 Good morning and welcome to this week's edition of Encompass Live. I am your host, Krista Burns, here at the Nebraska Library Commission. Encompass Live is the commission's weekly online event. Yes, it's a webinar that we do covers anything that might be of interest to librarians across the state and actually across the country. We do these sessions live every Wednesday morning at 10 a.m. central time, but they are all recorded. So if you're unable to join us on Wednesday mornings, you can go onto our website and see all the recordings and watch all the recordings of all the previous sessions we have done. We do mixtures, things here. We do presentations, interviews, mini-training sessions. Anything that's related vaguely to libraries, we'll put it on the show. We have commission staff that do sessions for us, episodes, and we have guest speakers that come in, which I guess David's going to have a mixture. The last Wednesday of the month, usually, almost every time, is our tech talk with Michael Sowers, next to me here, who is the Technology Innovation Librarian here at the Nebraska Library Commission. He comes on and shares some tech news of the month since the last time he did and sometimes brings in speakers as he has today over here to his left. I will just hand over to you and you can explain what you've got here and what you've brought in. Great. Thanks, Christa. Setting up tech talk each month is always kind of an experience. Usually, sometimes I'm finding people at the last minute. Sometimes I've got folks booked months in advance and I just want to share a quick little story about how today came about. I was at Internet Librarian last month, yeah, last month, October, right? Yes. I had just finished having lunch one day with Amy Mather from Omaha Public Library and I had forgotten my camera at the restaurant, so I had to run back to the restaurant and get my camera and then I realized I had left the power cord to my laptop and the room I had done a presentation in before, so I ran over to that room before that session started and ended up running into Rivka SAS, formerly of the Omaha Public Library. And then by this time, sessions had started and I wasn't sure where I was going to go and I just started flipping through the catalog and I went, hmm, artificial intelligence in the library, that sounds interesting. So I went over there and I kind of sit down towards the front and plugged my computer in and I went, wait a minute, this looks familiar. And it was Deanne and Laura from UNL here talking about their pixel project and I thought, okay, I'm in California and I am just running into everybody from Nebraska who was at this conference in less than an hour. So I sat through their presentation and then walked up to them and said, hey, how'd you like to be on the show? And I think you kind of looked at me like we've been waiting for you to call. So we have here today Deanne Ellison and Laura Dawes from UNL and I'm just going to kind of hand it over to them to give their presentation about this very interesting project of theirs and then we'll submit questions as we're going along and we'll get those in and I'm sure I will have some questions for them at the end. So why don't you go ahead and take it away? Okay, thank you. We're going to be going back and forth during our presentation so you get a chance to hear both of our perspectives. I'm Deanne Ellison and I'm the director for competing operations and so I'm the technical person behind Pixel, Pixel's mother if you would. And Lorna is here to provide more of a reference public service perspective. She joined the project about about a year into it and I was really happy to get a little different viewpoint than what I have. So I'm going to start by giving you a little bit of information. Oops. Yep. Click on it. There we go. Sorry about that. I'm going to start with a little bit of an overview of what the University of Nebraska-Lincoln is like. We're a land-grant institution founded in 1869, so a fairly old institution as far as that goes. We are recognized by the Carnegie Foundation as a major research activity institution. So research is an important part of what we do. We have three major missions, research service and education of course. There are about 18,000 students at UNL and about 2,500 faculty staff, so we're not a really large institution, but it's large enough so that we have interesting times trying to serve everyone in the libraries. For the libraries, we have seven libraries on campus right now. The largest is Love Library with six branch libraries. We work very closely with the Law College and their library there to share a catalog and many of the same resources. We also have the Archives for the Whole University System. That's the University of Nebraska-Lincoln, Omaha, Kearney, and the Med Center all located in Love. We have special collections as well, and that concentrates on rare and limited editions of books from the Great Plains in Nebraska in particular. We have a heavy collection of Willa Catherine materials, for example. As a research institution, research is also important to the libraries as part of our mission as well, and so Pixel fit pretty nicely into that particular function. We are hoping that we would be able to reduce staffing and potentially 24-hour chat service by substituting people for Pixel. We do not expect to replace reference librarians by any matter of means. Reference is important and important service for the libraries, but there's an awful lot of questions that come to the reference staff that could be answered by Pixel. Directional questions, really simple things. We don't expect Pixel to ever be mentoring somebody through a thesis or dissertation. We expect librarians to be providing high touch in those areas where that's important. But for the undergraduate at 2 o'clock in the morning who realizes their paper's doing three hours, Pixel could be a lifesaver. Why a chatbot? Start with, it is a conversational agent. So it's unlike lists of fax FAQs and other kinds of help pages that we've been creating for years. It gives one answer to a question that comes through and not a list. It is text-based. The one that we developed is text-based, which is hence the term chat. But it could include sound and visual effects. There are other chatbots around that do that. It's a technology that undergraduates are very familiar with. It's kind of like the gaming software that they use so much. The software that runs Pixel is based on a database that includes information, a single bit of information that is organized or managed through metadata. And that metadata matches against the person's query to bring back the result that the person is asking for. It's available 24 hours a day, seven days a week, so it's ready whenever the patron is there. All of the information in the website can be added into it. I tried to get all of the different pages included into the metadata into the database, so it can actually flatten the website so people don't need to know where to go to get information. It isn't intimidating to people. And from the chat logs, I can tell you people are not intimidated at all by Pixel. They're not afraid. Well, it can be kind of amusing when I see arguments going back and forth. You said this. No, I didn't. Yes, you did. Pixel does have personality, and that does happen sometimes at times. But people aren't afraid of asking this super question. And the other angle which sometimes gets overlooked is that you don't have to worry about the person on the other end. I know some parents will not let their children use chat because of the fear of a stranger danger, but a software program is not something that they would have to worry about. Finally, the software can handle multiple questions at a time, which is kind of nice. There are never going to be a queue set up for people waiting for an answer, and it could literally handle hundreds of questions if we get so lucky as to have that many. Now I'm going to turn it over to Lorna. Thank you, Ben. So what can Pixel do? Hi, I'm Lorna, and as Diane said, I came in about only a year ago, and what I really wanted to do with Pixel was have a lot of conversations with Pixel to see how friendly Pixel was and to see how close to a reference service Pixel could get. So we found that Pixel can help the reference librarian and the patron, and for the reference librarian, it can help, especially new librarians, navigate the web page. As Diane said, some of our web pages are very dense, and what Pixel does is Pixel will search the website for us and find what we need on the website. It can also access all those easy questions that as a reference librarian we think we should know, but when we're new again, we don't know some of the common questions. Some of the questions are in the frequently answered questions, and sometimes it can get very embarrassing when somebody comes and asks simple question, directional question, and you don't know. Well, Pixel is never new. Pixel knows everything, and Pixel will find those directions for you, which is really good. I love libraries. Sometimes those directions are impossible. It is, and Pixel can help. It is very difficult. And for the patron, of course, once again the simple answers, it can give database instruction, it can help them articulate in requests, and also Pixel doesn't show in patients like we do as humans, and sometimes that's very important. You know, Pixel will continue and continue. Pixel always remains polite, and that is something that we find very helpful. Even when the patron is? Even when the patron is. So what I did is I went through our guidelines, and we have research guidelines for how, as reference librarians, we should conduct ourselves at the table. And so I looked at the guidelines that in academic libraries, as Arusa guidelines, and those are the guidelines that we follow, and they have five guidelines that we should try to follow when we are given reference. And so when we're training staff in reference, these are the guidelines we follow. So I went through Pixel, and I wanted to see, and Diane went through, and so each time I went through a guideline, I went to examine how Pixel relates to this guideline, what she does, and what we can improve. The first thing, of course, is approachability. We know that when people go to the reference desk, it allows them to be approachable. And so we were very careful to make Pixel approachable. And I will go in and show you how approachable Pixel can be with just warm greetings. We positioned her prominently on the website. Originally, she was in a beta testing zone. Now she has moved to a prominent place on the website. So if somebody is going into the website, they can immediately go to Pixel. And I want to go to Pixel now, and let's see how approachable can I get rid of this? Let's see how approachable Pixel is. So I'm... Oh, no, there's no text. Right. So I'm just going to say hello. Let me see if Pixel will tell me who she is. Do you really think Hello is who you are? Okay. My name is Lorna. Now she could have been rude, but she wasn't rude. Hello Lorna, how can I help you? I think she's very approachable. I think she's very approachable. And so what I'm going to answer is, since it's coming out for Christmas, I'm going to ask Pixel if she has any books on baking. What do you have? It doesn't take spelling really well. Let's see what Pixel will say to us. I can look that up for you. She's a very kind person. All right. And she looks. But the thing about Pixel is Pixel is approachable and we made sure that she was approachable. Next thing we wanted, which sometimes can be very difficult as I've learned, we wanted Pixel to show an interest. So she number one, we know that Pixel does answer in a timely manner. Again, as humans sometimes, we cannot get to our patrons in a timely manner. If we're the only ones at the reference desk, then sometimes it's very difficult to get to them in a timely manner. And what Pixel does is she continues the communication, which is something we try to do on chat. And she also demonstrates a commitment to provide information. And if we look back at the thing that I did, she demonstrates a commitment here by telling you, number one, I can look it up for you. So I'm committed to finding it for you. And then she goes on to say, if you need illustrations, check the facet on the left of the catalogue. So what she has done is she searched our catalogue for us. She's also said to the patron, but if you don't want these books, if you want illustrations, you can look elsewhere. And so what I want to tell Pixel is I think 240. I think that's too many books. So I'm just going to say, I think that's too many. And see if she's still committed to talk to me or on she says, you really think that's many books and I say yes. She's still committed. You know, it's a funny thing. It's a good search. This is what she's saying, but she's very polite with that. So she does have the way where she is very committed. Again, another one of our guidelines is that we have to show that we are listening. And that is what Pixel does effectively. What she does is she has to show that she hears what you say. She can repeat the question, but she has to show that she understands what you say. And of course, this may be one of the difficult things to get a computer program to do because they can't see the body language. And this is one of the things that is very important on face to face and is not on chat. The people want to see the verbal cues. This is something that we are trying to get into Pixel, but of course you have to do it with text. But how does she show she's listening is a question. So I want to write a thesis for research paper. Right. This is going to be interesting. How do you write her? All right, let's see if she heard me. Okay. And the important thing is even if she doesn't hear me, I want to know. So does it have anything to do with science? And I'm going to say, of course it does. Why wouldn't it have anything to do with science? All right. So she heard me. But again, again, she is not sure. And this is where she shows that she's heard she might not be clear. And this is one thing that we will build into Pixel. It's those open ended questions. It's those neutral questions that get her to ask for something else. I've heard that you want resources on research and science. But if you're interested in more articles, I'm going to put you on to our resources page. And this is what Pixel does. Oh, wrong page here. Oh, I need to go back. So what we need to do here, the other thing that I'll show that Pixel does, which we don't do as librarians, which is very important, is the teaching aspect of reference. The teaching aspect of reference says that you actually try to make the page on an independent learner by showing them your search strategy. Pixel obviously does that because what Pixel does is Pixel actually brings up the whole list of where you can go on the website. And that is very important. And later on, Diane will come and show you the sequence in which Pixel searches. Pixel will go to a website and she will go to live guides and she will go. So where am I still on the sciences? Okay, so she will go on the resources. So I'm going to say this did not help. Let's see. I'm not sure what she's going to do now. What will you do with this information? Again, she's trying to say, what are you going to do with this? I'll get a paper. An assignment. Shall I put a paper? Let's see what Pixel will do. What will you do with this information? Alright, now Pixel will stop listening. So I'll say bye. Thank you. She still wants to know which is interesting because the last thing we have showed Pixel to do is follow-up. Follow-up is important because what Pixel and what we have to do is number one, know if we've helped the patron so that we can help the next patron and also encourage the patron to come back. And so Pixel loves to know if she was helpful and I just want to see if I can get Pixel to be rude, but I can't. So I said to Pixel, no Pixel, unfortunately, you weren't helpful, but she's sorry, which shows that she's polite and she wants to know what she can do to improve. So what we have tried to do with Pixel is Pixel also has to adhere to the same guidelines that we adhere to as reference. And for us, that's important because it ensures that she maintains a certain standard of service and we do these guidelines so that we maintain a standard of service. And that's how we get Pixel to maintain the standard of service. So I'm going to hand back over to Deanne and she's going to tell you what happens in the background. What Pixel does behind the scenes? Oh, you went I'm going to skip through these. These were things that Lorna talked about. Okay, so I'm going to kind of go under the hood at this point to tell you a little bit more about the technology of how it's built. There are a couple of ways to build a chatbot. There are commercial sites that you can work with, subscribe to, and they will post and manage the chatbot for you. I'm not going to talk about those today because we don't have that kind of money. There's also open source projects which are much more attracted to libraries. Two big ones that I'm most familiar with are Pandora Bots and Programo. Pandora Bots is a really nice site because it hosts the chatbot for people and for nonprofits it's free. So if you're interested in playing around with creating your own chatbot it's a great place to start. Programo is what we did at UNL. It's an open source code that was developed that we've done a little bit of modification on. As I go through this I'll explain a little bit about what we did. There are other chatbots around. There's Emma the catbot which was developed at it's actually a cab if you go to the site and it's a talking cat too so you might find that interesting. It was developed from interpublic library and then there's Stella and Lisa, those two European chatbots. Stella if you know German is a great one to go play with because that one has been around for a long time so it's one of the more developed ones that's out on the internet. For the database it started by pulling information from the library website, circulation hours, kind of basic questions that people will ask. How do I get a book for interlibrary loan? For community users can we check out materials? That kind of information. Those can get kind of complicated because we do have different rules for different types of patrons. Community users have a different circulation, policy than undergraduates and faculty do. We pretty much don't have to follow any rules. We use the built-in chat log and add information when pixel doesn't respond appropriately so we try to keep continually adding to the knowledge base behind that pixel uses to answer questions. We also use chat logs from the real people in the library, the librarians. We have 24 hour chat question point service as well so I mind those logs periodically to get information that I can add into pixel. Who is pixel? Pixel is a chat bot that was developed using ProgramO. We added some customization to the program, some display coding for bolding and the produce list so it displays a little bit better and the frame that appears at the bottom of the screen that you saw a minute ago. So the front end there's the box, text box where the person types in their question and then where the book is is the iframe that gets rotated based on the question the person puts in. If the question has a link, Pixel's program executes that link and displays it in the iframe on the bottom part. It displays the first one if there's more than one. Here's another example of that. Who is the president? So it goes out and pulls that information out and displays the image at the bottom with the link above it so if someone clicks on the link it opens it in the tab. That's helpful when they're going into the catalog because they probably want a bigger window to do searches with. A meal is the metadata that's behind Pixel and it means artificial intelligence markup language and the Alice Box site is the original site for all of the information on a meal. So if you want to know more about the metadata, that's a great place to start. This example is from that site. This is a really basic a meal file. It consists of a single category and the category is one bit of information. The beginning and ending tags are category. Then the pattern metadata is what the person inputs. So whatever is in the pattern is what the person has input. In this case, if somebody asks the question, what are you, then the template part is what the bot would respond. So the person asks, what are you? The bot would say, I am the latest result in artificial intelligence, which can reproduce the capabilities of the human brain with greater speed and accuracy. No pride in that statement. The thick part of the template is instructions to the bot. So that is setting variables, which is something that a meal supports and it's very helpful in keeping conversations going so that Pixel can understand from one statement to the next what we're talking about. So if the user after this typing in this statement would say, well what is the topic, what are we talking about? The bot would respond, we're talking about me because the topic has been set to me. So that basic concept of setting variables is really important for keeping the conversation going. There are in my little world three ways in which questions get asked. And this is one example of metadata from Pixel. It starts with people just sitting down and typing words. And I suspect this is happening because they know this is a library site, so they just sit down and start typing in keywords like they were doing a catalog search. So they might sit down and just say corn, in which case Pixel needs to know what to do with that. They could also say star corn, star being a wild card variable. So it will match on anything that comes before the word corn. So if they say would you have corn on growing corn, that would match that particular category. The same for the next one. If they said corn information please, then it would match on corn star. And then the last one star corn star. Do you have any books on raising corn for the paper I'm writing for school and it's probably due tomorrow morning. And could you please find me the information really fast and it goes on and on and on. Sometimes they get really, really, really long. So you have to be able to match on the key phrase and what works best is to look for the most specific match point that I can find. And so in this case it's corn. It could be paper for the paper I'm writing for school if there was no topic. But in this case I have a really specific match point corn so that's what I would be shooting for and matching in the metadata. It gets a little bit more complicated. The asterisk or star I talked about a minute ago. It's a wild card match. It's also can be used to refer back to as a variable field too. I'm not going to go into that right now. The underscore is a more what I would call a hard match and it has to be used very carefully. So an example if I had a pattern that said I underscore that would literally match on anything that a person put in if they start off with I want information on corn. I want information on your hours. I want information on the days of the week or I want to know what time it is. It would match all of those and reply with a single category. And that's not good. So it has to be used with a little bit of care. Then the other term is seri which is like a cross reference in library vocabulary. In the previous example I had four different versions for corn. While using the seri pattern I only have to put the actual information in one of them and then I use the others to refer back to that one which makes it a lot easier to manage instead of having to do everything four times I only have to do it once. And that's what the seri tag looks like. It has just the metadata that says seri between the tags and it refers back to whatever is above that. Subject and topic categories are something that I have had to implement for a library application and we always originally designed as a conversation bot just for people to go in and talk to a bot and not particularly give any kind of information. So I added these two categories areas to provide layers to keep the conversation going. These came from a library congress classification code. There's a topic layer and a subject one. The topic is a broader one so in the case of our corn example that would be agriculture and subject would be the narrow one which is corn. The purpose for these is to keep the conversation going and it relates back to the way libraries are kind of organized. Our databases tend to be more broader topics. We don't have a database on corn for example. We have a Gricola which includes a lot of information on corn but we don't have anything really specific. The same is true for the librarians assignments and for the lib guides that we're developing or the help guides that we're using. So there really are kind of two layers to it. We can do a direct search on corn and provide the information people need that are on books and things like that but to go to the next level the articles we need to move up to the topic level whether or not those two levels. So in the metadata for the category of corn there's the topic is set to agriculture the subject is set to corn so that will be used in subsequent conversation that could happen with this particular question. So the person would say corn in this case or they could use any of those four matching categories and it will bring this one up and the bot will return are you interested in books and then it does a search in our discovery tool on-core on the word corn and then it's going to follow that up with some additional options. So we start with the books that's the first thing it's going to offer which is based on the subject which is corn then it's the person says no it will offer articles and at this point it's going to the topic level which would be agriculture then it goes to lib guides if they say no I don't want articles either we're talking about somebody who really probably doesn't know where they're going and so a lib guide might be most helpful to help them get started and that will bring up agriculture lib guides that have been developed by librarians and then it finally goes to the librarians for that particular subject area some people have questioned well why do you put the librarian last right Lorna I think that was the first question she asked me and the answer is because whenever pixel offers to send people to librarians they never go there I think if they wanted to talk to a librarian they'd be using a chat service or go to the reference desk so at this point in time our experience is that that's not what people are looking for so that's the current order that of course could be changed at any time so the back end of this how this actually works in the previous example where the bot asked are you interested in books on corn now you see I use an underscore this is a hard match because if the bot previously said something that had that in it I want them to go to one of these new categories and this in this case the pattern is now so they're saying no I don't want books on I don't want books on corn so then it asks are you interested in articles on corn there are several databases that will have lists full text journals in your subject area and at that point if they say yes does that being responding to what was previously supplied by the bot then if they say yes then we will do the research articles and in this case we have some open source databases that are also available to the general public so it offers both of those so how does this work here's the first example I'm looking for information on raising corn and so it's offering the books first then the person says no so then it goes to the articles and now it's pulling up our agriculture articles so because of the search syntax is customized to the particular category it can be as specific as we want it to be which gives us a lot of versatility in providing exactly the right databases that least I think they need until somebody tells me they want change it. For the information this is coming from the logs and this is an example from a question point log and sometimes there is cryptic as this how do I make a refworks in this case I made the assumption that they're asking about a refworks account and so in this case it's supplying the information if you don't already have a refworks account you can go here to set one up and then it gives them URL links them to the site to set up an account and if you're not familiar with refworks if you're not coming from on campus you need to have a group account code to put in and most of the time people don't know that so by adding this series at the end of it it's pulling up that separate category that has information on the group code so you can actually combine two different answers when it makes sense to do that is it within this case Pixel also has a log here's another example I need help with my project this is a very generic sort of request and since it doesn't give you any kind of specific subject information what I do in these cases is I pull up a random question to ask as a follow up that ability to do random questions keeps the conversation fresh which I think is important because sometimes users will type the same thing and expecting the same answer and Pixel gives them a different answer so it makes it seem more like you're talking to a person some basic information about Pixel there are 80 currently 84,000 and growing categories in the database there's 204 spelling variations which isn't nearly enough that's something that I need to be working more on the ProgramO interface that I use for administering the bot has a built-in functionality for adding and editing single records and groups of records has a search feature it has a log for conversations and you can demo the chat which is useful when I'm adding new categories I want to test and make sure that they're responding the way that I want them to respond this is some Google Analytics from the past year that are kind of interesting especially when I compare them with our catalog and our discovery tool on Core for example, the average duration of visit on Pixel is almost 6 minutes which is actually longer than they spend on our catalog and our discovery tool so that at least tells me that people are engaged they're sort of enjoying talking with the bot sometimes a little too much I think there's also kind of interesting that 37, almost 38% are returning visitors which is actually higher than our catalog and it's kind of like our catalog in terms of its placement the discovery tool is the first thing on our website that we encourage people to search first so I think that these statistics at least demonstrate a lot of potential for engaging a particular niche of the population this is never going to be a tool for faculty and probably graduate students unless they want to know about finds and things like that and librarians won't use it except for locating information on our website as Lorna was talking about it's a pretty good way to find quick information so future plans for the bot we are going to continue to evaluate how it fits into our current reference structure we have a reference desk at Love Library, the main library we have 24-hour chat we have live guides we have instant messaging we have Twitter, we have Facebook we have the whole smorgasbord that everybody is doing and so we're experimenting with a whole bunch of things to see what's going to stick to the wall basically we're also going to be looking at the possibility or feasibility going into production with other libraries I'm looking at a grant there's a grant that I think Pixel would fit into so I'm looking for partners like a four year undergraduate college or even a public library to look at moving Pixel into a semantic web type architecture where there could be one database of shared information things like who's the president which is going to be the same regardless of where you're at but then there would be individual databases for each library that could be customized for their own search engine for their own hours and that kind of information so if anybody out there is interested they'll have my contact information and the course assessment we haven't done any direct user assessment yet and that's going to be something that we're going to try to work on in this coming year so that's it well thank you I've got a bunch of questions does anything come in from the audience at the moment let's give them a first crack here yeah Dave McStore from South Sioux City Public Library just said how much time was spent on putting in information for us general public libraries not a lot of staff what would be a recommended program for this that we don't have a lot of time to put all this information in I think that if I was going to I would look at scaling it down and probably say build one that just answers questions from your website the kind of FAQ answers what are your hours are you closed on Christmas how many books can I check out at a time and not try to be too ambitious in terms of adding a lot of other information if you scale it down I think it would be a whole lot easier to manage when I started this I spent a huge amount of time building the database and figuring out a lot of what was figuring out how it would work and whether it made sense for us or not but I think if you scale it down it could be a lot more manageable now I don't actually spend probably as much time as I should working on it going through the logs well and at least one of my questions you got the software and I'm assuming there wasn't really much of a database in it when you got it you had to build it actually that's not true that's good I didn't mention that where the baseline comes from that's a really good question because Emil has an open source community behind it to answer questions and the basic conversational part was already built in fact there's what are called pickup lines that come with the software that are designed for continuing the conversation and they were things like well what's your sign not something not something that was appropriate for a library that changed all of the pickup lines to things like coming back a project what is your topic has anybody ever asked what's your sign yes in fact Pixel's been asked out on dates to go do a movie constantly people are trying to pick up Pixel but she has a boyfriend so he's been asked that too well and as at least the only male in the room including Pixel I guess my question is do you have a conversation about male versus female Pixel well since I did it no okay unilateral decision yes I did it for the whole lot and it was pretty easy to change that picture to whatever I remember setting up some early social web stuff here and I was setting it up so I made it male and there was a conversation that was had I've seen her call at least one of them why is it male I had to pick one there was only two choices any others just speak up if anybody has any questions go ahead and type them into the question section of your go-to web interface on the right side of your computer and we can pass them on I'll kind of basically try to go in a general order I was writing them down as you were talking you kind of implied possibly this reducing for 24-7 chat yet the impression I guess you still have 24-7 chat do you see a point at which this does take over at 3 a.m. or well there's still it's still a very controversial project in the library to say the least and I don't know we're evaluating 24-hour chat it's very expensive for the hour of the question rate it's pretty cheap now since it's been built it's much more cost effective the question is can people be convinced to give up the person contact which seems to be hard for librarians to do so it's the librarians you need to convince that's right in fact I'm reading a couple of logs there were at least two people who were convinced Pixel was a human being and there was this argument going back and forth really person no I'm software no I think you're that went on for several lines so I mean it's it's possible I think but it's kind of it's a control issue I think I think I see both sides of that argument I'm not sure where I would come down there I remember for your presentation at IL that there's a point at which you need to apologize to Pixel there's this list of bad words that people can use that when somebody asks one of those Pixel will come back and say no you got to apologize I'm not going to talk to you that you apologize and that can go on for pages where the person refuses to apologize Pixel just says no I'm not going to talk to you I'm not talking to you I apologize and they do they do wow all right it's a behavior modification never heard so maybe a little more towards the technical ends of things have you tested this how well does it work it should work pretty well where it's going to get dicey is with the iframe whether that's technology supported by and most of the modern phones is pretty good with but somebody has an older phone it might not work so well with that one um the the and I want to ask you to go to it but so there when you're creating the XML that's behind this you have an editing interface like a form to fill you're not writing this by hand actually I ended up doing most of it because once you learn the metadata code is pretty easy to do and I create templates for the subject area so I just copy and paste and just change as I need to the parts of it and now goes a whole lot faster than using the editing interface you can do either neither yeah I am used no tab probe to create the category files and I have like I think I have 26 subject files that cover all of the all of the subject areas of the library congress classification and thousands of entries in them and do you know how many lines of XML you have in that no I have no idea hundreds of thousands of not really that would be interesting I was just like how many lines so far um what sort of I know you said you mind the chat logs um to to get new things in for that people are asking for have you gotten um larger faculty input as to what should be in here or reference staff or beyond how else are you adding the only one who's actually giving me much input is I have asked the reference people and they never supply anything but when I started this project over two years ago they said it wasn't necessary because they were going to build their own FAQ system which never happened either so now we're using live guides and I'm hoping that things will move forward a little bit that's probably where most of their energy is going right now since the live guides also just recently purchased live answers so that's another competitor we'll see how it goes and if you don't want to answer I know politics being what they are do you think that's a passive aggressive we don't want it to succeed we don't like the idea in the first place so they're just like I've got too much else to do all of the above and I think insane that I'm not saying I think part of it is sometimes it is not knowing I think at the beginning they really did not they felt threatened because they did not understand how pixel could be a part of the service and I think sometimes just like we're not threatened by Twitter and Facebook and all the others I think there is a place where we can offer services because our clients are different we can offer different services to each client and I think that's where pixel comes in so in a way I don't really see pixel even replacing chat because you do have some patrons who will prefer to use chat and you have some patrons who will always use pixel but as librarians we know that we are we want to provide a service to all and we want the service to be equivalent to all so pixel and I think now because so many new things are coming I think a lot of reference librarians are realizing that there is a place pixel will be able to answer certain questions and chat will be able to answer certain questions and the truth is I'm a librarian and I call our help best a lot you know I call because I know it's the quickest way for this answer but I call them for specific answers and then I will go to pixel for specific answers so I think it's a matter of time and slowly I think now the reference librarians are seeing where pixel is in the reference flow then they say this is a real help for us because while I'm dealing with one client here and I have my chat online another patron can come and I can say can you ask pixel that question because she can help you with that while doing this so I think it's moving on I really think it's a great opportunity idea I hope that answers yeah no that was a wonderful answer it's evolving it is very much a good idea and as you add more information it becomes more useful and so that it's kind of snowballs I guess for lack of a better and I think there are certain things that pixel will never do you know and I think we're always aware of that there are certain things and that's why I think pixel is very good in sending as you said the librarian comes last not really last it's a matter of pixel has a way because we're built it in of evaluating what sources are the best sources and so it will go to these sources and finally if these sources can't help you then you need to go personally to the library yeah and so it has that kind of new technology always does that there's a great 1945 Wilson library bulletin article about you know should we be doing reference over the telephone because it was a threat right because you know can you actually get that personal interaction with the person and you can't either body with and so and we've come past and then of course it was with chat you know when you get the emotives and the emotions how do you know what they're saying right question from the audience yeah we do have one from the audience many patrons want to talk what you were just talking about anyways but any patrons want to talk to a real live person also many people are not computer literate so how can these problems be addressed and I think that was what you're talking about that having all the different out there yeah we will never eliminate the telephone and and and we will never eliminate having a person there and but what we have tried to do with this virtual reference is to address the people who don't want to come in for the person so you do have a set of the population that are not like that that they don't want to or they don't see the need or they don't have the time to come in and they want to be able to do it from home and still get some kind of person and that's what pixel the other thing we're trying to do is to make it as low-tech as possible so all we have to do is type in right whatever comes to them it's very simple clean it is a very good I would think that very scary to someone who doesn't go on computer you were showing the google analytics there which was great can you measure the spikes to like end of semester well the big spike was when it was entered in the chatbot challenge oh okay interesting what is that the chatterbot challenge was they don't doing it anymore it was an annual chatterbox contest where experts went in and evaluated chatbot and then rated them so and it was we tied with Alice for 10th place I believe are you getting any sort of analytics regarding the users graduate undergraduate staff faculty no have you considered it or just not really a way to do it that's why we'd have to do assessment because we try very carefully to keep it anonymous I mean I can kind of tell from the questions I know the librarians when they come in you know our language right the way they ask you can tell kids we get an awful lot of elementary school people and they don't ask anything and everything I had this one conversation where a kid was complaining about being bullied those kinds of things are hard to program for but you have to be prepared for anything a lot of foreign people come in and try to practice English oh okay there is actually a chatbot for that but they're not using it can you program it to point them in that direction I haven't tried that yet what happens is that Pixel will say well we have books in Italian or whatever right and kind of maybe the last question I have specifically when you were going through and one of the answers said if you want to narrow it down click on the facet over on the left and that just shouted out at me a librarian word it did well you know I'm not sure I have a question but have you done I guess the question would be have you done usability testing with non-libraries no we haven't done any that's the next the assessment is the next stage we haven't done usability testing on our discovery tool or our catalog guide it's one of those weaknesses I think we have that we've kind of started addressing but that's what we call it as a facet oh I knew what you meant I knew what she meant I'm just thinking you know 18 year old student coming in all things are just really really hard you know that's one of the things we struggle with constantly arguing over what do you call last thing we call an open URL resolver you're about to get an argument started ask that question oh yeah well and I don't I don't necessarily have a better suggestion I just saw facet that is a librarian word it is but yeah that happens any other questions coming in from the audience not at the moment I don't really have any news this week other than windows 8 is out so you know hey that's kind of the big news since last month and I'll let you go microsoft.com I don't think we need to send people off off to that website I'm using it at home I actually like it but you know I'm weird that way a lot of people don't anyways so thank you very much this was wonderful I always try to I know what people are going to talk about but I try not to research it too much so I can kind of answer come up with questions on the fly and I had a whole bunch of them there and I got all the answers I was looking for and we had a couple of answers questions coming from the audience so thank you for that and I think that's so I'll end it thank you sure yeah alright okay well thank you very much for attending this week it has been recorded and we do have the powerpoint as well so that will be uploaded when the recording is available the powerpoint will be there and I've been grabbing some of the websites that we've been mentioning and putting them into our delicious account I know I still I missed something but I'll be going back and doing that so you'll have all the quick links to all those as well when the recording is up and ready so that will wrap it up for today's show but I hope you'll join us next time when our topic is book club kits reviews that's kind of hard to say here at the library commission we do put together collections of book club kits 10 or so books in a bag and we have a book club discussion on those titles and there are always new ones being added so Beth, Gomo, Debra Dracos and Lisa Kelly will be here with us next week talking about the new ones that are there and how you can use that program to run book club discussions at your library so please join us for that next week and Encompass Live is on Facebook if you are a Facebook user please feel free to follow us there there it goes and we'll post here anytime when episodes are coming up I had a reminder this morning join us right now for this week so anything that we're doing on Encompass Live you'll find the announcements on here so you can follow us there as well other than that that wraps it up and thank you very much Michael, Lorna, Deanne just a few thank yous from the audience and we'll wrap things up and hopefully we'll see you next time bye