Is the catalog metaphor dead?

March 2, 2022
Catherine Nicole Coleman
Black and white photograph taken between 1900-1920 of a room with stacks and stacks of catalog cards.


“It is astonishing what a different result one gets by changing the metaphor!”

George Eliot, The Mill on the Floss

We rely on metaphors to acclimate us to new environments and new technologies. But as we become familiar, comfortable, and even demanding of the affordances of technologies, we slough off the old metaphors. Apple eased us into managing our contacts on the computer by using the look of the Rolodex. The Rolodex metaphor was no longer relevant when a list of contacts became more readily understood as a social network. Metaphors also influence technological development. The metaphor we imagine brings with it expectations of how a technology should behave. The social network metaphor, for example, continues to drive feature developments in Facebook, Twitter, Linked In, and similar platforms, even if the underlying business model is closer to surveillance. 

The physical library catalog was both an organizing system and the primary point of discovery for library patrons. But as we have known for some time and had recently confirmed in patron interviews for the new library web site, in the digital age, patrons search Google before visiting the online catalog to find what they are looking for in the library.  Google Search, Google Books, and Google Scholar will not get you to the rich and diverse collections in a research library, but neither will the online catalog, albeit for different reasons. As Deputy University Librarian Philip Schreur explained in “The Academy Unbound: Linked Data as Revolution,” to discover a resource in the online catalog, a bibliographic record for it must be present in the system. But we do not have bibliographic records for many of the resources in the library and as the number of resources continues to grow exponentially, we will not be able to catch up. 

In response to this insurmountable problem, a metaphor that is gaining ground for information discovery is the knowledge graph. The graph model is the anti-silo. It is a response to the needs of researchers who are asking questions and performing analysis within and across collections. The knowledge graph as metaphor brings to mind the image of a network. In some cases, as with the Yewno discovery tool that you can find on the front page of the Stanford Libraries website, the metaphor takes shape visually. While visual networks are compelling and in many cases useful, they can paradoxically hide the underlying abstraction we employ to draw them. As my co-authors and I wrote in The Network Turn, the unbounded rhizomatic structure of the visual network suggests an infinite web of information. When we search the World Wide Web via the Google Search Engine it feels like unfettered discovery and yet is deeply and problematically limited by the way information is flattened.  Context and provenance, upon which we rely to judge the reliability of information, is ignored in favor of the speed and efficiency of getting to a result.

Experiments with machine learning in museums brings context back to discovery while adding the depth and breadth of interconnected content-level search. A museum project in the UK, Heritage Connector: Transforming text into data to extract meaning and make connections developed by The Science Museum with support from AHRC operationalizes the promise of linked data described by Schreur by applying machine learning to collections. As in the library, online discovery in museums is also based on records that are a digital version of the cards in the physical card catalog, a model that pre-dates digitization and is tied to the needs of collection management rather than discovery. The knowledge graph reaches outside of the museum to link records to Wikidata and add entities extracted from text descriptions in existing records.

A few days ago, here in our own backyard, the Computer History Museum in Mountain View demonstrated the results of their collaboration with Microsoft, which they described as cognitive search. (Explore the project: ) The goal of this first phase of the partnership was to make their vast, underutilized, collection of audio and video recordings available, discoverable, and easily explorable online. Two key features of the project are automated transcription and automated translation to make the collections accessible to patrons around the world.  The next step in their plan is to extract named entities from those transcriptions and connect them to external knowledge bases and authorities.


Ahnert, Ruth, et al. The Network Turn: Changing Perspectives in the Humanities. Cambridge University Press, 2020.

Dutia, Kalyan, and John Stack. "Heritage connector: A machine learning framework for building linked open data from museum collections." Applied AI Letters 2.2 (2021): e23.

Fensel, Dieter, et al. "Introduction: what is a knowledge graph?." Knowledge Graphs. Springer, Cham, 2020. 1-10.

Schreur, Philip Evan. "The academy unbound." Library Resources & Technical Services 56.4 (2012): 227-237.