Detecting population-level mental health responses to global negative events through social media language remains understudied, despite its potential for public health surveillance. While pretrained language models (PLMs) have shown promise in mental health detection, their effectiveness in capturing event-driven collective psychological shifts – especially across diverse crisis contexts – is unclear. We present a prototype evaluation of three PLMs for identifying population mental health dynamics triggered by real-world negative events. We introduce two novel datasets specifically designed for this task. Our findings suggest that DistilBERT is better suited to the noisier global negative events data, while MentalRoBERTa shows the validity of the method on the Covid-19 tidier data. SHAP interpretability analysis of 500 randomly sampled posts revealed that mental-health related vocabulary (anxiety, depression, worthless) emerged as the most influential linguistic markers for mental health classification.

Large language model classifiers do not directly offer transparency: it is not clear why one class is chosen over another. In this work, summaries explaining the suicide risk level assigned using a fine-tuned mental-roberta-base model are generated from key phrases extracted using SHAP explainability using Mistral-7B. The training data for the classifier consists of all Reddit posts of a user in the University of Maryland Reddit Suicidality Dataset, Version 2, with their suicide risk labels along with selected features extracted from each post by the Linguistic Inquiry and Word Count (LIWC-22) tool. The resulting model is used to make predictions regarding risk on each post of the users in the evaluation set of the CLPsych 2024 shared task, with a SHAP explainer used to identify the phrases contributing to the top scoring, correct and severe risk categories. Some basic stoplisting is applied to the extracted phrases, along with length based filtering, and a locally run version of Mistral-7B-Instruct-v0.1 is used to create summaries from the highest value (based on SHAP) phrases.

Sharing data containing personal information often requires its anonymization, even when consent for sharing was obtained from the data originator. While approaches exist for automated anonymization of text, the area is not as thoroughly explored in speech. This work focuses on identifying, replacing and inserting replacement named entities synthesized using voice cloning into original audio thereby retaining prosodic information while reducing the likelihood of deanonymization. The approach employs a novel named entity recognition (NER) system built directly on speech by training HuBERT (Hsu et al, 2021) using the English speech NER dataset (Yadav et al, 2020). Name substitutes are found using a masked language model and are synthesized using text to speech voice cloning (Eren and team, 2021), upon which the substitute named entities are re-inserted into the original text. The approach is prototyped on a sample of the LibriSpeech corpus (Panyatov et al, 2015) with each step evaluated individually.

Many automatic semantic relation extraction tools extract subject-predicate-object triples from unstructured text. However, a large quantity of these triples merely represent background knowledge. We explore using full texts of biomedical publications to create a training corpus of informative and important semantic triples based on the notion that the main contributions of an article are summarized in its abstract. This corpus is used to train a deep learning classifier to identify important triples, and we suggest that an importance ranking for semantic triples could also be generated.

As the quantity of publications increases daily, researchers are forced to narrow their attention to their own specialism and are therefore less likely to make new connections with other areas. Literature based discovery (LBD) supports the identification of such connections. A number of LBD tools are available, however, they often suffer from limitations such as constraining possible searches or not producing results in real-time. We introduce HiDE (Hidden Discovery Explorer), an online knowledge browsing tool which allows fast access to hidden knowledge generated from all abstracts in Medline. HiDE is fast enough to allow users to explore the full range of hidden connections generated by an LBD system. The tool employs a novel combination of two approaches to LBD: a graph-based approach which allows hidden knowledge to be generated on a large scale and an inference algorithm to identify the most promising (most likely to be non trivial) information. Available at https://skye.shef.ac.uk/kdisc