Creative Writing

27 pages
4 views

Conversational Agents in Libraries

of 27
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Share
Description
Conversational Agents in Libraries
Transcript
  THEME ARTICLES Artificially intelligentconversational agents in libraries Victoria L. Rubin and Yimin Chen  Faculty of Information and Media Studies, University of Western Ontario, London, Canada, and  Lynne Marie Thorimbert  Marigold Library System, Strathmore, Canada Abstract Purpose  – Conversational agents are natural language interaction interfaces designed to simulateconversation with a real person. This paper seeks to investigate current development and applicationsof these systems worldwide, while focusing on their availability in Canadian libraries. It aims to arguethat it is both timely andconceivable for Canadian libraries toconsider adopting conversational agentsto enhance – not replace – face-to-face human interaction. Potential users include library web site tourguides, automated virtual reference and readers’ advisory librarians, and virtual story-tellers. Toprovide background and justification for this argument, the paper seeks to review agents from classicimplementations to state-of-the-art prototypes: how they interact with users, produce language, andcontrol conversational behaviors. Design/methodology/approach  – The web sites of the 20 largest Canadian libraries weresurveyed to assess the extent to which specific language-related technologies are offered in Canada,including conversational agents. An exemplified taxonomy of four pragmatic purposes thatconversational agents currently serve outside libraries – educational, informational, assistive, andsocially interactive – is proposed and translated into library settings. Findings  – As of early 2010, artificially intelligent conversational systems have been found to bevirtually non-existent in Canadian libraries, while other innovative technologies proliferate (e.g. socialmedia tools). These findings motivate the need for a broader awareness and discussion within the LIScommunity of these systems’ applicability and potential for library purposes. Originality/value  – This paper is intended for reflective information professionals who seek agreater understanding of the issues related to adopting conversational agents in libraries, as this topicis scarcely covered in the LIS literature. The pros and cons are discussed, and insights offered intoperceptions of intelligence (artificial or not) as well as the fundamentally social nature of human-computer interaction. Keywords  Academic libraries, Public libraries, Canada, Intelligent agents, User interfaces,Information retrieval Paper type  Research paper Introduction Conversational agents: srcins and types As modern libraries continue to evolve in the information age, the problem of how tobest access information and take advantage of technological advances withoutintroducing new barriers remains a compelling puzzle. The impact of internet, andcomputer technologies on the library paradigm is unmistakable: even small ruralCanadian libraries may offer web sites and an online public access catalogue (OPAC). The current issue and full text archive of this journal is available at www.emeraldinsight.com/0737-8831.htm LHT28,4 496 Received 12 April 2010Accepted 23 May 2010 Library Hi TechVol. 28 No. 4, 2010pp. 496-522 q Emerald Group Publishing Limited0737-8831DOI 10.1108/07378831011096196  Where to next? We make the case that Natural Language Interaction (NLI) systems,human-computer interfaces designed to simulate conversation with a real person, arean effective, appropriate complement to many existing library services and may be thekey to unlocking solutions to future interactions with information.NLI is a part of the broader fields of Natural Language Processing (NLP) andArtificial Intelligence (AI). Most internet users have, perhaps unknowingly, alreadyhad some contact with NLP technologies in the form of search engines and machinetranslation tools and with AI – through video games and financial tools. NLP studiesthe structure, function, and use of language, and organizes it into computationalmodels to design and develop language-related software applications (Joshi, 1999). Thefield of AI focuses on intelligent computer systems, those that exhibit thecharacteristics we typically associate with intelligence in human behavior such asunderstanding language, learning, and problem solving (Barr and Feigenbaum, 1982).As a synthesis of these two fields, the promise of NLI is the ability to providebelievable, personalized, and human-like interaction with computers in naturallanguages like English or French.For our purposes, we identify two realms within NLI: “chatbots” (also known astext-based conversational agents, artificial conversation entities, chatterboxes, orsimply bots), where interaction is limited purely to text input and output; and“embodied conversational agents” (ECA) 1, where the computer interface “isrepresented as a human body, that uses its face and body in a human-like way inconversation with the user” (Foster, 2007, p. 828) and incorporates animations that aresynchronized with the system’s linguistic behaviors. Historically, text-based chatbotssimply aimed at free-flowing conversations with the user, while ECAs generally servea particular purpose, such as providing information to users or aiding them with atask. Due to animation, ECAs can also express believable human behavior duringconversations to enhance the social quality and enjoyment of human-computerinteraction. Survey of language technology availability in Canadian libraries To position our research in the context of the broader use of language-relatedtechnologies in libraries, we surveyed the current state of their availability in thelargest libraries in Canada. Using the latest available Canadian Public LibraryStatistical Report and Rankings as the basis (Public Library Statistics, 2008), weselected the ten largest public libraries by holdings (Public Libraries, see the Appendix,Table AI). By the same parameter from the Maclean’s, 2009 Rankings of CanadianUniversities (Maclean’s, 2009, 16 November), the top ten academic libraries werechosen (Academic Libraries, see the Appendix). The rationale was to obtain a sampleof Canadian libraries that are likely to have the resources and expertise to test andimplement current language technologies.The web site of each selected institution was examined for availability of language-related services and information about them (accessible electronically viatheir respective web sites). Our primary interest was artificially intelligent NLPapplications such as text-based question-answering, and more specifically chatbotsand conversational agents on the library web sites. We also inventoried broaderinternet-enabled services where text or speech were involved (see the Appendix). Artificiallyintelligent agentsin libraries 497  In particular, we looked for text- or audio-based services (such as OPACs anddownloadable e-books), language-related resources (online language learningpackages) and web site accessibility tools (such as audio/video tutorials, and foreignlanguage translations), online socially interactive services (such as live or instantmessage reference services, and social software applications), as well as other NLP/AIapplications (such as retrievable FAQs). In addition, we examined web sitedocumentation (missions, visions, core values, or strategic plans, where publiclyavailable) for statements pertaining to new technologies to see how the librariesaddress and present issues of technological innovations. Survey results: current status of broader innovative language-related technologies As expected, all 20 surveyed library web sites, provide access to OPACs anddownloadable e-books. Eight out of 23 public libraries (Toronto, Ottawa, Mississauga)and five academic libraries (UBC, Queen’s, University of Calgary, McGill, and Ottawa) – provide video tutorials to navigating their web sites. This is indicative of a new trendin help services: moving from searchable text manual style to more interactive andengaging brief presentations.All 20 libraries have a-synchronous e-mail reference services. Six public librariesand all but one academic library (Montre`al) have live or instant messaging reference(IM) services. Hours of operation of these services are typically restricted to officehours, for instance, Vancouver’s AskAway does not operate after 5 p.m., and isunavailable on Sundays. When unable to cope with the volume of simultaneousquestions, some IM reference services place patrons in live queues until a librarian isable to join the session: an estimated wait time for a question to the Virtual ReferenceDesk at the Toronto Public Library was 26 minutes (on 11 March 2010).All but two (the Winnipeg and Fraser Valley Regional Public Libraries) have atleast one form of social media application: Facebook, Twitter, blogs, RSS, Tagging, orYouTube (Social Media, see the Appendix). The proliferation of these applications maybe a fad, but represents a consistent institutional response to user interests in newinteractive technologies.More than a half of the library web sites provide alternative language support forweb site navigation through translations or alternative resources. English and Frenchbilingual navigation is common (in seven public libraries and in four academiclibraries – Alberta, McGill, Montre`al, Ottawa). In addition to English and French website access, extensive multilingual access support is provided in five public librarycases: Vancouver offers library web site navigation in six alternative languages,Edmonton – in nine, Mississauga – in ten, Ottawa – in 11, and Toronto – in 16 (website Translation, see the Appendix). As far as online computer-assisted languagelearning applications, only four institutions, all public libraries, offer online access tosuch software (specifically, the Calgary and Edmonton Public Libraries provide accessto Tell Me More Online courses; Ottawa and Hamilton – to Mango Languages).Four libraries show their dedication to technological innovation in their publiclyavailable online documentation. University of Saskatchewan Library acknowledgesthat “library users expect state-of-the-art computing equipment, robust networks, vastcollections of software, and other information resources as well as expert assistancewhen and where they need it” (University of Saskatchewan, 2010). Toronto PL states LHT28,4 498  that “[n]ew technologies extend access to global information beyond library walls”(Toronto Public Library, 2010). Ottawa PL “continuously review[s] current practices,make[s] improvements, leverage[s] technology . . . ” (Ottawa Public Library, 2010). UBCLibrary announces “the digital agenda is a major plank of their library” includingon-line access to “thousands of full-text e-journals, e-books, indexes and databases”,and “providing new and more efficient types of digital services . . . ” (University of British Columbia, 2010). The question is: to what extent are these libraries prepared toexperiment with groundbreaking state-of-the-art technologies, based on theirindividual priorities, criteria for adoption, and implementation capabilities? Survey results: current status of conversational agents Our survey confirms that NLI systems specifically are virtually non-existent in the top20 largest Canadian libraries as of early 2010. None of the top 20 surveyed librariesemployed embodied conversational agents for any of their online-accessible services.As for text-based NLI applications, there was only one rare approximation. Instead of astandard static FAQ list, the University of Western Ontario uses a text-basedartificially intelligent Ask Western Libraries service, an Instant Answer Agentdeveloped by the IntelliResponse aspiring to embrace “the new paradigm of the web,social media and mobile” (IntelliResponse, 2009). Ask Western Libraries retrievesbest-fit answers from the inventory of FAQs such as “What are western librarieshours?” (see Figure 1). Figure 1. Ask Western Libraries:a searchable text-basedinteractive FAQ list, theUniversity of WesternOntario Artificiallyintelligent agentsin libraries 499  The service takes the form of familiar information retrieval format, but its domain islimited to the pre-existing information on library and archives operations. The system“has been asked 58,834 questions since it was implemented in 2007” as a form of on-line support and self-help mechanism, and has been able to pair up “80 percent of asked questions to one best-fit response” (From personal communications with JenniferRobinson, communications and outreach librarian, March 25, 2010). Although far froma full-fledged NLI system, this application is still a step up from unattractive static listsand endless scrolling.In sum, the lack of conversational agents in Canada is not surprising consideringthat, to the best of our knowledge, there have been very few attempts of NLI systemapplications worldwide. Most notably, four systems were found in Germany and one atthe Mount Saint Vincent University Library in Halifax, Nova Scotia, which was notamong the surveyed Canadian libraries. These systems will be further illustrated anddiscussed in this paper. The current status of the availability of broaderlanguage-related technologies – and specifically NLI systems – motivates the needfor a broader discussion within the LIS community in order to heighten awareness of these systems’ applicability towards library purposes. Evolution of capabilities and internal workings of select prototypes We argue it is both timely and conceivable for Canadian libraries to considerincorporating conversational agents into their systems. To provide background and justification for this argument, we first review several NLI systems, from classicimplementations to state-of-the-art prototypes: how they interact with users, producenatural language, and control conversational behaviors. Four systems are discussed asimportant landmarks in the development of NLI systems towards more human-likeintelligent interactions – a task far more difficult than it may seem at first glance. Formost people, using and manipulating language is second nature, but the processesinvolved in even the simplest acts of communication are decidedly non-trivial: wordsrarely have only one meaning, grammatical structures can often be interpreted in avariety of ways, and the pragmatic intent of a sentence can be completely at odds withits semantic meaning (e.g. sarcasm). A vast amount of human knowledge about thephysical and social world is required to achieve any practical level of languageunderstanding and manipulation. The Classic ELIZA In 1966, Joseph Weizenbaum designed one of the first chatbots, ELIZA, whichsimulates a dialogue with a Rogerian psychotherapist. ELIZA could convincinglycreate a discursive environment where almost no real-world knowledge was requiredon the part of the program. Since the role of a therapist was understood to benon-judgmental and passive, the chatbot was designed to allow the patient to direct theconversation. As a result, ELIZA often rephrases patient statements in the form of aquestion (Weizenbaum, 1966). By taking advantage of common conversationalpatterns (e.g. “I hate my father”/“Why do you say that you hate your father?”), ELIZAis capable of producing authentic-sounding responses. Depending on how the userchooses to interpret and respond to ELIZA, the illusion that he or she is actually LHT28,4 500
Related Documents
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks