PROTOCOLS OF USERS MANIPULATING VISUALLY PRESENTED INFORMATION WITH NATURAL LANGUAGE
Original Publication Date: 1982-Dec-31
Included in the Prior Art Database: 2005-Sep-19
Software Patent Institute
Related PeopleAuthors: Other Related People:
This document contains a complete set of transcripts for a set of protocols collected at BBN by Candy Sidner, with the "~ help of Rusty Bobrow and Jeff Gibbons, in the spring of 1980. In all, eight protocols were collectedv two preliminary ones and six main ones. The two preliminary protocols are based on the task of designing a 1 bit adder and a 4 bit parallel adder as are three of the main protocols, while the other three make use of RL-1 as a database system with graphic representation. The purpose of these protocols, and hence the design of the tasks, was to obtain data about how people talk about graphically presented material which they are trying to manipulate in some way. In particular, we were interested in the kinds of references people made and what sorts of instructions they gave to the machine. The transcripts given here contain all the instructions given both to the user and to the person who simulated a natural ""` language understanding system, often referred to here as the system or the machine. All the graphic displays drawn and all the communication between user and system are given as well. The slides of graphic displays have not been editted or changed in any way. As a result they look somewhat cluttered, and the quality of the reproduction is poor. However, they will provide the reader with a better understanding of the interaction between user and "system." Bolt Beranek and Newman Inc. Report No. 5128 The preliminary protocols permitted us to observe potential problems in protocol collection. Two persons, henceforth referred to as A and B, played the part of the system in both protocols, while persons C and P were users (respectively) for those sessions. A and B divided the task of being the system into one person to draw displays and the other to communicate with the user. The user and system communicated with a two way link over 0100 terminals with a transcript maintained from a TELNET log file. Person B, who drew pictures, displayed them with a view graph projector and the user pointed at the display when needed. One of our discoveries was that pointing directly at the display by hand or by means of a light beam visual aid was unsatisfactory. It was clumsy and so inconvenient that the user found pointing unmanageable. Also we discovered that high load average on the machine slowed down the communication link enough to make communication too slow to be effective from the user's point of view. These problems were remedied by 1) using the machine when the system load could be kept lower by excluding other users, and 2) by building a device to make pointing possible. To permit both the system person and the user. to point, we used two view graphs whose images were superimposed on the same place on a screen as shown below.