Digital White Papers

37 WWW.ILTANET.ORG | ILTA WHITE PAPER KNOWLEDGE MANAGEMENT Automating Knowledge Delivery Using Knowledge Bots » Who can help me reset my Lexis password? » Do we have a fiy-state survey on [e.g., recordkeeping requirements, workplace training requirements, wage statement laws, etc.]? By fielding these types of questions, KM professionals help aorneys leverage pre-existing work product and thereby become more efficient. At the same time, the KM professionals answering the questions are becoming more efficient too: by analyzing these questions over time, KM departments can identify which resources aorneys request most frequently, which resources are most helpful, which resources should be developed or purchased, etc. By staying abreast of the firm's questions, a KM department can become increasingly efficient at providing comprehensive answers to those questions. To save time in answering frequently asked questions, our KM Counsel team (also known as practice support lawyers in other firms) saves prior responses that can be easily recycled. For the most frequently asked questions, KM Counsel can save the answers as "Quick Parts" in Outlook to auto-generate future email responses. While these techniques help to automate some of the question- answering process, they still require KM Counsel to be actively checking email, to know that someone on the team has answered the question previously, and to spend time and aention generating the response. Furthermore, despite timesaving techniques, a high volume of requests can consume a significant percentage of a KM Counsel team's time. Time is an especially precious resource for a lean department, and time spent on repetitive tasks, no maer how efficient, is time not spent on high-value, big-picture projects. The Solution: A Knowledge Bot As with any repetitive work process, our process for responding to frequently asked questions was ripe for automation. We had already compiled many template answers to common KM questions. However, having had prior experience and knowledge about chatbots and NLP, I knew a knowledge bot could handle many of the questions we received faster, cheaper, and around-the-clock. Knowledge bots are surprisingly not a "cuing-edge" new tool. In fact, knowledge bots are an older technology than email, floppy disks, the computer mouse, and pocket calculators. The first "automatic question- answerer" – "a computer program that answers questions phrased in ordinary English about stored data" – was created by Bert F. Green, Jr., et al. in 1961. That program, BASEBALL, was a simple chatbot that answered natural language questions (entered on punch cards) about baseball data for each game in the American League over one year. A few years aer BASEBALL, Joseph Weizenbaum created ELIZA, a chatbot with improved NLP abilities. ELIZA accepted input through a typewriter connected to a computer, processed the input through a "language analyzer," and then responded to the user's inputs according to a script. ELIZA's first extensive script "enabled it to parody the responses of a nondirective psychotherapist," because Weizenbaum wanted "to temporarily sidestep the problem of giving the program a data base of real-world knowledge." Despite its simplicity and its lack of a substantive knowledge base, "ELIZA created the most remarkable illusion of having understood in As with any repetitive work process, our process for responding to frequently asked questions was ripe for automation. We had already compiled many template answers to common KM questions.

• Cover