To find records that are similar to the currently opened record click the Find Similar Records button in the top left corner of record view. A new browser tab will open displaying up to the results with full functionality available to export, visualize, refine with include/exclude functionality, and more.
To return the most similar records to the currently opened record an algorithm is used. The algorithm reviews the publication title and abstract, to find keywords in each field that are relevant to the current record. An internal search is then conducted using those keywords to identify other records containing at least 40% of those keywords by default. Lastly, the records compiled from the internal search are analyzed for relevance using Lucene BM25 algorithm. The records that are most similar are then displayed in the Similar Records table, which opens in a new browser tab.
To modify the amount of similarity between the current record and those found in Similar Records, click the Modify button, depicted as a pencil on the similar records breadcrumb. Use the sliding scale to select the degree of similarity between 0.1 and 0.8, then click apply changes. At 0.1 (which equates to 10% of the identified keywords in the current record) the results will be very broad and at 0.8 (which equates to 80% of the identified keywords in the current record) the results will be near duplicate.