News

• Paper-Presentation at ICECCME 2025

  Our recent publication, “Silverbullet or Mystery Box – LLM-based Soft Skill Classification in Volunteering”, has been accepted and presented at ICECCME'25. The work was carried out in the context of our ongoing research within the CIvolunteer project.

  Volunteering represents a critical pillar for both essential infrastructures and the achievement of the Sustainable Development Goals (SDGs). In particular, skill-based volunteering plays a key role in ensuring that individuals can contribute effectively and that acquired competencies can be leveraged beyond individual engagements. However, systematically identifying and classifying relevant skills in volunteering contexts remains a significant challenge, as existing approaches are primarily tailored to structured job postings in the labor market.

  This work investigates how established skill classification methods perform when applied to volunteering opportunities, which are typically characterized by unstructured, task-driven descriptions and a strong emphasis on soft skills. Building on this analysis, we propose a novel approach based on cache- and retrieval-augmented generation techniques to enable efficient soft skill classification without requiring costly large language model (LLM) fine-tuning.

  The proposed methods are evaluated using a newly created dataset of volunteering opportunities with expert-labeled soft skills from a global platform. The results, assessed through both quantitative metrics and qualitative analysis, demonstrate that lightweight AI approaches can provide effective and scalable solutions for skill-based volunteer matching.

  This work contributes to advancing intelligent support for volunteer coordination and highlights the potential of modern AI techniques to strengthen the sustainability and impact of volunteer-driven ecosystems.

• Paper-Presentation at ICECCME 2024

  In early November, our colleague Christoph Gassner presented the publication “Towards Goal-Oriented Volunteering – LLMs to the Rescue?” at the 4th International Conference on Electrical, Computer, Communications, and Mechatronics Engineering (ICECCME). This work was developed as part of our CIvolunteer project.

  The sustainability of the volunteer sector is increasingly at risk—partly due to demographic changes—despite the United Nations (UN) identifying volunteers as key actors in achieving the Sustainable Development Goals (SDGs) outlined in the 2030 Agenda. In Austria, where a significant portion of critical infrastructure relies on volunteer work, this development poses a serious threat to the operational capabilities of many organizations. The CIvolunteer project aims to strengthen the volunteer sector and ensure that volunteers are sustainably matched with organizations, taking their individual preferences into account.

  The presented study examines the relationship between volunteers’ personal goals and suitable activities. In the first step, we analyzed the nature of personal goals and volunteer activities across various leading regional, national, and international platforms. The objective was to develop methods to identify activities that align with defined goals. Due to the unstructured nature of the available textual data, large language models (LLMs) were employed to establish connections between volunteers’ goals and available activities. To improve computational efficiency, we also investigated whether a pre-trained Cross-Encoder model could be adapted for this classification task using LLM-generated data. The resulting models were evaluated using standard metrics, descriptive statistics, and statistical tests to assess and quantify differences between the models.

  This work makes a significant contribution to leveraging modern AI technologies to develop sustainable solutions for the volunteer sector.