Virtual reference librarians (Chatbots)
Joseph Vincze
Introduction make you feel”, creating a simple IBM Watson’s cognitive computing
illusion of understanding. It was the service has been applied to various
Chatbots go by many different first attempt at human– computer commercial uses, but one area of
names: digital assistants, intelligent interaction. Some non-technical staff particular interest has been in
agents or virtual agents. Overall a who used ELIZA thought they were healthcare – enabling medical
chatbot is a conversational agent that talking to a “real” therapist and professionals to choose the best
conducts conversations via text or voice revealed personal information to the treatment plans for individual patients
commands. They are touted by the machine. Overall, this early chatbot with cancer. It would not be a stretch to
media and scholarly journals as the next was limited, being only able to respond assume that Watson’s ability to gather
digital transformation, redefining how to you with a question, triggered only and process vast amounts of
businesses will engage and delight by keywords and not very intelligent information can be used by libraries to
consumers. As computer programs are (Rubin et al., 2010). improve customer interaction. A library
capable of interacting with humans and patron can type in a search request, “I
In 1951, Alan Turing proposed a test
acting on their behalf, they are often am looking for XYZ”. Through a
to gauge computer intelligence in
anthropomorphized with personalities Watson-fueled search of the library’s
which a human being should not be
in the following commerce-related resources and through the use of
able to distinguish a machine from reference interview questions, Watson
applications: another human being based on the would make suggestions based on the
replies to questions put to both. The patron’s responses.
● customer relations;
criterion for the Turing Test is that a
● customer service;
computer is mistaken for a human more Types of chatbots
● ordering service; and
than 30 per cent of the time during a
● personal consultant.
series of 5-min keyboard conversations. There are two types of chatbots. The
On June 7, 2014, a milestone was first type functions based on
Currently, people are using messaging
achieved when a chatbot called Eugene predetermined rules and tends to be
apps more than they are using social
Goostman, programmed as a non- very limited. These chatbots are only as
networks. Messaging is at the heart of
native English speaker with the persona good as they were programmed to be
the mobile experience, and chatbots are
of a smart-alecky pre-teen male living and can respond only to very specific
particularly well suited for that
in the Ukraine, convinced 33 per cent of commands. If your response is outside
experience. A benefit of using chatbots
the judges at the Royal Society in the parameters of the commands, these
is that consumers will be able to types of chatbots will not know what
interact naturally in a conversational London that it was human – passing the
Turing Test. We are at the early stages you mean and reply unsatisfactorily.
way to obtain relevant information or to The second type functions based on
carry out a task, leading to a more of true artificial intelligence (or AGI –
Artificial General Intelligence). At this machine-learning algorithms. Using
humanized and personalized artificial intelligence and natural-
experience. Chatbots have been around time, the majority of AI systems are
designed to focus on specific domains, language processing, this type of
in various forms for years. We are chatbot is capable of understanding the
currently in the early phase of a chatbot thus allowing them to specialize in
context of what is being said, not just
revolution. This article will explore things and know them extremely well.
specific commands. This type learns
integrating chatbot applications into Some notable milestones in artificial
from experience, getting smarter with
libraries to improve reference services. intelligence (A.I.) are the following: each conversation it has with people
Chatbots have been around since the (Schlicht, 2016).
1960s. MIT Professor Joseph ● 1997 – IBM Deep Blue defeats
Weizenbaum created an early Natural Garry Kasparov – reigning world Adopters of chatbot technologies
Language Processing computer chess champion.
program named ELIZA which ● 2011 – IBM Watson becomes the There are two demographics that
emulated a Rogerian psychotherapist. Jeopardy-winning supercomputer. are early adopters of chatbot
Named after Eliza Doolittle in ● 2016 – Google AlphaGo defeats technology: Millennials and the
Pygmalion, ELIZA was designed to South Korean Go grandmaster Chinese. The Chinese leapfrogged
hold simple conversations prefixed with (Google claims that their system is directly into mobile, skipping desktop
psychotherapy phrases such as “tell me not pre-programmed – that it learns computing altogether with its
more about that” or “how does that from experience). browsers and drop-down menus.
LIBRARY HI TECH NEWS Number 4 2017, pp. 5-8, © Emerald Publishing Limited, 0741-9058, DOI 10.1108/LHTN-03-2017-0016 5
Chinese businesses recognize this fact chatbots can improve productivity and library’s visibility in the user
and often launch digital presence with be cost-effective, freeing library staff community.
a chatbot on WeChat while skipping a to focus on more demanding research ● The library can use the chatbot’s
webpage entirely. WeChat has and tasks uniquely suited to humans. conversation log to track the types of
penetrated all age groups in China. Chatbot conversations can be questions and resources requested to
For Millennials, chatbots are a natural combined with actual humans to allocate resources for library patrons.
extension of texting. Millennials, answer questions that are more
having grown up with the internet and complex – if the chatbot cannot help
responsive on-demand services, often Case study: Emma the MPL Catbot
the client, it will forward the question
cite phone calls as being overly to a librarian.
intrusive, even presumptuous – thus “Emma the MPL Catbot” (Figure 1)
● Chatbots have the potential to be a
their preference for texting. was developed to address proposed
better information delivery system
According to a 2014 Gallup poll, state budget cuts to Ohio’s public
than other forms of customer
Millennials use their phones to interaction, providing a livelier libraries. From 2009 until 2012, Emma
message far more than anything else. interactive experience than a search provided immediate information about
In this context, it makes sense that engine. the Mentor Public Library and
chatbots will serve as an efficient ● Customers can interact with a answered common reference questions
alternative, if not outright chatbot to find what they need from MPL’s 36,000 patrons. If Emma
replacement, for apps on mobile without interpreting and navigating could not answer a user’s question in a
devices. No longer having to webpages, thus simplifying their single response, she would guide or
download and register each new app access to information. Responsibility prompt the user to clarify their
to access a new service, mobile of locating requested information question, simulating a reference
messaging users will simply send a shifts from the user to the interview. The user would then be sent
text to their chatbot of choice to do a programmer of the chatbot. to the most likely library online
variety of activities. In fact, according ● Chatbots can create an even stronger resource based on keywords in the
to Dunn (2016), the number of such bond with young audiences. Moving user’s answer (Mentor Public Library
apps have peaked. A quarter of all the access of information from the and Pandorabots, 2017).
downloaded apps are abandoned after browser and mobile app to a chat Emma’s knowledge base contained
a single use. The average smartphone interface on their mobile devices that over 30,000 categories, with each
owner downloads almost zero new this generation is already familiar category consisting of an input
apps per month. They spend the with results in convenience and question, an output answer and an
majority of their time in just three engagement. optional context. User conversations
apps, and the amount of time people ● Chatbots can answer simultaneous
spend using those apps is stagnant with Emma were divided into three
questions from multiple users in a general categories (Iglesias, 2013):
(Dunn, 2016). manner that is consistent and
instantaneous. 1. General questions about the library –
● Chatbots are unruffled by rude representing 40 per cent of user
A case for automating reference
customers and remain patient and queries.
services with chatbots
polite throughout their interaction 2. Catalog or information searches –
Chatbot applications are as diverse with such users. representing 40 per cent of user
as the organizations that deploy them. ● Because of the anonymity of chatbot queries.
The ultimate goal of chatbots is to conversations, users are put at ease 3. Chat – representing 20 per cent of
replace the current interfaces that exist about asking “stupid” or user queries.
on computers and in mobile devices. embarrassing questions.
● Chatbots can be used as a marketing Some factoids regarding Emma
Libraries must be aware of “new”
tool for reference services, (Iglesias, 2013):
technologies and do their best to make
use of them to improve library services. familiarizing users about the variety
of available resources and services. ● In 2011, users had a total of 7,116
Repetitive, mundane tasks or activities
● Chatbots can reside on multiple conversations with Emma:
done by people are a form of wasteful
production and are ripe for automation. types of technology platforms – 2,223 questions pertaining to the
McNeal and Newyear (2013) identifies simultaneously (website, desktop library, and its services and
the following benefits of using chatbots PC, kiosk, etc.). policies.
as virtual reference librarians: ● Customers expect support to be – 2,551 user queries were sent to
available anywhere and anytime – the library’s catalog, databases or
● Chatbots augment traditional chatbots can provide 24/7 access to other online resources.
library staff; chatbot technology library services at the user’s
can help make up for budgetary or convenience, even when the library Emma answered a total of 4,774
staffing shortfalls. is closed. library-related questions; the cost for
● Delegating routine reference questions ● Chatbots can generate traffic to the providing this service was $0.14 per
for basic library information to library website and increase the use.
6 LIBRARY HI TECH NEWS Number 4 2017
Figure 1. Emma the Mentor Public Libraries Catbot AIML-based virtual agent designed
for libraries. Libraries are welcome to
participate in the development and
improvement of this chatbot by
editing a base repository version of
the code hosted on Google Code to
conform to their library’s specific
needs. There are two paths to the
InfoTabby Project participation:
“Non-Sharing Path” and “Sharing
Path”. Iglesias (2013) lists these two
advantages of the “Sharing Path”:
● The developer receives alerts of
and has access to the latest changes
to the base AIML code made by
other members of the library
community.
● The improvements made by the
developer are easily shared with all
community members.
Perhaps you are not into cats – or if you
want to create your own unique virtual
reference librarian from scratch,
● By November 2011, Emma’s enabling people to input knowledge Awesome Bots (see [Link]
correct-response rate for questions into chatbots. AIML is similar to BotCube/awesome-bots) has curated a
about the library and its services HTML; if you understand HTML, you list of bot-related resources (both text
averaged 90 per cent. can quickly learn AIML. Allison (2012) and voice bots) to assist you in that
● Emma regularly received marriage states two reasons for using AIML: endeavor.
proposals. 1. The metadata is flexible to
● Emma has received inquiries from Conclusion
accommodate the variety of
people outside the USA, including questions coming from library
from Russia and South America. Reference services via text
users.
2. The availability of AIML files messaging and chat exist today in
Emma resoundingly succeeded as a libraries; they are similar to a chatbot
which have been contributed by the
proof-of-concept that chatbots could experience, in that technology negates
open source community.
effectively handle the majority of the need for face-to-face interaction.
routine reference questions. Patrons’ The AIML metadata scheme was We can leverage what we know from
reactions to Emma have been invented by Dr Richard Wallace and these human chat sessions in the design
overwhelmingly positive, particularly a has been widely adopted by a large of a chatbot. Most libraries use some
hit among teens and young adults. In open source community of users sort of reference tracking software such
recognition of what was accomplished under the stewardship of the as Gimlet; this type of data can be used
with Emma, David Newyear the creator A.L.I.C.E. A.I. Foundation, as the basis of the chatbot’s knowledge
of “Emma the MPL Catbot” won the “promoting the development and base. Chatbots cannot replicate the
2011 Polaris Innovation in Technology adoption of A.L.I.C.E. and AIML free complexity of human interaction (both
John Iliff Award – the award provides a software”. A.L.I.C.E. (Artificial knowledge and emotional), but can
$1,000 honorarium to a library Linguistic Internet Computer Entity) provide a cost-effective way to answer
professional or library that has used or Alicebot has won the Loebner the majority of routine reference
technology and innovation as a tool to Prize three times – in 2000, 2001 and questions and direct users to the
improve services to public library users. 2004 – awarded to chatbots appropriate service. Libraries will need
considered by the judges to be the to find a balance between their human
Bot-building platforms most human-like. AIML and and virtual reference librarians by
ChatScript have performed well in redesigning work practices to leverage
Emma was created using AIML Loebner Prize competitions, the strengths that each possess. The
(Artificial Intelligence Markup providing the foundation for many ultimate goal of virtual reference
Language) metadata using the successful chatbots since A.L.I.C.E. librarians is not to replace humans or
Pandorabots hosting service, a free “Emma the Catbot” was released to face-to-face human interaction but to
open source-based community. AIML the open source community as streamline the work of the reference
is an XML-compliant language InfoTabby around 2011 as an department. Chatbot software is not
LIBRARY HI TECH NEWS Number 4 2017 7
something you install and forget about. ● Library users. Mentor Public Library and Pandorabots
Chatbots evolve over time, learning ● Library facilities. (2017), “Pandorabots”, available at: https://
with very interaction. Library staff will ● Intelligent agents. [Link]/static/html/Stories/Librar
need to examine the chatbot ● User interfaces. [Link] (accessed 23 March 2017).
conversation logs, and use the continual ● Information retrieval. Rubin, V.L., Chen, Y. and Thorimbert, L.M.
stream of queries to improve the quality (2010), “Artificially intelligent
and complexity of the chatbot’s conversational agents in libraries”, Library
responses. Adding to the chatbot’s REFERENCES Hi Tech, Vol. 28 No. 4, pp. 496-522.
knowledge base will require a certain Allison, D. (2012), “Chatbots in the library:
amount of time and effort to which is it time?”, Library Hi Tech, Vol. 30 No. 1, Schlicht, M. (2016), “The complete
librarians may or may not be able to pp. 95-107. beginner’s guide to Chatbots”, Chatbots
commit. Magazine, 20 April, available at: https://
Dunn, J. (2016), “Here’s more reason to [Link]/the-complete-begin
believe the app boom is over”, Business ner-s-guide-to-chatbots-8280b7b906ca#.5ft
Subject terms Insider, 14 October, available at: [Link] csqjry (accessed 20 March 2017).
● Library software. [Link]/app-boom-over-chart-ios-
● Information services. android-2016-10 (accessed 20 March 2017).
Joseph Vincze ([Link]@
● Library users. Iglesias, E. (2013), Robots in Academic [Link]) is Information Specialist
● Library science software. Libraries: Advancements in Library at Rutgers – The State University of
● Computer software. Automation, IGI Global, Hershey, PA, New Jersey, New Brunswick, New
● Chatbot. pp. 101-112, doi:10.4018/978-1-4666-3938-6.
Jersey, USA.
● Artificial intelligence. McNeal, M.L. and Newyear, D. (2013),
● Reference. “Introducing chatbots in libraries”, Library
● Libraries. Technology Reports, Vol. 49 No. 8, p. 5.
8 LIBRARY HI TECH NEWS Number 4 2017
