John Blake

The increasing volume of digital linguistic evidence and the expansion of virtual communicative spaces demand forensic linguistic tools that are both methodologically rigorous and accessible to a wider range of users, including legal practitioners. This paper introduces a bespoke, web-based authorship analysis tool designed to identify and visualize key textual patterns across questioned and known datasets. Operating through an intuitive graphical user interface, the tool offers an accessible alternative to command-line methods, expanding the reach of computational forensic linguistics to non-technical users. The tool integrates two core functionalities: (1) automated identification and comparison of shared n-grams, particularly bigrams and trigrams, within and between datasets, and (2) Key Word in Context (KWIC) exploration to support the interpretation of overlapping or distinctive lexical patterns. By combining these capabilities, the system enables both confirmatory and exploratory analysis of linguistic evidence, supporting investigations where authorship is contested or where stylistic consistency must be assessed. It has applications across the three domains of forensic linguistics: language as evidence, interaction in legal settings, and language of the law. In addition to its methodological contributions, the paper addresses key challenges in forensic semiotics: subjectivity, reproducibility, and transparency. It shows how computational approaches can mitigate the “subjectivity effect” by offering replicable, data-driven analyses of linguistic behaviour. Courtroom admissibility is also supported by producing clear, step-by-step analytical records and exportable logs that document each analytic operation. Our contribution aligns with the aims of computational forensic linguistics to enhance the interpretive rigour and evidentiary validity of linguistic analyses.

This study investigates the use of a pragmatics-focused computerized dynamic assessment (C-DA) of request-based L2 English email writing among Japanese university learners of English. The C-DA administers four email tasks to learners, automatically identifies perceived instances of pragmatic inappropriateness in texts and provides feedback, based on learner corpus perception data. The study compares the C-DA’s effectiveness in promoting learning among three participant groups: a graduated feedback group received feedback based on the concept of the zone of proximal development (ZPD); an explicit-only group received explicit feedback, and a control group did not receive feedback. The C-DA was administered twice, with a two-week delay between administrations. Both treatment groups showed evidence of development to a greater extent than the control both within a session and between sessions in terms of openings, closings and modification strategies. Comparing treatment groups, the graduated group responded more to feedback within a session for openings and closings, but not for modification strategies. Between sessions, there was no difference between treatment groups; however, across the entirety of the study, the graduated feedback group responded significantly more to feedback. Findings provide support for computerized approaches to L2 email writing feedback, and for approaches sensitive to a learner’s ZPD.

This study investigates the potential for Large Language Models (LLMs) to scale-up Dynamic Assessment (DA). To facilitate such an investigation, we first developed DynaWrite—a modular, microservices-based grammatical tutoring application which supports multiple LLMs to generate dynamic feedback to learners of English. Initial testing of 21 LLMs, revealed GPT-4o and neural-chat to have the most potential to scale-up DA in the language learning classroom. Further testing of these two candidates found both models performed similarly in their ability to accurately identify grammatical errors in user sentences. However, GPT-4o consistently outperformed neural-chat in the quality of its DA by generating clear, consistent, and progressively explicit hints. Real-time responsiveness and system stability were also confirmed through detailed performance testing, with GPT-4o exhibiting sufficient speed and stability. This study shows that LLMs can be used to scale-up dynamic assessment and thus enable dynamic assessment to be delivered to larger groups than possible in traditional teacher-learner settings.

This research contributes to the studies on the origins and transformations of Tchaikovsky’s Children’s Album, Op. 39 using the linguistic methods of discourse, metaphor, and comparative analysis to explore a number of connected questions and their impact on how the audiences and scholars perceive and understand the compositions. These methods are supported by the technology provided by computational linguistics, such as large language models along with music analysis algorithms based on signature pattern elicitation. This article examines how artificial intelligence technologies can shed light on the differing views on the Children’s Album. The meanings and implications of the published reordering of the pieces are explored. The influence of Schumann’s Album for the Young and the broader pedagogical and cultural significance of editorial transformations is investigated. Through this interdisciplinary approach, this study offers new insights into the compositional intent and interpretive possibilities of Tchaikovsky’s work. The presented results of the musicology, literary, computational, and linguistic analyses complement the few scholarly studies aimed at unveiling the intriguing metaphors and connections of the Children’s Album, which tend to remain in the shadows of his larger-scale piano and symphonic works.

This paper presents a novel algorithm that leverages cutting-edge machine-learning techniques to accurately and efficiently detect AI-generated texts. Rapid advancements in natural language processing models have led to the generation of text closely resembling human language, making it increasingly difficult to differentiate between human and AI-generated content. However, misuse of such texts presents a serious and imminent threat to the quality of academic publishing. This underscores the urgent need for robust detection mechanisms to ensure information quality, maintain trust, and preserve the integrity of research publications. Our proposed model outperformed existing algorithms for accuracy with less computational complexity. The proposed model is a feature-based hybrid deep learning network that leverages part-of-speech tagging and integrates Bidirectional Long Short-Term Memory (BiLSTM) networks with Attention modules. The initial module extracts local contextual features using convolutional layers, followed by BiLSTM layers that capture long-term dependencies from past and future sequences. An attention mechanism highlights critical sequence components, enhancing the model’s focus on relevant data. The outputs from the attention and initial modules are concatenated through a residual connection, ensuring comprehensive feature representation. This combination is then fed into dense layers for final classification, effectively balancing feature richness and computational efficiency. The proposed model was evaluated on two benchmark datasets, achieving 85.00% and 88.00% accuracy, respectively.