Twitter is increasingly investigated as a means of detecting mental health status, including depression and suicidality, in the population. However, validated and reliable methods are not yet fully established. This study aimed to examine whether the level of concern for a suicide-related post on Twitter could be determined based solely on the content of the post, as judged by human coders and then replicated by machine learning. From 18th February 2014 to 23rd April 2014, Twitter was monitored for a series of suicide-related phrases and terms using the public Application Program Interface (API). Matching tweets were stored in a data annotation tool developed by the Commonwealth Scientific and Industrial Research Organisation (CSIRO). During this time, 14,701 suicide-related tweets were collected: 14% were randomly (n=2000) selected and divided into two equal sets (Set A and B) for coding by human researchers. Overall, 14% of suicide-related tweets were classified as ‘strongly concerning’, with the majority coded as ‘possibly concerning’ (56%) and the remainder (29%) considered ‘safe to ignore’. The overall agreement rate among the human coders was 76% (average κ=0.55). Machine learning processes were subsequently applied to assess whether a ‘strongly concerning’ tweet could be identified automatically. The computer classifier correctly identified 80% of ‘strongly concerning’ tweets and showed increasing gains in accuracy; however, future improvements are necessary as a plateau was not reached as the amount of data increased. The current study demonstrated that it is possible to distinguish the level of concern among suicide-related tweets, using both human coders and an automatic machine classifier. Importantly, the machine classifier replicated the accuracy of the human coders. The findings confirmed that Twitter is used by individuals to express suicidality and that such posts evoked a level of concern that warranted further investigation. However, the predictive power for actual suicidal behaviour is not yet known and the findings do not directly identify targets for intervention

The tedious task of responding to a backlog of email is one which is familiar to many researchers. As a subset of email management, we address the problem of constructing a summary of email discussions. Specifically, we examine ongoing discussions which will ultimately culminate in a consensus in a decision-making process. Our summary provides a snapshot of the current state-of-affairs of the discussion and facilitates a speedy response from the user, who might be the bottleneck in some matter being resolved. We present a method which uses the structure of the thread dialogue and word vector techniques to determine which sentence in the thread should be extracted as the main issue. Our solution successfully identifies the sentence containing the issue of the thread being discussed, potentially more informative than subject line.

Yufei Wang, Ian Wood, Stephen Wan, Mark Dras, Mark Johnson. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

Necva Bölücü, Maciej Rybinski, Stephen Wan. Findings of the Association for Computational Linguistics: EMNLP 2023. 2023.

Food extrusion is a widely used processing technique that transforms raw ingredients into structured food products—foods with well-defined textures, shapes, and functionalities—through mechanical shear and thermal energy. Despite its broad industrial application, the absence of a standardised, structured dataset capturing extrusion research parameters has hindered research synthesis, product development, and process optimisation. To address this gap, we introduce a manually curated food extrusion literature dataset capturing publication details, product types, process parameters, formulation data, experimental variables, characterisation metrics, and study-level insights. However, while manually curated datasets are typically of high quality, their scope is limited by time and resource constraints. We propose a method to supplement the dataset using large language models (LLMs) and evaluate the accuracy of LLMs in extracting structured food extrusion data from the scientific literature. Our findings demonstrate that LLMs can effectively extract structured information. However, some challenges, such as hallucination and missing contextual details, remain, suggesting that human effort can be spent on validating the resulting data. This still represents significant time savings as validation is a less time-consuming task than data extraction. We argue that LLMs thus represent a viable tool in providing supplementary datasets, and we propose a method to leverage existing human efforts in dataset creation to improve data quality.