of eyewitness testimonies. Testimony quality is evaluated based on manually annotated content information provided by experts in testimony assessment, which measures adherence to the events. This is further complemented by fine-grained linguistic features, automatically extracted using linguistic analysis tools, to capture stylistic aspects. As a proof of concept, the analysis is performed on a corpus of 96 testimonies in Spanish describing two robbery cases. The results suggest that both mental reconstruction and narration times positively impact the accuracy of testimonies, as inaccuracies predominantly involve peripheral details. Furthermore, while the study confirms that educational level affects testimony quality, no significant differences were observed in the frequency of erroneous reports. This study contributes to the understanding of the relationship between cognitive strategies and the accuracy of witness statements, proposing an analysis approach applicable to forensic psychology for witness assessment. Full article

Between the 5th and 8th centuries AD, several (semi-)nomadic populations invaded ancient Tokharistan (Central Asia), introducing political, socio-economic and cultural changes that also affected pottery production. The study of ceramic materials thereby represents a useful tool for shedding light on the transformations related to such historical events. Unfortunately, no systematic research on ceramics from this region and this period has been conducted to date, and the information available mostly concerns the formal characteristics and imprecise relative chronologies. Aiming to contribute to the knowledge of Early Medieval pottery production in the territory north of the Amu Darya, we present a preliminary investigation on tableware and common wares recovered in the settlements of Khosijat Tepe, Shurob Kurgan, Balalyk Tepe and Dabil Kurgan. This paper provides new data on the vessels’ morphological/stylistic characteristics and relative chronologies, allowing their contextualisation and categorisation. In order to explore the manufacturing processes and their provenance (production areas), a multidisciplinary approach that combines WD-XRF, powder XRD and thin-section optical microscopy was carried out, allowing us to determine the chemical, mineralogical and petrographic compositions, respectively. The results suggest that most vessels consist of local/regional products. The choice and processing of the raw materials are broadly similar. However, slight variations are observed among the ceramics from distinct sites, related to the procurement of clayey sediments from different sources. Although a certain degree of morphological and technological continuity regarding the preceding Kushano-Sasanian pottery tradition is attested for some functional categories, new shapes and decorations appear, confirming the introduction of new practices in the region. Full article

Far from the uniformity of language and the ideal speaker, it is assumed that language varies even within the same speaker, and that such variation is intrinsic to this speaker. This intra-individual variation has been understood mostly as stylistic variation (Hernández-Campoy & Cutillas-Espinosa, 2012) and is becoming increasingly important (Werth et al., 2021) in the scholarly literature. Admitting that stylistic variation exists, I argue that not all intra-individual variation is conditioned and conscious, according to my data. The analysis will be carried out with data from the Basque linguistic atlas (EHHA) (Aurrekoetxea & Videgain, 2010–2020), which examines systematically the intra-individual variation of speakers in the elicitation phase. Based on these data, the aim of this paper is to show that there are also non-conditioned and unconscious types of variation, and to give a preliminary outline of a theory that accommodates these types of variation. Full article

Weibo sentiment analysis has gained prominence, particularly during the COVID-19 pandemic, as a means to monitor public emotions and detect emerging mental health trends. However, challenges arise from Weibo’s informal language, nuanced expressions, and stylistic features unique to social media, which complicate the accurate interpretation of sentiments. Existing models often fall short, relying on text-based methods that inadequately capture the rich emotional texture of Weibo posts, and are constrained by single loss functions that limit emotional depth. To address these limitations, we propose a novel framework incorporating a sentiment graph and self-supervised learning. Our approach introduces a “sentiment graph” that leverages both word-to-post and post-to-post relational connections, allowing the model to capture fine-grained sentiment cues and context-dependent meanings. Enhanced by a gated mechanism within the graph, our model selectively filters emotional signals based on intensity and relevance, improving its sensitivity to subtle variations such as sarcasm. Additionally, a self-supervised objective enables the model to generalize beyond labeled data, capturing latent emotional structures within the graph. Through this integration of sentiment graph and self-supervised learning, our approach advances Weibo sentiment analysis, offering a robust method for understanding the complex emotional landscape of social media. Full article

Background: Artificial intelligence poses a critical challenge to the authenticity of journalistic documents. Objectives: This research proposes a method to automatically identify AI-generated news articles based on various stylistic features. Methods/Approach: We used machine learning algorithms and trained five classifiers to distinguish journalistic news articles from their AI-generated counterparts based on various lexical, syntactic, and readability features. BERTopic was used to extract salient keywords from these articles, which were then used to prompt Google’s Gemini to generate new artificial articles on the same topic. Results: The Random Forest classifier performed the best on the task (accuracy = 98.3%, precision = 0.984, recall = 0.983, and F1-score = 0.983). Random Forest feature importance, Analysis of Variance (ANOVA), Mutual Information, and Recursive Feature Elimination revealed the top five important features: sentence length range, paragraph length coefficient of variation, verb ratio, sentence complex tags, and paragraph length range. Conclusions: This research introduces an innovative approach to prompt engineering using the BERTopic modelling technique and identifies key stylistic features to distinguish AI-generated content from human-generated content. Therefore, it contributes to the ongoing efforts to combat disinformation, enhancing the credibility of content in various industries, such as academic research, education, and journalism. Full article

Facial Expression Recognition (FER) has been widely explored in realistic settings; however, its application to artistic portraiture presents unique challenges due to the stylistic interpretations of artists and the complex interplay of emotions conveyed by both the artist and the subject. This study addresses these challenges through three key contributions. First, we introduce the PortraitEmotion3D (PE3D) dataset, designed explicitly for FER tasks in artistic portraits. This dataset provides a robust foundation for advancing emotion recognition in visual art. Second, we propose an innovative 3D emotion estimation method that leverages three-dimensional labeling to capture the nuanced emotional spectrum depicted in artistic works. This approach surpasses traditional two-dimensional methods by enabling a more comprehensive understanding of the subtle and layered emotions often in artistic representations. Third, we enhance the feature learning phase by integrating a self-attention module, significantly improving facial feature representation and emotion recognition accuracy in artistic portraits. This advancement addresses this domain’s stylistic variations and complexity, setting a new benchmark for FER in artistic works. Evaluation of the PE3D dataset demonstrates our method’s high accuracy and robustness compared to existing state-of-the-art FER techniques. The integration of our module yields an average accuracy improvement of over 1% in recent FER systems. Additionally, combining our method with ESR-9 achieves a comparable accuracy of 88.3% on the FER+ dataset, demonstrating its generalizability to other FER benchmarks. This research deepens our understanding of emotional expression in art and facilitates potential applications in diverse fields, including human–computer interaction, security, healthcare diagnostics, and the entertainment industry. Full article

This article discusses how the affinitive elements in 2 Samuel 1, 3, and 4 can be well-understood using type-scene concepts based on the concept of family resemblance and prototype theory. Applying the “type-scene” concept flexibly enables various arrangements of similar motifs and scenes, or even their absence, in the three murder stories, conveying associated messages. The repetitive motifs and scenes employed are the key characters “going out and coming back”; a background of death; killing a Saulide; mourning; judging or executing the killer; a “blood” curse; and elegy. The successive stylistic variations and modifications in conversation, narration, and elegy, in terms of motif alternations and changes in order, imply that, while David refrains from taking advantage and defends the Saulides’ honor, he treats Saul, Abner, and Ishbosheth differently according to their high and low political values. The increasingly fast pace of the narratives’ logical flow—variously determined by the length of conversation (in both dialogue and soliloquy), mourning, lamentation, and their changes in order and even their absence—indicates that, as the Saulide demise becomes gradually fatal, the narrator’s focus on their remaining members diminishes. Consequently, the establishment of David’s kingship based on his innocence is intensified. Full article

What happens when recognized and diverse conditioning factors of linguistic variation are omitted from analysis and/or are not analyzed under a single analytical procedure? This paper explores the consequences of such a choice on data interpretation and, consequently, (socio)linguistic theorization. Utilizing Twitter-style English in the Philippines (EngPH) as a case study, I employ the Twitter Corpus of Philippine Englishes (TCOPE) primarily to investigate and elucidate variations in three morphosyntactic variables that have been previously examined using a piecemeal approach. I propose a holistic quantitative approach that incorporates documented linguistic, social, diachronic, and stylistic factors in a unified analysis. The paper illustrates the impacts of adopting this holistic approach through two statistical procedures: Bayesian regression modeling and Boruta feature selection with random forest modeling. In contrast to earlier research findings, my overall results reveal biases in non-unified quantitative analyses, where the confidence in the effects of certain factors diminishes in light of others during analysis. The adoption of a unified analysis or modeling also enhances the resolution at which variations have been examined in EngPH. For instance, it highlights that presumed ‘universals’, such as the hierarchy of linguistic > stylistic > diachronic > social factors in explaining variation in some domains, is contingent on the specific variable under examination. Overall, I argue that unified analyses reduce data distortion and introduce more nuanced interpretations and insights that are critical for establishing a well-grounded empirical theory of EngPH variation and language variation as a whole. Full article

Jade artifacts produced in prehistoric China continue to generate extensive scholarly interest. In the absence of textual data, inferring how works functioned in Jade Age communities remains challenging. This paper focuses on Hongshan 红山 culture (4500–3000 BCE) jades, a distinctively styled corpus primarily recovered from late fourth millennium BCE graves in northeastern China. Recent finds within and beyond the Hongshan core zone have enriched the jade inventory and expanded the known scope of its stylistic variations. The analysis sheds light on enigmatic types, reveals the complex representational nature of this corpus, and clarifies the mimetic intentions that resulted in the soft rounded forms characteristic of the style. Most objects examined were unearthed at Hongshan ceremonial centers and have sound excavation pedigrees. Their study relies on contextual archaeological data and comparative visual analysis and draws on the broader Hongshan material world. Further considerations include environment, funerary practices, materiality, cognition, and human anatomy. Ultimately, the paper uncovers new paradigms of figural representation that should open fresh investigative avenues for specialists of early China. Preliminary evaluation of jades unearthed further south at Lingjiatan 凌家滩 and Liangzhu 良渚 sites suggests that some late Neolithic societies adopted Hongshan practices. Current evidence hints at members of prehistoric communities attempting, through jade works, to rationalize their physical circumstances and assert their social power by symbolically fusing with elements of their environments. Full article

Performative style is an important sociolinguistic variable among politicians, who accomplish agentive goals through speech. Examining 32 Spanish politicians, this article focuses on four Andalusian Spanish phenomena: the fronting of /t͡ʃ/ and the deletion of coda /s/, resyllabified intervocalic onset /s/, and intervocalic /d/. The analysis first looks at overall community production norms for the variants then turns to examine the style-shifting patterns of one individual who deviates from these norms. This individual is examined through a consideration of lectal focusing in interaction to track moment-by-moment variation. While coda /s/ and intervocalic /d/ deletion show usage patterns governed by regional and contextual factors, the deletion of onset /s/ and fronted /t͡ʃ/ reflect social variation and style-shifting. While politicians do not blindly follow partisan norms, normative expectations exist at the regional level that they can choose to depart from due to individual motivations and political affiliation in order to carry out identity work. This study combines quantitative examinations of community and individual variation to contribute to our understanding of style-shifting behavior in political speech and how politicians use linguistic tools to take on oppositional identities in the public sphere. Full article

The biblical figure of Balaam is ambivalent in many ways, including his religious affiliation. In the attestations of the Balaam tradition within the Hebrew Bible, different conceptions of this literary character are sometimes blended, sometimes unconnected. In the prose of Num 22–24, he is portrayed as a worshipper of YHWH and even confers with him nightly, albeit being a stranger from far away. In the oracles, a variety of divine designations is used without apparent reason, while the prose alternates between YHWH and Elohim only. The article focuses on the third and fourth oracles (Num 24:3–9, 15–19) and explores their use of divine designations against the backdrop of Num 22–24. The variation can be interpreted as a stylistic feature, literary conception, spoil from different religious traditions, and/or theological statement. Full article

The study presented here concerns an analysis of the structural vulnerability of the arm of the church of the Castello di Santo Stefano Abbey in Monopoli in Puglia (Italy). The ancient monument is located on a small “Peninsula” between two ancient natural harbors, 3 km from Monopoli and 5 km from the archaeological site of Egnazia. Studies on the abbey-castle, have revealed different construction and deconstruction phases relating to historical periods of the Roman age, late antique age, medieval age, modern and contemporary age. The castle is typically attributable to a closed courtyard with four arms. The last one, close to the wall on the port of Tramontana, was added at a contemporary time. The arm of the church is characterized by the crypt with opus reticulatum walls from the two surviving bays of the former Benedictine church, from the residential palace of the 18th century, and from the residential extensions of the 20th century. The north-east and north-west arms, born in the 16th century as defensive earth-filled walls with respective bulwarks (north-east pentagonal and north-west at the tip of a spear), were redeveloped between the 17th and 18th centuries into residential buildings with panoramic loggias and overlooking new ditches enclosed by high walled defenses. The study was conducted with non-invasive methods such as: thermographic surveys; archaeological survey; stratigraphic survey of the walls; stratigraphic survey of plants; photogrammetric survey; 3D returns; Morpho-typological survey of the building material making it possible to acquire data on degradation, on the historicization of instability, on the phases of “growth” of the 18th century and of stylistic “reconstructions” of the 19th century. The data acquired, compared with each other, documented: the deterioration and instability of the foundation wall structures (the crypt) and that of the elevated structures: surviving spans of the church and residential building. Alveolization of the building material, deformations, variations in the geometric shape, lack of resistance of the foundation walls due to the typological and qualitative characteristics of the material (opus reticulatum in soft tuff) with high porosity and problems of the church roof system. Full article

This article illustrates how the aesthetics of two types of Cambodian music—pin peat and Cambodian hip hop—enact Cambodian–Buddhist ethics and function as ritual practices through musicians’ recollections of deceased teachers’ musical legacies. Noting how prevalent historicist and secular epistemologies isolate Cambodian and, more broadly, Southeast Asian musical aesthetics from their ethical and ritual functions, I propose that analyses focusing on Buddhist ethics more closely translate the moral, religious, and ontological aspects inherent in playing and listening to Cambodian music. I detail how Cambodian musicians’ widespread practices of quoting deceased teachers’ variations, repurposing old musical styles, and reiterating the melodies and rhythms played by artistic ancestors have the potential to function as Buddhist rituals, whether those aesthetic and stylistic features surface in pin peat songs or in hip hop. Those aesthetic practices entail a modality of being historical that partially connects with but exceeds historicism’s approach to Buddhism, temporality, and history by enacting relations of mutual care that bring the living and dead to be ontologically coeval. Such relational practices bring me to conclude with a brief discussion rethinking what post-genocide remembrance sounds like and feels like. Full article