We'll Take it from Here: Developments We'd Like to See in Virtual Reference Software

Steve Coffman

Virtual reference services - providing patrons with live, real-time reference over the Web - have suddenly become a very popular topic in the library community. Any conference program with "virtual reference" or "digital reference" or "24/7" or any of the variety of other euphemisms we use for live, online reference on the Web, is guaranteed to be packed. New electronic discussion lists and discussion groups are popping up like mushrooms. The first articles on the subject have already appeared in American Libraries and Library Journal, and dozens more are being churned out as we speak. But even more telling are the numbers of libraries that have actually begun to implement it. In September 1999, there were no more than five libraries that had implemented any kind of live virtual reference service, or that even knew what it was. Today, less than eighteen months later, there are over two hundred libraries from all over the world that have started offering live online reference in one guise or another, and more are joining the fray everyday.

Of the more than two hundred libraries worldwide that have started offering live online reference services, only a few have tried to develop their own versions of virtual reference software - often based on an existing chat application of some sort. Most have chosen to go with one of the many commercially available applications like eGain Live, Cisco's Webline, LivePerson, Humanclick, or a number of others - all of which were originally designed to allow e-commerce companies to provide live interactive customer service over the Web. And all of us have been struggling - with varying degrees of success - to take these commercial customer service applications and modify them for library reference purposes. As the Product Development Manager for LSSI's Virtual Reference software, I am a veteran of those struggles, and have the scars to prove it. Over the past two years now, I have worked to adapt a number of commercial e-commerce applications for reference purposes, and it has not always been an easy process.

On the surface, answering a customer's questions about a pair of denim jeans and answering a patron's question about the etymology of the word denim (or anything else) would seem to have much in common. In reality, however, there are some pretty fundamental differences between online customer service and online reference and the software that is necessary to support it. In the first place, customer inquires about denim jeans or any other product can almost always be answered with information from the company Web site alone. So as long as the software works well with the company Web site, there is no problem, at least for the e-commerce people.
But that is hardly the case in reference. To answer the question about the etymology of denim (it comes from de Nîmes or "from Nîmes," France, by the way, for those who were dying to know), a librarian might want to check at least one or two good etymological dictionaries and perhaps a few other resources, none of which are actually part of the library's own Web site. And the next reference question is likely to require browsing a whole different set of resources. So the software that needs only work well with the company's site when used in e-commerce applications must be compatible with thousands, perhaps even hundreds of thousands, of other resources when it is used for reference. And there are plenty of other difficulties too. These software packages lack a good "on-hold" function because it assumes that a customer service rep will be able to quickly answer a question and move on to the next one without significant delay. True enough, when you are answering a question about a pair of jeans but not when you are a librarian browsing all over the Web looking for the answer to a reference question while the patron waits. And while the built-in knowledge bases that come with many of these packages may work quite effectively for answering product questions where there are a limited number of facts about a limited number of items, they have not performed well in library reference work where there can be almost infinite numbers of questions and answers.

The list of shortcomings could go on and on. It is important to note, however, that despite their limitations, commercial Web-based customer service applications are still the best resource available to libraries that want to venture into virtual reference on the Web. By adapting these existing applications instead of going it alone, we have been able to leverage millions of dollars in investment flowing into the e-commerce arena and take advantage of it for our own purposes. As a result, libraries now have access to applications that offer functionality and refinement far beyond anything we could have developed by ourselves. And the best of the Web customer service applications on the market offer a whole suite of interactive technologies that have proved very useful for live, online reference. Among the best applications featured are the following:

  • full two-way co-browsing (you can see what the patron sees, and visa versa);
  • form-sharing (the ability to share search strategies and other text in Web forms);
  • the ability to share files and slideshows with a patron;
  • scripted messages and bookmarks to help handle routine functions and requests;
  • basic knowledge bases;
  • logging and capture of reference sessions for future analysis;
  • full session transcripts e-mailed to both the patron and librarian;
  • queuing and routing of incoming questions; and
  • the ability to conference or transfer a call with other libraries on the system.

So, while commercial Web customer service software is hardly perfect, it does provide a good foundation upon which we can build the new virtual reference systems we will use tomorrow. The question now becomes, where do we take it from here? What changes and refinements will we need to make to this software, and what new functions and systems will we need to develop to build virtual reference applications that will more effectively meet the needs of libraries on the Web.

There are no definitive answers here of course, and virtual reference systems will always be a work in progress, just like any other software. However, based on a few years experience with the current software, I think we've got a pretty good idea of where some of our biggest problems are and the issues we should be tackling first. Here's what's on our short list at LSSI - and from my conversations with others working in this field, these are problems and frustrations shared by us all.

Co-Browsing and Collaboration

In general the co-browsing and collaborative capabilities of Web-based customer service software need to be significantly improved for virtual reference systems. The fundamental problem is that to be truly effective, virtual reference systems must be able to escort the patron anywhere on the Web - most particularly through those proprietary databases we subscribe to. We are spending a lot of money on those databases and it would be nice if we could use the virtual reference software to show our patrons how to use them more effectively. Unfortunately, none of the existing software works with more than a few of our databases right out of the box. Proxy-server-based co-browsing like that used in LSSI Interact software and Cisco's Dynamic Content Adapter - where a single computer does the browsing and sends the content back to all the parties on a session - seems to be the best solution to this problem. But there are still many databases that cannot be co-browsed effectively even with proxy-server-based systems. Each problem database needs to tackled and resolved on a case-by-case basis - and those of us who are developing software in the field are gradually working through the problem databases one by one. But it is an arduous process, and while we are making progress, there is still a long way to go. We would also like to see a tighter integration between the database and virtual reference in other areas as well. For example, it would be nice if the database producers could include a virtual reference link on all pages of their database - something that might say "Didn't find what you were looking for? Click here for help," a sort of "point of need" button the patron could click on to access a librarian anytime they needed help.

In addition to being able to co-browse a broader range of our databases we would also like to see some enhancements to the collaborative tools available in a reference session. For example, it would be nice to have a "virtual marker" that could be used to highlight words, phrases, and passages of a Web page we sent to the patron. It would also be nice if we could scroll the patron's screen. Most applications only allow you to send the Web page, not scroll it up and down for the patron.


The communications technology is another area that could use some improvement for virtual reference purposes. Right now, most live customer service software uses chat as the primary method of communication between the agent and the customer. While it is better than nothing, chat leaves something to be desired as a communications tool for almost any function, but particularly for reference, where the ability to conduct an effective reference interview requires careful attention to nuances of voice and inflection that are lost in chat. Moreover chat is much more time-consuming than regular voice communication because you have to type everything out - and then you have spelling and typing errors that can creep in and make even the best of us look like fools.

For all of these reasons and more, it is hoped that chat is an interim technology which will soon give way to something much more humane like voice. Indeed, this transition seems to be already well underway. Many people already have two telephone lines or high-speed Internet connections using cable or DSL and in these cases, of course, it is already possible to speak with the patron over the phone while using the virtual reference software to handle the co-browsing on the Web. But for those people who do not yet have access to a second phone line, the answer seems to be Voice over Internet Protocol (VoIP). VoIP is a developing Internet protocol that allows both voice and data to be passed over the same Internet connection. In a VoIP session, a librarian and a patron would co-browse the Web and talk with each other either using headsets plugged directly into their computers, or by using the built-in microphone and speakers that now come standard on many PCs. While studies show that many PCs installed in people's homes and offices are VoIP capable, for some reason very few people take advantage of that feature as of yet. This is likely to change over the next few years as the network gets faster, voice quality improves, and people begin to explore the technology as a way of making cheap long distance phone calls.

One potential downside to relying on voice instead of chat in virtual reference is the risk of losing that nice chat transcript that gives such insight into the reference process. However, it is possible to capture the voice as a sound file and to store the Web pages pushed during the session along with it. The problem is that analyzing sound files would be a good deal more cumbersome and tedious than reviewing chat transcripts - and that is tedious enough. Perhaps the best solution would be to explore ways of using voice-to-text software to convert voice records to searchable text files that could be analyzed much as we are using chat transcripts now.

Networked Reference Services

One of the greatest potentials of the new virtual reference software is that it could serve as a very effective platform for the development of shared and networked reference services. Up until now, reference has been a pretty insular activity with each library basically handling its own questions as best it could. That method works just fine for routine reference questions that can be easily answered within the expertise and resources of the local library, but it breaks down as soon as you get a "zinger" that can't be answered effectively with your own resources. In those cases, it would be nice if the question could be simply and easily transferred to somebody with the resources and expertise to answer it. The problem is that we have never had the technology that allows us to quickly and easily transfer and share questions among one another. Virtual reference software has the potential to change all of that. The software allows the transfer of calls or conferences with librarians from any library using the system and it can be done live, in real-time while the patron is still online. So we now have the basic network structure that makes it possible to share reference, but there is still a lot of work that needs to be done before libraries can really take advantage of this new capability.

One thing needed is a good question tracking and routing system that will help us keep track of who is handing a question for whom, what its status is, what time and costs were required to answer it, and whose account it should be charged to. The Collaborative Digital Reference Service at the Library of Congress has done some initial work in this area, but much remains to be done. We also need to look at some sort of standardized record format for the questions themselves so they can easily be passed back and forth between libraries using different virtual reference systems, just as the MARC record allows us to interchange bibliographic data between libraries using different automated systems.

Finally, once we have this infrastructure in place, libraries can be expected to begin developing specialized reference services that they will offer to others on a subscription basis. For example, it is easy to imagine a library with a strong business reference service using the virtual reference software to make that service available to other libraries on a subscription basis. Librarians at subscribing libraries could either transfer calls to the business reference service or patrons could be routed directly to the business reference service from a live link on the subscribing library's Web page. Business is just one of the many specialized subject areas where subscription reference services would make sense. Other obvious fields include law, medicine, sci/tech, statistics, foreign language reference services, and a number of other specialties that we "general practitioner" librarians often have difficulty with.

Online Reference Collections and Knowledge Bases

One of the things that distinguishes libraries from other sources of information on the Web is that we provide access to current, authoritative, and unbiased data in a wide variety of subject areas. The problem is that a great deal of that information is still locked up in our print reference collections and not available over the Web. So, if we are serious about moving reference to the Web, it is high time we began to work with the major reference publishers to begin to move some of those key resources to the Web. And it is not just a matter of making an electronic replica of the print source; the Web allows us to use these resources in ways we never could in print. At a very minimum, we should be looking at developing a reference search engine that allows us to keyword-search the full text of our entire electronic reference collection regardless of which publisher created the source. Plus, our electronic reference sources should be able to learn and improve as we use them to answer questions. For example, librarians should be able to bookmark or annotate the sources to help others find the answers to difficult questions more easily, and we should also be able to refer others from a source to supplemental material we may have found in answering a reference question. So when somebody listed in one of our biographical dictionaries dies, for example, we should be able to annotate the entry with their obituary plus references to the spate of articles that typically appear after their demise. If we could incorporate some of these ideas, our reference collections would no longer be just a static collections of sources, they would be living and breathing things which would develop and improve as we worked with them. The ideas here are not entirely new; you can begin to see the result of the first efforts to develop some of these new reference tools in products like Xrefer ( www.xrefer.com) where a variety of core reference sources are searchable in a single search engine. This is a step in the right direction, but much remains to be done.

We also need to take another look at licensing, especially for reference purposes. Most standard license arrangements for electronic resources are based either on FTE or on simultaneous users, where the assumption is that patrons will be accessing these resources online all on their own. This works all right for general purposes, but it falls down when the resources are used for virtual reference. In the first place, how do you count collaborative browsing? Does each party on a session count as one simultaneous user? Or does each virtual session count as one? But, more importantly, current licensing arrangements tend to overstate the value of reference sources that are not used heavily. We believe it is time we got together with the publishers and see if we could come up with a licensing arrangement specifically for virtual reference where the librarian and the patron were sharing the resource in a virtual reference session. This could help us address the issue of what counts as a simultaneous user and also help insure that some of our less commonly used reference material was available at a more moderate price. This sort of reference licensing could also give librarians an opportunity to try out a resource for reference purposes, and if it proved popular, they could then license it for direct patron use at a higher cost. This is not just idle speculation. LSSI is currently working with a number of reference publishers to develop exactly this sort of specialized licensing for virtual reference - and hopefully these efforts will bear fruit.

Finally, it is about time we realized that reference publishers are not the only sources of content for our reference collections. In fact, librarians create what is potentially some of the best and most useful content every time we work out the answer to a reference question. We all know that questions recur, and if we could somehow access the work another librarian had done before, there would be no need to start over answering every question from scratch. The problem is that up until now there has been no easy way to preserve our work, so our efforts vanished with the patrons as they strolled out the door. Now, for the first time, the virtual reference software has given us the tools to capture and preserve the reference content each of us is creating on a daily basis. The software records and preserves a complete transcript of every reference session including all of the chat conversation between the librarian and the patron as well as the titles and addresses of all Web pages pushed, and files, slides, and other content that was transferred during the session. And it does it all automatically. There is nothing the librarian has to do - not even make a hash mark.

Of course, now that we've finally got this information, the question is, what are we going to do with it? People have only just now begun to experiment with this transaction data, but several possible ways of using it seem to be shaping up.

The simplest approach is to use elements from the transcripts of previous questions as a sort of crude FAQ so you can run new questions through to see if they have already been answered. You can see one such system in operation at WebHelp ( www.webhelp.com), a major commercial reference service. When you type in your question at WebHelp it automatically runs a keyword search against all of the other question transcripts in its archives, and brings back a list of possible matches from previous questions asked on the system. If you don't find your answer within those results, you can then click on the button to get live help. The system has its limitations - people can ask questions in many different ways so there are quite a few duplicates, and the list of Web pages pushed appears without any context since the chat transcript is not displayed to preserve customer privacy. But there are some great advantages as well: the entire FAQ can be produced automatically so you don't have to wait for answers to be written, edited, and published as in a more formal knowledge base, and new questions are available for searching immediately. So, if you suddenly get a run on a very current question because of something that has appeared in the news (for example, What is mad cow disease?) there is a strong likelihood that patrons may be able to find the answer to the question in your automatic FAQ before they ever reach you.

The next step up would be to take the raw content in those saved reference transcripts and use it to create a more formalized knowledge base. As mentioned earlier, many of the commercial customer services applications come with a basic knowledge base of some sort already built-in. While those knowledge bases may be adequate for commercial purposes, they are woefully inadequate to meet our needs in reference where we must figure out ways to store, retrieve, and maintain what could easily be millions of records. There are a number of parties that have begun to examine this issue, but nobody has yet been able to demonstrate a system that will effectively answer a very broad range of reference questions without frustrating both the librarian and the patron. One of the largest attempts to come up with a similar system on the commercial side was the AskJeeves search engine, and anybody who has used that knows the limitations of very large knowledge bases.

Finally, we can analyze the raw transcripts and see what we can learn from them about how people are using our reference services and the kinds of questions they are asking. Ultimately the objective is to anticipate our patrons' questions and to design our library Web sites and electronic collections so that most patrons can easily find the information they need without having to ask. This one is a no-brainer, and many libraries have already begun reviewing their transcripts to see what lessons they can glean from them. At this point, this is pretty much a manual process, and that will work fine as long as there are a modest number of transcripts. As reference volume increases, however, it will be difficult to continue this sort of analysis by hand, and we will need to take a look at some of the software applications used to process and analyze large data warehouses and information stores in the commercial world.

Those are some of the developments you can expect to see in virtual reference software over the next one-to-two years as we work to take something that was originally built for selling sweaters, stocks, and vacation packages, and transform it into something we can use to help people find the information they need. This is hardly an exhaustive list (and doubtless many of you could suggest things that should be added) yet it is a very tall order, even as it stands. It is a very good thing, then, that so many wonderful, enthusiastic, and energetic librarians have been jumping into this field in recent months. As you can see, we have a whole lot of work to do, and if we are to be successful, it will require every bit of commitment, enthusiasm, and energy we can muster - and then some. But if we are successful, we will have done much more than just create a new piece of software; we will have made a set of tools that promises to fundamentally change the face of reference as we have known it. Now, let's get going.

   Steve Coffman ( coffmanfyi@earthlink.net) is Product Development Manager, Virtual Reference Services, at Library Systems and Services, LLC (LSSI).