利用人工智能改进数据发现——挪威国家数据门户.pdf

1、Norways National Data Portal improving data discovery usingAI Data.norge.no is the national data portal of NorwayThe database consists of RDF-descriptions compatible to specifications and standards(i.e.dcat-ap-no)In this presentation I will demonstrate how and why we turned to using Artificial intel

2、ligence(AI)and Large Language Models(LLMs)Introduction Datasets are often described using a field specific jargon.Users looking for datasets did not always know the terms used in the descriptions of the datasets they were interested in.Users did not know what data they needed but often knew what the

3、y wanted to use the data for.Background what were the user needs?What alternatives did we consider?Make the data owners improve their descriptionsTune relevance of elastic search engineImprove documentation and information about the searchExperiment with Artificial intelligence We wanted more experi

4、ence with AI The Minister of Digitalisation and Public Governance in Norway expressed that she wanted more of the public sector to use AI.LLM technology works well to roughly navigate and extract information from large text hubs using natural language.Short way from idea to market.Possible to alter

5、as we gathered more insight and knowledgeWhy AI?Why LLM search?Worries before we startedCostQualityLack of AI-competence Security and privacy Cross sector cooperation and initiative We had knowledge about our database.We had well structured metadata Others had time and resources to experiment with A

6、I.What made it possible to start experimenting?Elapsed time:3 weeks.Intentionally made with the same tech stack we already use.Cost:170 EUR/month of which 90%is cost related to Postgres database.Limitations:Based on one single extraction from the dataset catalogue.No dynamic update of data.The proto