Further research and design considerations

Research questions:

• What a new researcher thinks they’re searching (full text; webpages; records for materials/objects; digitised copies)

• What people think will happen when they click on a subject, genre or contributor tag

  • Does it expand search but only in that category?

  • Does it narrow search to {my search} + {oil paintings}

• What constitutes a ‘failed’ search for a user? One hypothesis is that too many results may result in the user leaving before they’ve got what they need (fatigue, not enough time to look through the results list)

Feature ideas:

• Highlighting matched search terms within work

• Browsing a list of sequential reference numbers

  • WMS: “There is a left anchored browse in the current library catalogue, and I use it a lot, e.g. to get up the photographs around V 27687: enter V 2768 and browse the results http://catalogue.wellcomelibrary.org/search~S5/o?SEARCH=v+2768&searchscope=5 Until recently it was possible to do something similar in the Collections search (then called Works) by entering a truncated search V002768*. That was also possible in MIRO and was very useful, but that option has now disappeared. Unlike MIRO, the results in Works did not always appear in numerical order.”

• Surface authority records as part of name searching

• Designing search result ‘cards’ not only for works, but also contributors, subjects, genres; and how these appear in a list of search results

• Concepts, genres, etc - deliver aggregation page rather than full list in search results

• Deliver the collection-level record as a search result if query includes people, orgs represented in archives

• It’s very niche – but would be useful to be able to search on ISBN & ISSN

• Searches relating to corporate bodies can be confusing – this is due to the data & versions of names, not necessarily the search, but if there are ways to display all options that would be good. As an example, The National Health Service springs to mind.

• Aggregation pages should include all relevant filters (date filter and sort options most important)

• Sort by reference number, accession number (sequential)

• Better handling of zero results

Further feedback

The phrase "by or about" appears a couple of times, but, if we start with persons, at the moment we have three options: find works by X; find works about X; find any mention of X. That's very useful.

For instance in the Wellcome catalogue we have:

• works catalogued as by Hippocrates and not about him: 1,161 works

• works catalogued as about Hippocrates and not by him: 769 works

• any mention of Hippocrates: 2,346 works

When we're looking for works about Hippocrates, we don't want to have trudge through over a thousand editions of his works as well (they are not "relevant"). But when we're looking to see what Wellcome has by Hippocrates, those 1,161 editions that were irrelevant before are relevant now, but this time, we don't want to trudge through all the 769 critical and historical works as well (they are not "relevant").

So some way of linking the search terms to the facets in which they appear (author; name-as-subject; title-as subject, since there are at least two journals called Hippocrates) seems to me essential to relevance. It's not just a matter of getting certain words higher up than other words in a list of results.

The same applies to titles: sometimes you want works which have that title (the innumerable editions of Hamlet), sometimes works about works which have that title (the even less numerable works about that play).

In the CONTRIBUTORS section [p. 6] "Where there are multiple contributors with similar names, creation date used as secondary criteria to disambiguate." I can't see this working, as the creation dates of the works are not given, only those of the manifestation, which may be long after the author's death. The access points for the contributors usually have some disambiguation features already e.g. life dates, LCNAF number, etc.