Full Record – Author

Ref Type: Journal Article

Authors: Moore,Mary, Ahmed,Terry, Glazer,Adam,

Title: Using and automated knowledge agent for reference and customer service

Publication Full: Journal of the Medical Library Association

Year: 2004 Volume: 92 Issue: 2

Start Page: 271 Other Pages: 273

Descriptors: NATIONAL libraries, MEDICAL libraries, CUSTOMER services, AUTOMATION, INFORMATION services,


The article discusses a study regarding the use of an automated customer service agent, developed by the National Library of Medicine (NLM) using software from NativeMinds, for reference and customer service. NativeMinds uses a proprietary language called Neuroscript, a variation on Java-enabled C++. Natural language processing identifies key concepts and discards irrelevant information. Reference librarians reviewed customer service question logs, identified common questions, scripted pattern lists and categories, and wrote responses to questions they thought were most likely to be asked. The agent was given the name Cosmo (Customer Service M. Owl) and was graphically represented by an owl. Scripting, testing, and auditing of the agent continued, and Cosmo was placed on the external Contact NLM Customer Service Web page. The service now includes more than 900 answers to common questions. In April 2003, users asked the agent 466 questions. Cosmo answered 51% of these questions correctly and 22% incorrectly, 20% were out of scope for the agent's job description and 7% were out of scope for NLM. The service is available at all hours, provides immediate and consistent responses to questions, relieves boredom for humans from answering repetitive questions, has the potential for cost savings, and refers more difficult questions to human agents when a personal response is needed. Cosmo is performing acceptably. Having an added choice for customer service that is available at all times of day and night and can answer basic, repetitive questions can benefit both clients and librarians. From the perspective of quality filtering of information and managing information overload, intelligent agents may be very useful.


ID: 358; M3: Article; Moore, Mary 1; Email Address: moorem@mail.nlm.nih.gov Ahmed, Terry 2; Email Address: ahmedt@mail.nlm.nih.gov Glazer, Adam 2; Email Address: glazera@mail.nlm.nih.gov; Affiliation: 1: Head, Reference and Customer Service, National Library of Medicine, 8600 Rockville Pike, Bethesda, Maryland 20894 2: Reference Librarian, National Library of Medicine, 8600 Rockville Pike, Bethesda, Maryland 20894; Source Info: Apr2004, Vol. 92 Issue 2, p271; Subject Term: NATIONAL libraries; Subject Term: MEDICAL libraries; Subject Term: CUSTOMER services; Subject Term: AUTOMATION; Subject Term: INFORMATION services; NAICS/Industry Codes: 519120 Libraries and Archives; NAICS/Industry Codes: 519190 All Other Information Services; Number of Pages: 3p; Illustrations: 1bw; Document Type: Article

Links: NULL

Extra Data: NULL


Retrieved Date:January 1, 1970

Editors: NULL

Book Title: NULL

Series Title: NULL

Publisher: Medical Library Association

Place Of Publication: NULL

Issn Isbn: 15365050