Search

This paper presents a case study on Early Medieval burial practices using AGNES, an intelligent search engine for Dutch archaeological grey literature. Traditionally, it is assumed that cremations phased out at the start of the Early Middle Ages, when the inhumation practice became more numerous. However, recent research (Lippok, 2020) shows that cremations might be more prevalent than previously assumed. Due to research efforts being concentrated on furnished inhumations, cremations and other types of burials have not received their share of research interest. It is suspected that unknown Early Medieval cremations may be found in grey literature research reports. The rapidly growing document collection requires more efficient methods to search through this big data, as manual searching is too time intensive. AGNES uses machine learning to allow searching on archaeological concepts (such as time periods and artefacts) in full texts, solving common problems with synonymy and polysemy. This paper describes a controlled search for Early Medieval cremations in the Netherlands, and a comparison of the new information uncovered from the reports to prior knowledge on the topic. The queries resulted in 2541 hits. Twenty-three Early Medieval cremations that were previously unknown to experts were uncovered, and 31 (of 77) known sites were identified. Forty-one possibly interesting documents were noted, where it was not clear from the report if Early Medieval cremations were present. 2446 documents were not relevant to the study for various reasons. The 23 new sites are an increase of 30% over the existing knowledge of experts. In the last 20 years only nine new sites featuring Early Medieval cremations were discovered, so being able to add another 23 is a major development. Adding previously unknown Early Medieval cremations to the dataset challenges the existing bias for inhumation graves, and supports striving towards viewing the Early Medieval burial repertoire as more heterogeneous and representative. This indicates that AGNES is useful for archaeological research, and the uncovered information can lead to a more cohesive view of the past. • We created an intelligent, entity-based search engine for archaeological texts. • The system uses Machine Learning to detect and find archaeological entities. • A case study found 23 new Early Medieval cremations using the system, a 30% increase. • The new cremations shows that they are more common than previously thought. • The results show the system outperforms string search on metadata records. [ABSTRACT FROM AUTHOR]