Lab

Image Retrieval

During the analysis of the requirements of the similar items search (see Section “Search Infrastructure:: Similar items search” in this wiki), the possibility of searching by image was considered. The experiments done on image retrieval and its evaluation can be found here:

D6.1 Advanced image discovery report: https://pro.europeana.eu/files/Europeana_Professional/Projects/Project_list/Europeana_DSI-2/Deliverables/d6.1-advanced-image-discovery-report.pdf
Demo url: http://image-similarity.ait.ac.at/solr/lire.html
Additional internal documents:

Presentation from David with his initial assessment, interesting wrt the vision of the product: https://docs.google.com/presentation/d/1UN0AjqfjKcaHH0dMSzBFnNxNMP5PYdSYgmFiDkeeVrA/edit
M6.1 Advanced image discovery development plan https://basecamp.com/1768384/projects/12623747/messages/65659915
Datasets selected: https://docs.google.com/spreadsheets/d/1iG7zWppd_8T1LtDk61sQZpq2z39bmkryMq93tnr6mKE/edit#gid=0
Brief: https://docs.google.com/document/d/1hRHHLEhD4jSTPvBZo0yiFZntJVBy7GeZGHiRjCgDfPg/edit
Fields and facets being considered: https://docs.google.com/document/d/1z5SGBFw7LT3XqZXNq9zUyEQjjemzztRgr9kexcK9HbQ/edit#heading=h.n1eny9r04dkd
References: https://app.assembla.com/spaces/europeana-apis/documents/difOFmuber54k-dmr6bg7m/download/difOFmuber54k-dmr6bg7m
Related evaluation work: feedback from the pilot evaluation task: https://basecamp.com/1768384/projects/5774755/messages/69246838 (very brief description in Analysis.docx page 5 under Q5 section.)

Named Entity Recognition

The use of Named Entity Recognition (NER) in our collection is a way of enrichment that can contribute to reduce the problem of multilinguality, as well as increase the precision in the search. A review of the NER tools and an experiment to compare the performance of four different tools has been conducted and can be found in the following internal documents: