Typed-similarity Sample Clauses

Typed-similarity. In addition to the similarity degree between items, we also want to know why two or more items are similar to each other. One of the new features of the second prototype is that the similarity is going to be broken down in types of similarity. PATHS developers have selected seven similarity types: similar author or creator, similar description, similar event or action, similar location, similar people involved, similar subject and similar time period. The first step is to check with real users whether these types make sense. We manually selected pairs of items from Europeana with the previous similarity types and created a gold standard with them. The gold standard is composed of 175 pairs of items, 25 pairs for each of the similarity types. Of course, a pair can be similar in more than one of these similarity types. Previously, we had a gold standard with 295 pairs of items. These pairs may not be similar between them. We are going to include these pairs, both related and unrelated, to the gold standard, in order analyze people’s responses. Now, we plan to use an online survey to ask people about the similarity between two items we are showing. The design of the survey is shown in Figure 1. Note that this survey shown here is a pilot survey, not the final one. Respondents will be able to select different similarity types for the items shown, one or more, or even none. They also will have a textbox area to write any other similarity type they feel appropriate.