Global Negative Correlation Learning: A Unified Framework for Global Optimization of Ensemble Models

Artificial intelligence extracts key insights from legal documents to predict intimate partner femicide

Legal documents serve as valuable repositories of information pertaining to crimes, encompassing not only legal aspects but also relevant details about criminal behaviors. To date and the best of our knowledge, no studies in the field examine legal documents for crime understanding using an Artificial Intelligence (AI) approach. The present study aims to fill this research gap by identifying relevant information available in legal documents for crime prediction using Artificial Intelligence (AI). This innovative approach will be applied to the specific crime of Intimate Partner Femicide (IPF). A total of 491 legal documents related to lethal and non-lethal violence by male-to-female intimate partners were extracted from the Vlex legal database. The information included in these documents was analyzed using AI algorithms belonging to Bayesian, functions-based, instance-based, tree-based, and rule-based classifiers. The findings demonstrate that specific information from legal documents, such as past criminal behaviors, imposed sanctions, characteristics of violence severity and frequency, as well as the environment and situation in which this crime occurs, enable the correct detection of more than three-quarters of both lethal and non-lethal violence within male-to-female intimate partner relationships. The obtained knowledge is crucial for professionals who have access to legal documents, as it can help identify high-risk IPF cases and shape strategies for preventing crime. While this study focuses on IPF, this innovative approach has the potential to be extended to other types of crimes, making it applicable and beneficial in a broader context.

One Model is Not Enough: Ensembles for Isolated Sign Language Recognition

In this paper, we dive into sign language recognition, focusing on the recognition of isolated signs. The task is defined as a classification problem, where a sequence of frames (i.e., images) is recognized as one of the given sign language glosses. We analyze two appearance-based approaches, I3D and TimeSformer, and one pose-based approach, SPOTER. The appearance-based approaches are trained on a few different data modalities, whereas the performance of SPOTER is evaluated on different types of preprocessing. All the methods are tested on two publicly available datasets: AUTSL and WLASL300. We experiment with ensemble techniques to achieve new state-of-the-art results of 73.84% accuracy on the WLASL300 dataset by using the CMA-ES optimization method to find the best ensemble weight parameters. Furthermore, we present an ensembling technique based on the Transformer model, which we call Neural Ensembler.