Serendipity and readability: Building an engaging online collection site with limited resources
AbstractThe Sarjeant Gallery is a small regional gallery in Whanganui, New Zealand. The Gallery does not have its own Web development or IT staff. The Gallery looked for engaging ways for visitors to browse their collection without assuming a prior knowledge of the collection contents. How could the collection be presented so that general visitors could make the most use of it? What interfaces could be used without requiring the Gallery staff to spend significant time reworking the existing cataloguing records and images? What areas were worth focusing on with the limited time and budget available? With the new website, the Gallery had three broad aims: introduce innovative features, make the site as accessible as possible, and meet current technical best practice. The Sarjeant Gallery staff were open to prototyping and experimentation and worked closely with Vernon Systems to see what might be possible with their collection data. Key fields of basic collection metadata are combined to create natural sounding sentences that are easier to read. Automated analysis of the collection images provided us with further options. The Gallery has received many positive responses to the site. Visitors are delighted by new discoveries as they browse artworks of similar colors and explore based on the subject keywords added by Google's Cloud Vision tool. Interesting connections between works from different artists and periods are emerging based on the colors, shapes, and image orientations detected by the computer vision tools employed on the website.
Keywords: collections, image analysis, computer vision, small museums, mobile
The temporary closure of the main public building for Sarjeant Gallery Te Whare o Rehua Whanganui (the Gallery) made the task of providing online access to the Gallery’s collection a high priority. The Gallery engaged Vernon Systems Limited (VSL) to build a new website to showcase the collection.
Online collection sites can be complex development projects involving bespoke development and organization-specific branding work. The total budget for this project was less than $NZ 20,000 ($US 15,000), so we needed to consider how the site could be built within a relatively small time frame and budget. Online collection websites are changing from being search driven to more generous interfaces that provide a wider range of methods for the visitors to explore a collection, particularly visitors that may not already have detailed knowledge about what the collection contains (Whitelaw, 2015). This paper looks at how we gathered a list of possible features, and developed the website through a series of prototypes.
About the Sarjeant Gallery
The Sarjeant Gallery Te Whare o Rehua Whanganui is located in the lower North Island of New Zealand. The neo-classical purpose-built Gallery opened in 1919 and the collection is one of the most comprehensive surveys of New Zealand art history in a regional center. The collection contains works in all media, including paintings by old masters and contemporary artists, drawings, photographs, sculptures, and ceramics.
Following the 2012 Canterbury, New Zealand earthquakes, the Gallery’s Category I listed heritage building was assessed as meeting only 5% of the current building code and classed as an earthquake-prone building. Plans were implemented to relocate to a smaller temporary site, called Sarjeant on the Quay, to maintain public programs and safety during the fundraising for building redevelopment. The fundraising of NZD 35 million is now complete and the two-year rebuild is poised to begin.
The Sarjeant Gallery has a compact team comprising of 11 full-time equivalent staff with no Web developer. The Gallery did not have the resources available to spend hours retrospectively inputting relevant data and keywords into the collection records. During the redevelopment period there was an urgent need for online access to the collection as the temporary exhibition space is significantly smaller, with less opportunity to provide the public with physical access to the collection. However, as all available funds were channelled towards the redevelopment, no funds were available for an online project.
The aim was to create a site that would be easy to use, fun to navigate, and assumed no prior knowledge of the collection. The Gallery wanted the collection data to be presented in flexible, dynamic ways, enabling the visitor to make unexpected discoveries and encouraging them to linger on the site.
In June 2017 VSL, who already supplied the Gallery’s collection management system, offered to work in partnership with the Gallery to develop an online collection website. Jennifer Taylor Moore had previously worked with VSL for eight years as a systems consultant in Europe and had a well-established working relationship with the team. This made for a very enjoyable process during the four months of experimentation and development.
Research, prototyping and design
The Sarjeant Gallery was open to prototyping during the project and worked closely with VSL to see what might be possible with their collection data, including researching how to make the content more findable (Solas, 2010), and providing options for both collection search and browse features (Brenner, 2015). We explored a wide range of existing online collection sites to gather a list of potential features and user interface options. Sites reviewed included Auckland Art Gallery, Clyfford Still Museum, Cooper Hewitt, National Galleries Scotland, and Victoria and Albert Museum. We also looked at successful online shopping sites, such as Amazon, for examples of search result filtering.
A group of ten users with various backgrounds, interests and ages were involved in the prototype testing. Half of these users were Gallery and VSL staff that were not involved in the development and design, with the other half being external testers. The testers were asked for general feedback on the experience of searching and browsing the collection and were also asked to perform basic tasks to find particular artworks or groups of artworks on the website. The key changes that resulted from user testing were the following:
- adding links to the main Gallery website and the Collection website in the website header
- adding shortcuts near the top of the home page for the various options for accessing the collection. e.g. a shortcut for “Colors.”
- reducing the length of the general introductory text
- adding an option on the home page to view a list of all of the artists on the website
- simplifying the copyright statements on each artwork and adding a link to a more detailed page about the image use that each licence type allows
We also referred to research on types of museum visitors as a starting point. Falk (2009) identified five types of visitors: explorers, facilitators, experience-seekers, professional/hobbyists, and rechargers. We organized the potential features by these visitor types, with the primary goal being to meet the needs of the explorers.
Explorers—motivated by personal curiosity:
- ensure the online collection is easy to browse
- provide simple options to filter the collection to focus on areas the visitor is interested in
- consider color, image orientation, subject, object type, and acquisition period as potential options to interconnect the collection
Facilitators—motivated by other people and their needs (organizing a visit to the gallery):
- highlight artworks that are currently on view in the physical gallery
- connect sections from the main website for planning a visit
Experience-Seekers—motivated by the desire to see and experience a place:
- provide navigation by color
- create new connections from automated subject keywords
- provide sets that guide people to interesting parts of the collection
Professional/Hobbyists—motivated by specific knowledge-related goals:
- provide advanced search options to aid researchers seeking specific knowledge or a particular artwork
Rechargers—motivated by a desire for a contemplative or restorative experience:
- provide a clean, uncluttered website where the artworks are the primary focus
Due to time and budget constraints we decided to defer work supporting the “facilitators” visitor type, as this required more work on the source data and additional work from the Gallery staff to keep the content up-to-date.
The site was built on top of VSL’s online collection application, Vernon Browser, so we started initially with a working wireframe — unbranded templates with no customization. This allowed us to get the data on a test site at the beginning of the project. With the full data immediately available in a basic search and browse interface, we could see both the forest and the trees—areas where the fields held consistent information across the whole dataset, and individual outliers that could be addressed by editing the data or altering the user interface. During the prototyping we went through iterations of exporting and exploring the data, with VSL providing advice on where the Gallery’s limited time could be directed toward for improving the data.
For example, most of the artist records in the collection database included the artist’s place of birth. We found that by adding nationalities to the 40 countries of birth recorded in the system we were able to display nationality as a search option for the majority of the artists. We also selected the artists whose works represented over one third of the collection’s works (approximately 100 artists) and added links to the Wikipedia and Te Ara (the encyclopedia of New Zealand) pages about these artists. Where additional resources exist on the Internet, we should be making an effort to connect to these. “Ultimately, building or maintaining silos will only lead to isolation, and in an increasingly interconnected world, that is the last things that museums and cultural heritage institutions should want” (Connolly & Bollwerk, 2016).
The working prototype was useful for assessing some of the more subjective ideas. For example, we were inspired by the National Galleries Scotland online collection website which uses primary colors from the artwork image as the background color on the artwork detail page. From our testing on the prototype we decided that this was distracting for artworks with strong colors and we did not include this concept in the final design. We also found that it was difficult to meet contrast ratios that met accessibility standards with some combinations of colors.
Another idea that was suggested was to display a graph of the dominant colors for collection artworks in each decade of production. We hoped that this would visualize trends in the selection of colors by artists over time. However, in the prototype, we were able to see that common neutral colors were dominant in the artworks for all periods in the collection. This idea has not been included in current design, but may be included at a later date with the neutral colors filtered from the graph.
Weighing the pros and cons of microsites
We looked at the pros and cons of building a microsite for the online collection versus an online collection fully integrated into the main site. A microsite is independent from the main organization website, so it can be built using different technology and it can have its own navigation options, layout, and branding. A fully integrated site can provide cross content searching, such as a search on “Picasso” leading the user to search results from website blog posts, event listings from an events calendar, and collection records from the institution’s online collection. However, providing options to search and browse collection data is similar to the complexity of building online shopping sites. Incorporating these features into the product used for the main institution website can be more costly than using a separate, dedicated website for an online collection or online shop (Ironpaper, 2016).
We have implemented a microsite as this has allowed us to base the site around an existing working product, and to limit the amount of customization required. In total, approximately seven days of development time were spent altering the branding and layout of Vernon Browser for this project.
Exploring by color and image orientation
Automated color analysis was an immediate candidate for extending access to the collection. Several existing online museum and gallery websites incorporate color as a navigation option (Cope, 2013; Hincliffe & Whitelaw, 2015).
Extracting the dominant colors out of the images was straightforward; however, building navigation based on the colors proved much more difficult than we expected. The challenge was that there were 16 million possible colors in the source images, so any precise color detected in an image was rarely present in any other artwork in the collection. We looked at reducing this to a basic 16 color palette (as you see as a search filter on Flickr, for example), but after testing we settled on the 140 named CSS colors supported by modern Web browsers. Magenta and fuchsia have the same color code, so our final palette has 139 colors.
We used Sven Woltmann’s Java port of Color Thief, an open source color extractor, to extract up to five of the most dominant colors in the primary image for the artwork. We then wrote our own decision tree to find the best match from our 139 named colors for each raw color detected. The palette size is a trade-off: a large palette provides closer matches to the original colors in the images, but a smaller palette allows more images to be grouped together.
We displayed the CSS color names on the site and the color name and color groups (red, green, etc.) were indexed as part of the text for the artwork. This allowed visitors to use color names in their searches, such as “blue house.”
One element of metadata we already had was the orientation of the image. i.e., landscape, portrait, or square. We added this as one of the filter options, and we also indexed the orientation as a text keyword.
Subject tags from Google Vision
Computer vision is moving forward quickly and covers many types of automated image analysis including facial recognition, color analysis, subject keywording, and text recognition. Cogapp, a digital media company working in the cultural sector, has an excellent site presenting their trials of three of the popular image tagging tools: Clarifai, Microsoft Azure Computer Vision API, and Google Cloud Vision API (Hindle, 2017). From reviewing Cogapp’s tests and our own internal testing, we settled on the Google Cloud Vision API. So far, we have only been using the subject tagging feature, but the API provides other features such as detecting text within an image.
The original plan was to make the Vision API subject tags a private field to aid the curators when they create artwork sets to share with the public. However, we were surprised at how good the results were. The tags were not always perfect, but the new connections between the artworks were almost always interesting. The tags were retained on the public detail page for each artwork.
Many of the works in the Gallery’s collection did not have subject descriptions. Adding automated subject tags provided a rich range of new terms to find related artworks. The painting Cattle on a Beach, by Henry Schouten, was a typical example. The tags were mostly correct and enabled us to jump to all of the related artworks that depict “pasture” or “herd.” However, the work was incorrectly tagged as “goat” and “goats.” The consensus was that the additional navigation options the tags provided were still worthwhile. VSL developed options for the curators to disable selected tags on an individual record or permanently delete specific dubious tags across the whole site. Many of the photographic artworks were tagged by Google Cloud Vision as “stock photography”; this was an example of a controversial tag that was deleted across the whole site.
By combining the text keywords provided by subject tags, named colors and image orientation, the site provided new opportunities for exploring the collection. For example, you can search for pictures of a house with the color white in landscape orientation. We wanted users to be engrossed in their browsing of the collection, serendipitously discovering the connections between the works.
Making the whole collection available
By publishing the entire collection online, it is easily accessible and shareable for the first time. No collection catalog is perfect (perfection is the enemy of done), and the decision was made early in the project to publish all of the records for the Sarjeant Gallery’s collection. This is a huge achievement for a regional gallery with a small staff. The Gallery focused on copyright clearance and some minor data clean up, while VSL worked on the technical implementation.
Publishing the entire collection has many advantages (Kapsalis, 2016; Johnson, 2015). There are more artworks for the software to connect, and these connections encourage exploration and serendipitous discovery by the visitors (Walter, 2012). Google Analytics can be used to gain a broader picture of the areas of the collection the visitors are looking at. The volume of online visitors grows as the volume of pages on the site grows. While there is a risk to publishing information which is incomplete or incorrect, having more eyes on the data provides an opportunity to improve the collection catalogue. We were able to use the full catalogue to give a sense of scale within the site. Everywhere we displayed categories, including on the home page, we displayed a live count of how many works fall into each category.
Natural language and readability
Like most internal catalogues, the source data for the Gallery’s collection can appear dry and formal. For the Sarjeant Gallery, a small number of artwork records already had more detailed curatorial descriptions, but for most records, there were only the core keywords (object type, production place, and record type for example) and basic text (medium and measurements for example). It was not feasible to add more descriptive text to large numbers of records with the resources available, (although this is a task for the longer term), so we had to explore ways to make the best use of the existing data.
We were inspired by the Cooper Hewitt online collection site’s use of natural sounding sentences that were generated by their website software (Walter, 2012). The Cooper Hewitt’s website (https://collection.cooperhewitt.org/) uses elements from the collection record such as the object’s production place to create full sentences. In our case we have combined the object type, production place, and production date data into a sentence on the page for each artwork. By doing an automated search on the object type, we were able to display a count of the related artworks of the same type. This met our goal to give the user a sense of the scale of the collection. A typical collection record had data in the following form: “object type: print”; “production place: New Zealand”; “production date: 1982.” The online collection website now presents this as a full written sentence in the following form: “This is one of 866 prints in our collection. It was made in New Zealand in 1982.” Object type and production place both appear as links so that the visitor can easily jump to all of the works in the same category.
We employed a similar approach for the record type to make it clearer what the record represented, and to help surface the links already in the data to related collection records. We hard-coded the sentence version of each record type, with the generated sentence containing the link to the related records. e.g. “archive series” became “an archive series,” and “page from a book” became “a page from a book.”
We wanted the site to comply with accessibility standards from day one. It is often more difficult to change the design after launch, and accessibility compliance can often be forgotten after the rush of the first phase of development.
The site was tested for compliance with Web accessibility standards to meet a WCAG 2.0 AA standard. We used the Wave browser plugin (WAVE, 2017), so that compliance could be checked during the development work within our internal development environment. We also used WebAIM’s contrast checker (WebAIM, 2017) to check contrast ratios for all of the key colours in the branding. The compliance checkers helped us spot problems such as missing image alt (alternative) text. Image alt text is used by screen readers for users with impaired sight and it is also used by search engines such as Google and Bing to help determine the subject of an image. We also considered the wording used throughout the site, such as on buttons and in links.
Create Once, Publish Everywhere (COPE)
The site uses responsive templates to provide layouts for mobile, tablet, and desktop screens. Through these templates, the same data can be displayed in the most appropriate way for the device screen size (Ludden, 2013). For example, on smaller screens we display smaller image sizes (this also helps with speed over slower connections) and we changed the menu into a collapsed single-level list.
We used standard HTML metadata tags and Open Graph metadata tags to help search engines index the site and improve sharing on social media sites (Moth, 2013). Open Graph tags ensured that the correct image, title, and description were copied when users shared a page on social media. Social media sharing buttons were included on the artwork and artist details pages to streamline sharing.
Lastly, the website was built on top of an API to provide potential options in the future for data re-use and new interfaces. The API is used to share the records with New Zealand’s cultural heritage aggregator, DigitalNZ, thereby providing another location for visitors to discover the collection. By publishing the entire collection in a form that is useable on many device formats, is easily shareable, and is accessible from multiple sites and search engines on the Internet, we embraced the concept of “Create Once, Publish Everywhere”(Collections Trust, 2016).
No dead ends
The search results page provided options for filtering the results, including by image orientation and color. Apache Lucene was used to retrieve search facets, providing live counts of the number of records matching each of the displayed categories. The search filters made it easy to browse on subsets of the collection. However, for the more serious researcher, it was still useful to get a complete list of the terms available in key fields on the site. With that in mind, we included an advanced search screen to display all of the options for fields such as production place, artist nationality, medium, and object type.
We believe that an online collection site should not exist in a vacuum (Connolly & Bollwerk, 2016). Two hours of work went into cross-checking artists against pages on Wikipedia. Wikipedia pages were found for many of the artists that were well-represented in the collection. We also added links for artists with biographies in Te Ara (the encyclopedia of New Zealand), and links to a handful of other online resources. Gallery staff will be adding Wikipedia pages for some of the artists which are not documented on Wikipedia, and are adding links from Wikipedia back to the Sarjeant Gallery’s artist pages to help people discover the website.
The options to find related records by color and Google Cloud Vision subject tags provide ways for the user to explore further. So far the Google Analytics statistics look promising, with a greater than three minutes average session time and an average of five pages viewed per session.
The Gallery is thrilled with the overwhelming and positive responses they have had from visitors to the site. From informal discussions with visitors and the review of Google Analytics visitor statistics, the feature that people most seem to enjoy is searching by color. Visitors love the ease of use, the dynamic filter options, and the serendipitous discoveries they can make. The Gallery has actively promoted the site via social media, regularly posting pictures of some of the discoveries that can be made by searching on a particular key word or color as well as topical searches to do with current events; for example, the “Movember” movement by searching the keywords “facial hair.” The Gallery has sent out press releases to the media and had articles published in local newspapers and online, and they have been speaking regularly about the new site to their volunteers and various stakeholder groups. At New Zealand’s recent annual National Digital Forum conference for the gallery, library, archives and museum sector, the Gallery won the award for Digital Exhibition or Collection, drawing significant national attention to the site.
Gallery staff are currently working steadily through requesting copyright permission from artists and copyright holders to increase the number of images that can be displayed on the site. Every artist approached so far has responded positively. One artist has temporarily suspended her agreement while she seeks permission from her subjects first.
Because the Gallery is operating out of a temporary site, and not at their 98 year old heritage building, visitors can sometimes be disgruntled by the time they arrive because they have often tried to visit the closed building first. Recently a couple visited wishing to view a specific painting and expected to see it displayed on the wall. They were deeply disappointed to hear that it was not currently on display (though it had been displayed relatively recently) and wished to view it regardless. The Curator of Collections was in a meeting and unable to give them immediate unplanned access to the collection store. Instead, front of house staff offered them a virtual viewing of the work online and they departed much happier, having been able to see and discuss the work. The Gallery is currently in the process of setting up a kiosk inside the gallery which they hope will provide Gallery visitors access in this type of situation.
Next steps for the website
- Add further collection highlights, including having one for items currently on display;
- Continue to work through copyright permissions to increase the number of images that can be displayed;
- Add to the underlying data, creating meaningful interpretive text to each record, while continuing to improve the consistency of the data overall;
- Identify the subjects in portrait images that contain Māori subjects, then request permission from them or their descendants to display the images;
- Look at ways to encourage visitors to look beyond the home page by regularly changing content on the home page (explore the collection);
- Investigate additional features, such as quantitative displays of data for aspects such as acquisition decades, or visually demonstrate trends over time, such as the percentage of representation of women artists over collection decades;
- Improve support for the “facilitator” visitor type, who may be using the online collection to plan a visit. Possible changes include highlighting the works currently on display and adding links on the online collection website to Web pages about visit planning on the main website.
A working prototype with a full set of collection records allowed us to identify fields that worked well for browsing and searching, and to direct the Gallery’s limited resources to the areas of data that could be quickly improved. We recommend using an agile development process for online collection development projects, as users can test design concepts early in the process and changes to the design can be made more easily (Hegley, 2016).
Building the online collection as a stand-alone microsite simplified the project. The user interface for an online collection should cope with the data volume and complexity and should ensure that the content is findable and usable. A fully integrated site would have provided the visitor with a single integrated experience, but this would have required migrating to different content management system or investing significant time in development work.
Automated image analysis creates opportunities to search based on data already embedded in the images, such as color ranges and image orientation. However, we found that mapping colors to a smaller, more useable palette was not easy with the tools we tested. Generating subject keywords provides a different range of keywords than would typically be added by internal catalogers, and proves to be a huge advantage for organizations without the resources to do this manually. The keywords are not always exact, so organizations should factor in time to review at least a selection of the generated keywords.
By publishing the entire collection and embracing the concept of “Create Once, Publish Everywhere,” this new site allows the Gallery to highlight collection items or sets that tie with current events and internal exhibitions. Full online access to the collection is generating increased positive exposure for the Gallery, providing new statistics on where visitor interest has been drawn, and providing the public with digital access to artworks which they do not have physical access to. The site can be accessed at https://collection.sarjeant.org.nz.
Auckland Art Gallery website. (2017). Consulted September 1, 2017. Available https://www.aucklandartgallery.com/
Brenner, S. (2015). “Reconsidering searching and browsing on the Cooper Hewitt’s Collections website.” MW2015: Museums and the Web 2015. Published February 1, 2015. Consulted January 11, 2018. Available https://mw2015.museumsandtheweb.com/paper/reconsidering-searching-and-browsing-on-the-cooper-hewitts-collections-website/
Clyfford Still Museum online collection website. (2017). Consulted September 1, 2017. Available https://collection.clyffordstillmu seum.org/
Collections Trust. (2016). “Create Once, Publish Everywhere (COPE).” Last updated 2016. Consulted January 11, 2018. Available http://collectionstrust.org.uk/resource/create-once-publish-everywhere-cope/
Connolly, R. & E. Bollwerk. (2016). Positioning Your Museum as a Critical Community Asset: A Practical Guide. Lanham, Maryland: Rowman & Littlefield.
Cooper Hewitt online collection website. (2017). Consulted September 1, 2017. Available https://collection.cooperhewitt.org/
Cope, A. (2013). “All your color are belong to Giv.” Last updated February 13, 2013. Consulted September 1, 2017. Available https://labs.cooperhewitt.org/2013/giv-do/
Falk, J. (2009). Identity and the Museum Visitor Experience. Walnut Creek, California: Left Coast Press Inc.
Hegley, D., M. Tongen, & A. David. (2016). “The Agile museum.” MW2016: Museums and the Web 2016. Published January 15, 2016. Consulted January 31, 2018. Available https://mw2016.museumsandtheweb.com/paper/the-agile-museum/
Hincliffe, G. & M. Whitelaw. (2015). “Colouring digital collections: Challenges and opportunities for the use of colour metadata in cultural collections.” MWA2015: Museums and the Web Asia 2015. Published August 14, 2015. Consulted January 10, 2018. Available http://mwa2015.museumsandtheweb.com/paper/colouring-digital-collections-challenges-and-opportunities-for-the-use-of-colour-metadata-in-cultural-collections/
Hindle, A. (2017). “Automated image analysis with IIIF.” Last updated Tuesday June 20, 2017. Consulted January 11, 2018. Available https://blog.cogapp.com/automated-image-analysis-with-iiif-6594ff5b2b32
Ironpaper. (2016). “6 Benefits of Microsites.” Last updated Friday October 28, 2016. Consulted January 11, 2018. Available http://www.ironpaper.com/webintel/articles/6-benefits-of-microsites/
Johnson, K. (2015). “No Detail Goes Unnoticed When Art Is a Click Away.” Last updated January 29, 2015. Consulted January 11, 2018. Available https://www.nytimes.com/2015/01/30/arts/design/art-museums-are-increasingly-adding-their-collections-online.html
Kapsalis, E. (2016). “The Impact of Open Access on Galleries, Libraries, Museums, & Archives.” In Smithsonian Institution Archives. Consulted January 11, 2018. Available https://siarchives.si.edu/sites/default/files/pdfs/2016_03_10_OpenCollections_Public.pdf
Ludden, J. (2013).“Responsive Design: How Museums Can Thrive in the Universe of Desktops, Tablets, Smartphones, and More.” In N. Proctor & R. Cherry (eds). Silver Spring, MD: Museums and the Web. Published November 6, 2013. Consulted January 11, 2018. Available https://mwa2013.museumsandtheweb.com/paper/responsive-design-how-museums-can-thrive-in-the-universe-of-desktops-tablets-smart-phones-and-more/
Moth, D. (2013). “Three types of social structured data and why you need them.” Last updated April 16,2013. Consulted January 11, 2018. Available https://econsultancy.com/blog/62540-the-three-types-of-social-structured-data-and-why-you-need-them
National Galleries Scotland website. (2017). Consulted September 1, 2017. Available https://www.nationalgalleries.org/
Solas, N. (2010). “Hiding Our Collections in Plain Site: Interface Strategies for “Findability.” In J. Trant and D. Bearman (eds). Museums and the Web 2010: Proceedings. Toronto: Archives & Museum Informatics. Published March 31, 2010. Consulted January 10, 2018. Available http://www.archimuse.com/mw2010/papers/solas/solas.html
Victoria and Albert Museum website. (2017). Consulted September 1, 2017. Available https://www.vam.ac.uk/
Walter, M. (2012). “Getting lost in the collection (alpha).” Last updated October 2, 2012. Consulted August 20, 2017. Available https://labs.cooperhewitt.org/2012/lost-collection-alpha/
WAVE. (2018). “WAVE Browser Extensions.” Consulted September 20, 2017. Available http://wave.webaim.org/extension/
WebAIM. (2017). “Color Contrast Checker.” Consulted October 1, 2017. Available https://webaim.org/resources/contrastchecker/
Whitelaw, M. (2015). “Generous Interfaces for Digital Cultural Collections.” In Digital Humanities Quarterly: Alliance of Digital Humanities Organizations. Published September, 2015. Consulted September 20, 2017. Available http://www.digitalhumanities.org/dhq/vol/9/1/000205/000205.html
Rowe, Paul and Taylor Moore, Jennifer. "Serendipity and readability: Building an engaging online collection site with limited resources." MW18: MW 2018. Published January 14, 2018. Consulted .