What is the possible impact of artificial intelligence on library automation if it is to be applied?
Artificial intelligence has gained impressive capabilities in many aspects of society. This area of computer science continually makes new breakthroughs, which, for better or worse, introduce technologies able to perform activities of increasing complexity previously possibly only through human intelligence. The realm of libraries has already seen some impact from this type of technology and should expect even more in the future.
The key concept of artificial intelligence surrounds the ability for computers to perform tasks that traditionally rely on human intelligence. This concept goes beyond the automation of routine tasks. A bank ATM, for example, performs actions that are usually performed by a human teller simply by presenting menus of available services and responding according to pre-defined scripts. Robotics can likewise substitute for human labor in performing mechanical or logistical tasks without necessarily imitating human thought. Artificial intelligence goes beyond rote tasks to complex problem solving or other activities that take on characteristics of human cognition.
Computer technology, including artificial intelligence, has already reshaped the nature of library reference services. In former times, the library reference desk was considered the definitive source for factual questions. Resources such as Wikipedia as well as Google and other search engines able to provide instant access to authoritative and non-authoritative documents have siphoned almost all factual questions away from library reference desks. The latest wave of technologies layers advanced technologies from the realm of artificial intelligence into the mix. Digital assistant services—such as Google Home, Amazon's Alexa, and Apple's Siri—take advantage of voice recognition, natural language processing, and machine learning to tap into information available on the web and proprietary data repositories to respond to ever more complex requests. While these services abound with concerns related to privacy, they nevertheless have further solidified a technical substitute to the library reference desk for routine fact-based questions. Persons with sensitive issues would hopefully seek out a librarian or other professional committed to safeguarding privacy.
Diverting routine factual questions from library reference services isn't necessarily a major loss. Many libraries instead channel their efforts into more in-depth information services for their clientele and other areas of strategic involvement with the communities they serve. Many librarians spend more time providing expert assistance with research projects, conducting bibliographic instruction or research methods sessions, or other activities beyond the reach of computerized assistants. A strain of artificial intelligence that also has made inroads into the library domain is machine learning. A crucial aspect of library work relates to the ability to organize, describe, and provide access to large bodies of content. Machine learning describes a type of artificial intelligence where computers refine or perform new tasks based on processing a body of data. Traditional computer programming follows an algorithm to process and analyze data. With machine learning, the computer changes its behavior based on the data.
Library discovery services represent an interesting use case of machine learning. The common approach for discovery services today includes the creation of massive central indexes created through the processing citation or full text, representing some approximation of the totality of the body of scholarly communications. Publishers and aggregators provide copies of the materials in their content products to discovery service providers that then ingest them into their central index. These discovery services can then be searched by library users to gain access to the original materials available on the publisher's delivery platform. This brute force indexing of citations does not fall into the realm of artificial intelligence since it is performing a routine task—albeit at large scale—taking advantage of standard search and retrieval technologies.
A new product from Yewno provides one example of using machine learning to support the search and exploration of large bodies of documents based on related concepts rather than keyword matching and relevancy ordering. Yewno has developed a process based on machine learning to not just index keywords, but to identify and extract concepts from documents. It is not merely finding matches of words in a text to a pre-defined ontology, but rather identifying concepts based on their semantic context within a document and linking those concepts to occurrences throughout a corpus of material. (Smart Libraries Newsletter featured Yewno in its April 2017 issue).
It is likely that artificial intelligence will make its way into many other aspects of traditional library activities over time. In some areas, such as broad discovery, we can expect artificial intelligence to result in new tools and interfaces with capabilities not previously possible with traditional search and retrieval technologies. I anticipate that various aspects of artificial intelligence can be tapped to improve the accessibility of large-scale digital collections. Machine learning can power automated video description systems to facilitate the creation of search and retrieval systems for large digital video collections with more sophistication and at a scale not affordable solely through human processing.
Artificial intelligence should not always be seen as a threat to skilled workers. Tools based on artificial intelligence can also supplement the work of librarians. I would anticipate that new tools will be developed to create metadata to describe library resources. I see these kind of tools not as replacing the role of catalogers but as a means to amplify their work. Automated tools may be able to make a first pass at resource description; however, in most cases this work will require expert human intervention to ensure expectations related to following appropriate standards or neutrality and objectivity. I believe that it will be quite some time until libraries will be able to rely on entirely mechanical processes to create highquality metadata to describe their core collections.
Artificial intelligence poses both a threat and opportunity for libraries. When approached proactively, these tools and technologies can be explored and exploited to facilitate library work and strengthen their position. On the other hand, some may see artificial intelligence as a way to eliminate or bypass libraries in providing information services. It will be important for libraries to continually assess each aspect of their work and shape services in ways that provide value beyond what might be delivered instead through technologies driven through artificial intelligence.