Maratsi Maria Ioanna

Abstract:
Open government data has been in the spotlight for several years, emphasising the need for publicly available datasets to foster societal innovation and enable da-ta-driven decision-making. However, it is not sufficient for data to be available through an open data portal or repository to be usable by end users and for them to fully unveil its potential; it also needs to be accessible and intelligible. Target-ing this aspect, this study builds on the synergy of open government data and emerging technologies, such as Large Language Models (LLMs) to demonstrate that technologically enabled open data portals can have enormous possibilities for end-users to understand an available dataset efficiently, and in an interactive man-ner. This implementation is built on the LLMs Retrieval Augmented Generation (RAG) functionality which fetches datasets after a user’s choice from the Greek Open Data Portal, allowing the user to ask questions about this dataset using nat-ural language. The Greek open data portal provides access to datasets from dif-ferent domains through the Application Programming Interfaces (APIs) which makes it easier to bridge the gap between complex data and non-technical users who want to use it. This work explores the implementation potential of such a pipeline in a real-world application using the Greek Open Data portal as data source, and by utilising a conversational smart agent to interact with the open da-tasets available through the portal, it brings forward new capabilities in data-user interaction and efficient data exploration. Apart from the implementation, this study explores the pitfalls and shortcomings of such an endeavour (conversation-al agents) which requires real-time calculations and deterministic responses.

Abstract:
As a novel Artificial Intelligence (AI) application, ChatGPT holds pertinence not only for the academic, medicine, law, computing or other sectors, but also for the public sector-case in point being the Open Government Data (OGD) initiative. However, though there has been some limited (as this topic is quite new) research concerning the capabilities ChatGPT in these sectors, there has been no research about the capabilities it can provide to government concerning its wide range of functions and activities. This paper contributes to filling this gap by investigating the capabilities that the ChatGPT can provide concerning one of most recently initiated and novel, and at the same time most promising, activities of government that aims to fuel the emerging data economy and society: the opening of large amounts of government data; furthermore, we investigate the public values that can be promoted through the use of ChatGPT in the area of OGD by both the data publishers as well as their users. At the same time, we investigate the issues that the use of ChatGPT in the area of OGD can pose, which can reduce the capabilities identified as aforesaid as well as the benefits and public values that can be generated from them. For these purposes interviews with 12 experts have been conducted and their responses have been analyzed. Finally, based on our findings we have developed a research agenda concerning the exploitation of ChatGPT application in the OGD domain.