Contemporary large language models (LLMs) may have utility for processing unstructured, narrative free-text clinical data contained in electronic health records (EHRs) -- a particularly important use-case for mental health where a majority of routinely-collected patient data lacks structured, machine-readable content. A significant problem for the the United Kingdom's National Health Service (NHS) are the long waiting lists for specialist mental healthcare. According to NHS data, in each month of 2023, there were between 370,000 and 470,000 individual new referrals into secondary mental healthcare services. Referrals must be triaged by clinicians, using clinical information contained in the patient's EHR to arrive at a decision about the most appropriate mental healthcare team to assess and potentially treat these patients. The ability to efficiently recommend a relevant team by ingesting potentially voluminous clinical notes could help services both reduce referral waiting times and with the right technology, improve the evidence available to justify triage decisions. We present and evaluate three different approaches for LLM-based, end-to-end ingestion of variable-length clinical EHR data to assist clinicians when triaging referrals. Our model is able to deliver triage recommendations consistent with existing clinical practices and it's architecture was implemented on a single GPU, making it practical for implementation in resource-limited NHS environments where private implementations of LLM technology will be necessary to ensure confidential clinical data is appropriately controlled and governed.

Understanding the characteristics of swarming autonomous agents is critical for defense and security applications. This article presents a study on using supervised neural network time series classification (NN TSC) to predict key attributes and tactics of swarming autonomous agents for military contexts. Specifically, NN TSC is applied to infer two binary attributes - communication and proportional navigation - which combine to define four mutually exclusive swarm tactics. We identify a gap in literature on using NNs for swarm classification and demonstrate the effectiveness of NN TSC in rapidly deducing intelligence about attacking swarms to inform counter-maneuvers. Through simulated swarm-vs-swarm engagements, we evaluate NN TSC performance in terms of observation window requirements, noise robustness, and scalability to swarm size. Key findings show NNs can predict swarm behaviors with 97% accuracy using short observation windows of 20 time steps, while also demonstrating graceful degradation down to 80% accuracy under 50% noise, as well as excellent scalability to swarm sizes from 10 to 100 agents. These capabilities are promising for real-time decision-making support in defense scenarios by rapidly inferring insights about swarm behavior.

Many studies have shown that human languages tend to optimize for lower complexity and increased communication efficiency. Syntactic dependency distance, which measures the linear distance between dependent words, is often considered a key indicator of language processing difficulty and working memory load. The current paper looks at diachronic trends in syntactic language change in both English and German, using corpora of parliamentary debates from the last c. 160 years. We base our observations on five dependency parsers, including the widely used Stanford CoreNLP as well as 4 newer alternatives. Our analysis of syntactic language change goes beyond linear dependency distance and explores 15 metrics relevant to dependency distance minimization (DDM) and/or based on tree graph properties, such as the tree height and degree variance. Even though we have evidence that recent parsers trained on modern treebanks are not heavily affected by data 'noise' such as spelling changes and OCR errors in our historic data, we find that results of syntactic language change are sensitive to the parsers involved, which is a caution against using a single parser for evaluating syntactic language change as done in previous work. We also show that syntactic language change over the time period investigated is largely similar between English and German for the different metrics explored: only 4% of cases we examine yield opposite conclusions regarding upwards and downtrends of syntactic metrics across German and English. We also show that changes in syntactic measures seem to be more frequent at the tails of sentence length distributions. To our best knowledge, ours is the most comprehensive analysis of syntactic language change using modern NLP technology in recent corpora of English and German.

Cyber risk refers to the risk of defacing reputation, monetary losses, or disruption of an organization or individuals, and this situation usually occurs by the unconscious use of cyber systems. The cyber risk is unhurriedly increasing day by day and it is right now a global threat. Developing countries like Bangladesh face major cyber risk challenges. The growing cyber threat worldwide focuses on the need for effective modeling to predict and manage the associated risk. This paper exhibits a Machine Learning(ML) based model for predicting individuals who may be victims of cyber attacks by analyzing socioeconomic factors. We collected the dataset from victims and non-victims of cyberattacks based on socio-demographic features. The study involved the development of a questionnaire to gather data, which was then used to measure the significance of features. Through data augmentation, the dataset was expanded to encompass 3286 entries, setting the stage for our investigation and modeling. Among several ML models with 19, 20, 21, and 26 features, we proposed a novel Pertinent Features Random Forest (RF) model, which achieved maximum accuracy with 20 features (95.95\%) and also demonstrated the association among the selected features using the Apriori algorithm with Confidence (above 80\%) according to the victim. We generated 10 important association rules and presented the framework that is rigorously evaluated on real-world datasets, demonstrating its potential to predict cyberattacks and associated risk factors effectively. Looking ahead, future efforts will be directed toward refining the predictive model's precision and delving into additional risk factors, to fortify the proposed framework's efficacy in navigating the complex terrain of cybersecurity threats.

Document question answering is a task of question answering on given documents such as reports, slides, pamphlets, and websites, and it is a truly demanding task as paper and electronic forms of documents are so common in our society. This is known as a quite challenging task because it requires not only text understanding but also understanding of figures and tables, and hence visual question answering (VQA) methods are often examined in addition to textual approaches. We introduce Japanese Document Question Answering (JDocQA), a large-scale document-based QA dataset, essentially requiring both visual and textual information to answer questions, which comprises 5,504 documents in PDF format and annotated 11,600 question-and-answer instances in Japanese. Each QA instance includes references to the document pages and bounding boxes for the answer clues. We incorporate multiple categories of questions and unanswerable questions from the document for realistic question-answering applications. We empirically evaluate the effectiveness of our dataset with text-based large language models (LLMs) and multimodal models. Incorporating unanswerable questions in finetuning may contribute to harnessing the so-called hallucination generation.

This short paper describes the first steps in a project to construct a knowledge graph for Brazilian history based on the Brazilian Dictionary of Historical Biographies (DHBB) and Wikipedia/Wikidata. We contend that large repositories of Brazilian-named entities (people, places, organizations, and political events and movements) would be beneficial for extracting information from Portuguese texts. We show that many of the terms/entities described in the DHBB do not have corresponding concepts (or Q items) in Wikidata, the largest structured database of entities associated with Wikipedia. We describe previous work on extracting information from the DHBB and outline the steps to construct a Wikidata-based historical knowledge graph.

We study metric learning from preference comparisons under the ideal point model, in which a user prefers an item over another if it is closer to their latent ideal item. These items are embedded into $\mathbb{R}^d$ equipped with an unknown Mahalanobis distance shared across users. While recent work shows that it is possible to simultaneously recover the metric and ideal items given $\mathcal{O}(d)$ pairwise comparisons per user, in practice we often have a limited budget of $o(d)$ comparisons. We study whether the metric can still be recovered, even though it is known that learning individual ideal items is now no longer possible. We show that in general, $o(d)$ comparisons reveals no information about the metric, even with infinitely many users. However, when comparisons are made over items that exhibit low-dimensional structure, each user can contribute to learning the metric restricted to a low-dimensional subspace so that the metric can be jointly identified. We present a divide-and-conquer approach that achieves this, and provide theoretical recovery guarantees and empirical validation.

A mid-day drunk-dial from a friend has changed one man's life forever. Noland Arbaugh, 29, rose to fame after being revealed as Neuralink's first patient to receive its brain chip, but it all started when his friend called slurring his words in September. Arbaugh was paralyzed eight years ago during a diving accident. The friend called Arbaugh to tell him about Elon Musk opening up human trials and urged him to apply and even helped him fill out the form. Just five months after he was approved for the Neuralink trial, Arbaugh had a cutting edge brain chip embedded in his skull.

U.S. government spending on artificial intelligence has exploded in the past year, driven by increased military investments, according to a report by the Brookings Institution, a think tank based in Washington D.C. The report found that the potential value of AI-related federal contracts increased by almost 1,200%, from 355 million in the period leading up to August 2022, to 4.6 billion in the period leading up to August 2023. This increase was almost entirely driven by the Department of Defense (DoD). The total amount committed by the DoD to AI-related contracts increased from 190 million in the period leading up to August 2022 to 557 million in the period leading up to August 2023. The total that the DoD might spend on AI-related contracts if each contract were extended to its fullest terms grew even faster, from 269 million in the period leading up to August 2022 to 4.3 billion in the period leading up to August 2023.

The New York Times reports that Israel's military intelligence has been using an experimental facial recognition program in Gaza that's misidentified Palestinian civilians as having ties to Hamas. Google Photos allegedly plays a part in the chilling program's implementation, although it appears not to be through any direct collaboration with the company. The surveillance program reportedly started as a way to search for Israeli hostages in Gaza. However, as often happens with new wartime technology, the initiative was quickly expanded to "root out anyone with ties to Hamas or other militant groups," according to The NYT. The technology is flawed, but Israeli soldiers reportedly haven't treated it as such when detaining civilians flagged by the system.