AI search tools like Semantic Scholar and the gaming search engine Splore promise to extract meaning beyond the entered keywords, aiming to understand user intent. This is particularly crucial for researchers seeking specific concepts.
The potential of AI, particularly Natural Language Processing (NLP) and Semantic Machine Learning, is significant. When combined with traditional methods, these can derive concepts effectively. However, their success relies on robust metadata, including consistent subject headings and detailed abstracts or full texts.
The current landscape of scholarly metadata is fragmented, with inconsistent and limited data availability. This restricts AI's ability to accurately determine article concepts. Initiatives like Crossref and OpenAlex are working towards a more open and comprehensive metadata record, which is crucial for enhancing AI search capabilities.
For-profit publishers and indexing databases, including Scopus and Web of Science, are also advancing AI methods. Despite excellent curation, they face challenges due to variable metadata quality and reliance on author-assigned keywords.
To unlock AI's full potential in scholarly search, a unified and interoperable metadata source is necessary. Business models that restrict access to full metadata present a barrier, but the landscape is evolving. Cross-comparison between metadata sources and automated AI methods to enhance metadata are promising developments.
As AI continues to shape academic searching, professionals and researchers need to understand both its capabilities and current limitations. The evolving scholarly metadata landscape holds the key to realizing AI's promise in scholarly research.
This article was originally published by INFORTODAY and has been edited by Glodom.
