keyword frequency analysis

This paper addresses critical challenges in NLP research—including ill-defined objectives, absent stakeholder representation, and weakened application intent—by proposing the first interpretable three-stage framework to automatically identify the research positioning underlying models and datasets (i.e., objectives, methodologies, stakeholders, and applicable contexts). The method integrates structured information extraction, symbolic rule-based reasoning, and context-aware inference, validated on a newly annotated hate speech dataset and established fact-checking benchmarks. It systematically uncovers three emerging trends in NLP: objective generalization, strengthened scientific orientation, and a shift toward human-AI collaboration. The framework consistently outperforms strong LLM baselines on fact-checking and hate speech detection tasks. Applied to recent top-tier conference papers, it precisely diagnoses objective ambiguity and application–research misalignment, thereby promoting societal relevance and impact-driven NLP research.

This study addresses the inflated Type I error rates in traditional dispersion-based semantic breadth hypothesis tests, which are susceptible to interference from differences in semantic directionality. To resolve this issue, the authors propose a Householder-aligned permutation test: first, Householder reflections are employed to align the mean directions of word embeddings, effectively neutralizing directional confounds; then, a non-parametric permutation test is conducted on the aligned contextualized word embedding clouds to accurately detect genuine differences in semantic breadth. This approach is the first to successfully disentangle semantic breadth from directional variation, achieving a 32.5% reduction in Type I error rates while maintaining high sensitivity to true semantic differences. Furthermore, GPU-accelerated batch processing yields a 23-fold speedup over CPU-based baselines.

This study investigates the cross-cultural applicability of the Media Frame Corpus (MFC) framework beyond English and U.S. contexts, focusing on political-economic news in Brazilian Portuguese. Method: We construct FrameNews-PT—the first annotated corpus for this language—following MFC guidelines, conduct multiple rounds of expert annotation, and perform rigorous inter-annotator agreement assessment. Contribution/Results: Systematic validation confirms that all 15 MFC frames are broadly applicable, requiring only minor guideline adaptations. We identify a “fallback” phenomenon wherein annotators consistently resort to generic frames for emerging topics, revealing dependency on broad conceptual categories. Zero-shot cross-lingual transfer of language models underperforms fine-tuned baselines, underscoring the critical role of context-sensitive, culturally grounded annotation. The study proposes that cross-cultural frame application necessitates adaptive refinement aligned with local semantic practices, establishing both a methodological paradigm and an empirical benchmark for global media framing research.

This study addresses the challenge of modeling bidirectional semantic change—such as concurrent sense gain and loss—in contexts where slang coexists with standard usage. To this end, the authors introduce two complementary benchmark datasets: BD-LSC for cross-temporal semantic change analysis and ST-WSD for fine-grained word sense disambiguation, both featuring instance-level sense annotations that explicitly capture sense acquisition, loss, and stability. A comprehensive evaluation employing contextual embedding clustering, supervised learning, Transformer-based models, and large language models—including GPT-4o—reveals that GPT-4o achieves the highest performance in Exact Sense Match and multi-label accuracy. Nevertheless, all models attain only a Macro-F1 of approximately 0.5 on rare slang senses, underscoring the persistent difficulty of this task.

This work addresses the challenge of automatically identifying and framing entity narrative roles (e.g., protagonist, antagonist, innocent victim) in media texts. Methodologically, it introduces the first fine-grained, cross-lingual, and interpretable narrative analysis system: (1) a novel 22-category role taxonomy; (2) a two-stage sequence labeling and role classification model built upon multilingual pretrained language models; and (3) integrated interactive knowledge graphs, cross-document frame comparison, and timeline-based evolutionary visualization. The system supports Bulgarian, English, Hindi, Russian, and Portuguese, and demonstrates empirical efficacy in analyzing narratives around the Russia–Ukraine conflict and climate change. It is open-sourced and deployed as a real-time online analytical platform. Key contributions include: the first fine-grained narrative role taxonomy; multilingual, dynamic entity role tracking; and an interpretable, frame-comparative analytical paradigm tailored for media discourse analysis.

This study investigates how dataset characteristics constrain the effectiveness of quantitative methods in detecting semantic change within historical linguistics. By comparing two diachronic approaches—quadripartite conceptual modeling applied to the EEBO-TCP corpus and SynFlow analysis on the Royal Society Corpus—the research examines differences in conceptual operationalization, underlying data assumptions, and diachronic interpretability. The findings highlight the limitations of purely lexical frequency-based methods and demonstrate that data structure, including temporal granularity and textual representativeness, fundamentally determines which types of semantic shifts can be reliably identified. This work thus offers methodological guidance for historical semantic research and advocates for the development of quantitative paradigms better aligned with the specific properties of historical linguistic data.

This study addresses the limitations of traditional metaphor analysis, which often focuses narrowly on source domains and struggles to uncover differences in the semantic frames activated by metaphors within complex discourses. To overcome this, the paper proposes an integrated framework that combines Conceptual Metaphor Theory with computational linguistics, leveraging natural language processing and semantic frame analysis to automatically identify salient metaphors in discourse and enable fine-grained cross-frame comparisons. Applied to climate news corpora, the method successfully detects both established and novel metaphorical frames. Furthermore, it reveals that while conservative and liberal media may employ the same source domains in discussions of immigration, they invoke markedly distinct semantic frames, thereby underscoring the constitutive role of metaphor in shaping political discourse.

This study systematically evaluates the effectiveness of generative large language models (LLMs)—including GPT and Claude—for news framing analysis, benchmarking them against bag-of-words models, encoder-only Transformers, and human coding. Using a high-quality, manually annotated dataset comprising six months of U.S. news coverage on the 2022 mpox outbreak, the work adopts methodological pluralism to assess performance across framing identification, classification, and interpretation stages. Results indicate that generative LLMs underperform relative to human coders—and in certain cases, even lag behind smaller, domain-specific models—underscoring the indispensable role of human oversight and validation. The primary contribution is the articulation of a “human–AI staged collaboration” framework for computational framing analysis, which leverages complementary strengths across methods. This paradigm advances reproducible methodology selection and provides a practical roadmap for computational communication research. (149 words)

This study addresses the effectiveness of computational approaches for automatically detecting generic news frames in political communication. By systematically evaluating the performance of five pre-trained language models—BERT, RoBERTa, DeBERTa, DistilBERT, and ALBERT—the research constructs and leverages the first high-quality annotated dataset of Swiss election news, enabling assessment of model robustness beyond U.S.-centric contexts. The work contributes a comprehensive empirical comparison of BERT variants for news framing detection, releases the annotated dataset and fine-tuned models as open-source resources, and validates the applicability of the best-performing model for political communication analysis. These contributions establish a reproducible methodological foundation for computational social science research on media framing across diverse sociopolitical settings.