Aug 2026· Zenodo (CERN European Organization for Nuclear Research)
Natural Language Processing Techniques
Abstract
This paper introduces a novel self-adaptive grammar rule engine based on graph neural networks. The engine leverages a graph representation of the input text to automatically infer grammatical rules and generate more accurate translations. We propose a method that dynamically adjusts the graph structure based on contextual information, enabling the engine to learn and optimize rules for improved translation quality. The core mechanism involves a graph neural network (GNN) that transforms text into a graph representation, allows for reasoning and prediction, and iteratively refines the graph structure to achieve optimal results. The proposed approach offers a significant advancement over existing methods, providing a flexible and adaptable grammar rule system capable of handling complex linguistic structures. We demonstrate the engine's effectiveness through a series of benchmark tests focusing on natural language translation, capturing both syntactic and semantic nuances. The results showcase the engine's ability to produce translations that closely match human-generated outputs while exhibiting improved accuracy and fluency.
Agile methods continue to gain popularity. In particular, the Scrum method appears to be on the verge of becoming a de-facto standard in the industry, leading the so called Agile movement. While there are success stories and recommendations, there is little scientifically valid evidence of the challenges in the adoption of Agile methods in general, and Scrum in particular. Little, if anything, is empirically known about the application and adoption of Scrum in a multi-team and multi-project situation. The authors carried out an ethnographically informed longitudinal case study in industrial settings and closely followed how the Scrum method was adopted in a 20-person department, working in a simultaneous multi-project R&D environment. Altogether 10 challenges pertinent to the case of multi-team multi-project Scrum adoption were identified in the study. The authors contend that these results carry great relevance for other industrial teams. Future research avenues arising from the study are indicated.
A. Marchenko, P. Abrahamsson· Agile Conference· 59 citations· ⚡11
Context: While there are many success stories of achieving high reuse and improved quality using software platforms, there is a need to investigate the issues and challenges organizations face when transitioning to a software platform strategy. Objective: This case study provides a comprehensive taxonomy of the challenges faced when a medium-scale organization decided to adopt software platforms. The study also reveals how new trends in software engineering (i.e. agile methods, distributed development, and flat management structures) interplayed with the chosen platform strategy. Method: We used an ethnographic approach to collect data by spending time at a medium-scale company in Scandinavia. We conducted 16in-depth interviews with representatives of eight different teams, three of which were working on three separate platforms. The collected data was analyzed using Grounded Theory. Results: The findings identify four classes of challenges, namely: business challenges, organizational challenges, technical challenges, and people challenges. The article explains how these findings can be used to help researchers and practitioners identify practical solutions and required tool support. Conclusion: The organization's decision to adopt a software platform strategy introduced a number of challenges. These challenges need to be understood and addressed in order to reap the benefits of reuse. Researchers need to further investigate issues such as supportive organizational structures for platform development, the role of agile methods in software platforms, tool support for testing and continuous integration in the platform context, and reuse recommendation systems.
Yaser Ghanam, F. Maurer, P. Abrahamsson· Information and Software Tec...· 41 citations· ⚡3
This article is a contribution towards an understanding of open access (OA) publishing. It proposes an analysis framework of 18 core attributes, divided into the areas of bibliographic information, activity metrics, economics, accessibility, and predatory issues. The framework has been employed in a systematic analysis of 30 OA journals in software engineering (SE) and information systems (IS), which were selected from among 386 OA journals in Computer Science from the Directory of OA Journals. An analysis was performed on the sample of the journals, to provide an overview of the current situation of OA journals in the fields of SE and IS. The journals were then compared between-group, according to the presence of article processing charges. A within-group analysis was performed on the journals requesting article processing charges from authors, in order to understand what is the value added according to different price ranges. This article offers several contributions. It presents an overview of OA definitions and models. It provides an analysis framework born from the observation of data and the existing literature. It raises the need to study OA in the fields of SE and IS while offering a first analysis. Finally, it provides recommendations to readers of OA journals. This paper highlights several concerns still threatening the adoption of OA publishing in the fields of SE and IS. Among them, it is shown that high article processing charges are not sufficiently justified by the publishers, which often lack transparency and may prevent authors from adopting OA.
D. Graziotin, Xiaofeng Wang, P. Abrahamsson· Scientometrics· 21 citations· ⚡1
Structure-based machine learning algorithms have been utilized to predict the properties of protein-protein interaction (PPI) complexes, such as binding affinity, which is critical for understanding biological mechanisms and disease treatments. While most existing algorithms represent PPI complex graph structures at the atom-scale or residue-scale, these representations can be computationally expensive or may not sufficiently integrate finer chemical-plausible interaction details for improving predictions. Here, we introduce MCGLPPI, a novel geometric representation learning framework that combines graph neural networks (GNNs) with the MARTINI molecular coarse-grained (CG) model to predict overall PPI properties accurately and efficiently. This framework maps proteins onto a concise CG-scale complex graph, where nodes represent CG beads and edges encode chemically plausible interactions. The GNN-based encoder is tailored to extract high-quality representations from this graph, efficiently capturing the overall properties of the protein complex structure. Extensive experiments on three different downstream PPI property prediction tasks demonstrate that MCGLPPI achieves competitive performance compared with the counterparts at the atom- and residue-scale, but with only a third of the computational resource consumption. Furthermore, the CG-scale pre-training on protein domain-domain interaction structures enhances its predictive capabilities for PPI tasks. MCGLPPI offers an effective and efficient solution for PPI overall property predictions, serving as a promising tool for the large-scale analysis of biomolecular interactions.
Yang Yue, Shu Li, Yihua Cheng et al.· bioRxiv· 14 citations
Accurate modeling of protein-peptide interactions is essential for understanding fundamental biological processes and designing peptide-based drugs. However, predicting the complex structures of these interactions remains challenging, primarily due to the high conformational flexibility of peptides. To support a fair and systematic evaluation of recent deep learning (DL) approaches, we introduce PepPCBench, a benchmarking framework tailored to assess protein folding neural networks (PFNNs) in protein-peptide complex prediction. As part of this framework, we curated PepPCSet, a data set of 261 experimentally resolved complexes with peptides ranging from 5 to 30 residues. We benchmark five full-atom PFNNs, including AlphaFold3 (AF3), AlphaFold-Multimer (AFM), Chai-1, HelixFold3 (HF3), and RoseTTAFold-All-Atom (RFAA), using comprehensive evaluation metrics. Our benchmarking reveals meaningful performance differences among these methods and highlights the influence of peptide length, conformational flexibility, and training set similarity on prediction accuracy. While AF3 shows strong performance in structure prediction, further analysis indicates that confidence metrics correlate poorly with experimental binding affinities, underscoring the need for improved scoring strategies and generalizability. By providing a reproducible and extensible framework, PepPCBench enables a robust evaluation of PFNN-based methods and supports their continued development for peptide-protein structure prediction.
Silong Zhai, Huifeng Zhao, Jike Wang et al.· Journal of Chemical Informat...· 13 citations· ⚡1
Molecular representation learning (MRL) has shown promise in accelerating drug development by predicting chemical properties. However, imperfectly annotation among datasets pose challenges in model design and explainability. In this work, we formulate molecules and corresponding properties as a hypergraph, extracting three key relationships: among properties, molecule-to-property, and among molecules, and developed a unified and explainable multi-task MRL framework, OmniMol. It integrates a task-related meta-information encoder and a task-routed mixture of experts (t-MoE) backbone to capture correlations among properties and produce task-adaptive outputs. To capture underlying physical principles among molecules, we implement an innovative SE(3)-encoder for physical symmetry, applying equilibrium conformation supervision, recursive geometry updates, and scale-invariant message passing to facilitate learning-based conformational relaxation. OmniMol achieves state-of-the-art performance in properties prediction, reaches top performance in chirality-aware tasks, demonstrates explainability for all three relations, and shows effective performance in practical applications. Our code is available in our https://github.com/bowenwang77/OmniMol public repository. AI models for drug discovery often struggle with real-world, incomplete data. Here, the authors present OmniMol, a framework using hypergraphs to improve predictions of molecular properties, addressing challenges of imperfect data annotation and enhancing model explainability.
Assistant Professor Pat Pataranutaporn describes a new interface that lets everyday users glimpse inside an AI's neural network before their chatbot ever says a word.
MIT News · Artificial Intelligence· news.mit.eduJul 6, 2026
PhD student Rachel Sava, winner of the Envisioning the Future of Computing Prize, explores transformative improvements and dystopian risks of neural technology.