More than 80% of people who commit suicide disclose their intention to do so on social media. The main information we can use in social media is user-generated posts, since personal information is not always available. Identifying all possible emotions in a single textual post is crucial to detecting the user’s mental state; however, human emotions are very complex, and a single text instance likely expresses multiple emotions. This paper proposes a new multi-label emotion graph representation for social media post-based mental health classification. We first construct a word–document graph tensor to describe emotion-based contextual representation using emotion lexicons. Then, it is trained by multi-label emotions and conducts a graph propagation for harmonising heterogeneous emotional information, and is applied to a textual graph mental health classification. We perform extensive experiments on three publicly available social media mental health classification datasets, and the results show clear improvements.

Background: To better understand domestic violence, data sources from multiple sectors such as police, justice, health, and welfare are needed. Linking police data to data collections from other agencies could provide unique insights and promote an all-of-government response to domestic violence. The New South Wales Police Force attends domestic violence events and records information in the form of both structured data and a free-text narrative, with the latter shown to be a rich source of information on the mental health status of persons of interest (POIs) and victims, abuse types, and sustained injuries. Objective: This study aims to examine the concordance (ie, matching) between mental illness mentions extracted from the police’s event narratives and mental health diagnoses from hospital and emergency department records. Methods: We applied a rule-based text mining method on 416,441 domestic violence police event narratives between December 2005 and January 2016 to identify mental illness mentions for POIs and victims. Using different window periods (1, 3, 6, and 12 months) before and after a domestic violence event, we linked the extracted mental illness mentions of victims and POIs to clinical records from the Emergency Department Data Collection and the Admitted Patient Data Collection in New South Wales, Australia using a unique identifier for each individual in the same cohort. Results: Using a 2-year window period (ie, 12 months before and after the domestic violence event), less than 1% (3020/416,441, 0.73%) of events had a mental illness mention and also a corresponding hospital record. About 16% of domestic violence events for both POIs (382/2395, 15.95%) and victims (101/631, 16.01%) had an agreement between hospital records and police narrative mentions of mental illness. A total of 51,025/416,441 (12.25%) events for POIs and 14,802/416,441 (3.55%) events for victims had mental illness mentions in their narratives but no hospital record. Only 841 events for POIs and 919 events for victims had a documented hospital record within 48 hours of the domestic violence event. Conclusions: Our findings suggest that current surveillance systems used to report on domestic violence may be enhanced by accessing rich information (ie, mental illness) contained in police text narratives, made available for both POIs and victims through the application of text mining. Additional insights can be gained by linkage to other health and welfare data collections.

The significance of mental health classification is paramount in contemporary society, where digital platforms serve as crucial sources for monitoring individuals' well-being. However, existing social media mental health datasets primarily consist of text-only samples, potentially limiting the efficacy of models trained on such data. Recognising that humans utilise cross-modal information to comprehend complex situations or issues, we present a novel approach to address the limitations of current methodologies. In this work, we introduce a Multimodal and Multi-Teacher Knowledge Distillation model for Mental Health Classification, leveraging insights from cross-modal human understanding. Unlike conventional approaches that often rely on simple concatenation to integrate diverse features, our model addresses the challenge of appropriately representing inputs of varying natures (e.g., texts and sounds). To mitigate the computational complexity associated with integrating all features into a single model, we employ a multimodal and multi-teacher architecture. By distributing the learning process across multiple teachers, each specialising in a particular feature extraction aspect, we enhance the overall mental health classification performance. Through experimental validation, we demonstrate the efficacy of our model in achieving improved performance.

Despite recent advances in large language models (LLMs), most QA benchmarks are still confined to single-paragraph or single-document settings, failing to capture the complexity of real-world information-seeking tasks. Practical QA often requires multi-hop reasoning over information distributed across multiple documents, modalities, and structural formats. Although prior datasets made progress in this area, they rely heavily on Wikipedia-based content and unimodal plain text, with shallow reasoning paths that typically produce brief phrase-level or single-sentence answers, thus limiting their realism and generalizability. We propose DocHop-QA, a large-scale benchmark comprising 11,379 QA instances for multimodal, multi-document, multi-hop question answering. Constructed from publicly available scientific documents sourced from PubMed, DocHop-QA is domain-agnostic and incorporates diverse information formats, including textual passages, tables, and structural layout cues. Unlike existing datasets, DocHop-QA does not rely on explicitly hyperlinked documents; instead, it supports open-ended reasoning through semantic similarity and layout-aware evidence synthesis. To scale realistic QA construction, we designed an LLM-driven pipeline grounded in 11 high-frequency scientific question concepts. We evaluated DocHop-QA through four tasks spanning structured index prediction, generative answering, and multimodal integration, reflecting both discriminative and generative paradigms. These tasks demonstrate DocHop-QA's capacity to support complex, multimodal reasoning across multiple documents.