The document discusses the development of Kornelia, a chatbot for the Kornhaus public libraries in Bern, Switzerland. A team including students, an IT professional, and a designer worked to create Kornelia to bridge the gap between libraries and digital users. The team developed Kornelia using Program E and the AIML language. She was tested internally and with users to improve her knowledge and usability over time. Kornelia has been online since July 2009 to help answer questions about the Kornhaus libraries.

1. BRIDGING THE VIRTUAL AND THE PHYSICAL SPACE Kornelia – A chatbot for public libraries Eva Dohrmann Jasmin Hügi Nicole Scheurer Angela Trummer René Schneider Faculty of Information Science Haute école de gestion, Geneva (Switzerland)

2. Table of contents Chatbot Project: Teamwork Kornhaus libraries IT Professional Student Group Designer Testing Conclusion

3. Chatbot: Definition Combination of the terms “chat” and “robot” Simulation of an interlocutor Server application replying questions posed via Internet in a input field Chatbot – Project: Teamwork – Testing – Conclusion

4. Chatbot: In libraries Means to bridge the gap between libraries and digital natives Chatbots were usually implemented in scientific, academic libraries Needs of public libraries’ customers are different Punctual questions such as “Is the library open?” Chatbot – Project: Teamwork – Testing – Conclusion

5. Chatbot: Examples Stella  Bibliotheksystem Universität Hamburg http://www.sub.uni-hamburg.de/ BoB  Free University of Bozen http://web.inf.unibz.it/bob/?popup=yes&language=en Chatbot – Project: Teamwork – Testing – Conclusion

6. Project: Teamwork Kornhaus libraries Student group IT professional Designer Chatbot – Project: Teamwork – Testing – Conclusion

7. Kornhaus libraries: Context Network of 21 public libraries in Bern Chatbot – Project: Teamwork – Testing – Conclusion

8. Kornhaus libraries: Why a chatbot? Implementation of a new information system A chatbot hosted on the homepage Accessible 24/7 Fun approach to the library and its catalogue Meeting point between people, information and technology Birth of Kornelia Chatbot – Project: Teamwork – Testing – Conclusion

9. IT professional Chatbot – Project: Teamwork – Testing – Conclusion

10. IT Professional: The chatbot’s architecture Chatbot – Project: Teamwork – Testing – Conclusion

11. IT Professional: Technical development Program E Written in PHP and uses MySQL Chats saved in log files Chatbot – Project: Teamwork – Testing – Conclusion

12. IT Professional: Program E Chatbot – Project: Teamwork – Testing – Conclusion

13. Student group Chatbot – Project: Teamwork – Testing – Conclusion

14. Student Group: AIML - Definition Artificial Intelligence Mark-up Language XML derivate Content can theoretically be structured with only three tags Content based on rules containing requests and their corresponding answers Chatbot – Project: Teamwork – Testing – Conclusion

15. Student Group: AIML - Rules <category> <pattern>What is your name</pattern> <template>My name is Kornelia.</template> </category> Question in the “pattern tag” Answer in the “template tag” Chatbot – Project: Teamwork – Testing – Conclusion

16. Student Group: AIML - Wildcards <category> <pattern>Who is *</pattern> <template> I don’t know who that is.</template> </category> Special characters (“*” or “_”) which replace one or more words Just a fraction of the input will be encoded The chatbot recognizes the phrase structure Chatbot – Project: Teamwork – Testing – Conclusion

17. Student Group: Content Three types of content Social chat General chat about libraries Kornhaus focused chat Chatbot – Project: Teamwork – Testing – Conclusion

18. Student Group: Problems (1) Large variety of question’s formulation Information depends on the particular libraries (ex. Opening hours) No connection between the chatbot and the catalogue  leading to confusion Chatbot – Project: Teamwork – Testing – Conclusion

19. Lack of information  divergence between Kornelia’s knowledge and the expectations of the librarians Informal / formal formulation of the questions / answers Student Group: Problems (2) Chatbot – Project: Teamwork – Testing – Conclusion

20. Designer Chatbot – Project: Teamwork – Testing – Conclusion

21. Designer: Graphical design Discussion about the chatbot’s gender Female  more trustworthy Choice between four character types Suitable for the Kornhaus and timeless Neither too young or skinny, nor too sexy Artificial style Chatbot – Project: Teamwork – Testing – Conclusion

22. Designer: Development Chatbot – Project: Teamwork – Testing – Conclusion

23. Designer: Facial Expressions neutral happy toughtful explanatory angry

24. Testing: Internal testing The log files allowed the continual improvement of Kornelia’s knowledge Monthly analysis of the log files  percentage of the correctly answers had increased Implementation of a quality gate  50 essential questions Chatbot – Project: Teamwork – Testing – Conclusion

25. Testing: Usability test Carried out by people not involved in the project and without any further information User and expert focused evaluation Axis of the test: Kornelia’s ability to answer The design The quality of her user-friendliness Chatbot – Project: Teamwork – Testing – Conclusion

26. Testing: Results Strengths: Innovative Fascinating Entertaining Weaknesses: Irrelevant questions Vague answers Lack of professional knowledge Old fashioned and unattractive Chat example as guide was missing Chatbot – Project: Teamwork – Testing – Conclusion

27. Conclusion Constant implementation of Kornelia’s knowledge Online since beginning July 2009 http://www.kornhausbibliotheken.ch/index.php?option=com_wrapper&view=wrapper&Itemid=64 Chatbot – Project: Teamwork – Testing – Conclusion

28. Kornelia

29. Thank you for your attention!