Search Engine Characteristics: Running multiple strings of search terms revealed issues regarding the choice of boolean or "natural language" types of search engine. In each case, a search with three terms yielded nothing. This lead me to the conclusion that the searches were boolean, i.e.., if each search term was not associated with at least one image, the search would not find anything.
RecommendationA searcher from the general public might want the option of a "natural language" type of search engine so that if some of their search terms (but not others) corresponded to an image, they would have the option of sifting through those images. I thought a system similar to the one in Lexis (a legal research database) might be helpful where search results are statistically ranked according to desired search terms.
The ideal search engine might also accommodate technical users. Those with the training and experience to utilize advanced search tools should have access to those tools.
Search Results There was a large disparity among the various MESL sites regarding the amount of images recovered.
Recommendation One problem I have noted with searching data in a database form is that one can never be sure whether the search terms used are recovering all of the desired data. This general problem is apparent in the MESL sites where the same search strings yield different search results. In general, there seems to be two means to make large amounts of data accessible to the user: "pre-indexing" and search engines. The drawbacks to a search engine seem to be that the utility varies with the skills of the user and the fields used to index the images.
An example of a "pre-indexed" system would be an index to an encyclopedia. In this type of system, the information is organized under subject headings and sub-headings. I refer to this type as pre-indexed because prior to the searcher approaching the data, someone has chosen headings and sub-headings under which to organize the data. Although one could search the same data using a search engine, it helps many users to be able to see the headings and sub-headings under which the images are indexed.
It seems that the MESL sites might benefit from a pre-indexed system such as the one found in Yahoo. Such a system would pre-index the data as well as provide a boolean search engine (possibly a natural language search engine as well). The Michigan site seems to come the closet to this ideal in that Michigan allows the searcher to link to a page where you can view all of the search terms encompassing a particular field. This is very helpful in that it allows the searcher to tailor their query in accordance with the way the data is organized.
Illinois: 1) Illinois provides a useful help & explanation button which is linked to search tips. 2) Site had the ability to search across all of the museums or just a particular museum's holdings. 3) Although the Illinois images were sorted according to 32 possible fields, the search engine only searched across 5 fields (creator name, creator culture, object title, object type and subject). The search results were displayed using 7 fields which included "date" and "credit" as well. 4) Images included thumbnail, medium and full-size images. Nice layout with good sized thumbnail results and brief information regarding each record.
Michigan: 1) Site had the ability to search across all of the museums or just a particular museum's holdings. 2) 10 searchable fields (title, creator, culture or nationality, role, begin date, technique, medium, style, period, subject and type). 3) You can also delineate which fields you would like to search across. 4) Michigan allows the searcher to link to a page where you can view all of the search terms encompassing a particular field. This is very helpful in that it allows the searcher to tailor their query in accordance with the way the data is organized. 5) Results are shown through thumbnails. 6) Choice of S, M, L, or XL images when selecting a thumbnail. 7) Images are very high quality although L and XL are to large to be viewed on a 15" monitor (must be very nice on a large monitor however). Search Results 1) Poverty (4) 2) New Orleans (22) 3) Colorado (10)
American University 1) 10 searchable fields (creator/maker, object title, object type, begin date, subject, style-period, culture/nationality, creation place, creation technique and material/medium. 2) Results are indexed according to the museums in which the images originated. Search Results 1) Poverty (2) 2) New Orleans (1) 3) Colorado (30)
Cornell 1) Cornell allows you to delineate the number of images to appear on each page. This is a nice option because you can have all your images appear on one "long" page rather than having to flip back and forth between pages. 2) Thumbnails displayed with a high resolution option. 3) The high resolution option did not appear to be any better than Michigans although both were very nice. 4) When a thumbnail is chosen it also displays additional information about the image. Searches Results 1) Poverty (2) Including "Savage and Train" 2) New Orleans (11) 3) Colorado (10)
University of Maryland 1) Less information on the opening screen than the other sites. 2) Images are displayed as a link, then thumbnail, then very large image. Might have been nicer if it was displayed as a thumbnail then a choice of images. Search Results 1) Poverty (1) (Savage) 2) New Orleans (none) 3) Colorado (none)
University of Virginia 1) The images can be displayed in three ways 1) thumbnail with brief record, 2) thumbnail with "checkbox" or 3) brief record only. Search Results 1) Poverty (4) 2) New Orleans (21) 3) Colorado (31)