IRList Digest Saturday, 22 Mar 1986 Volume 2 : Issue 15 Today's Topics: Email - Delay on CSNET mail for Virginia Tech Call for Papers - News needed for IRList! Interest Statement - Information Management for Scholars Reply - Use of typesetting codes in IRList Query - Mail to ETH? Workers on intelligent front ends? Query and Abstract - Mail to DFN? Abstract on expert systems and IR Article - Part 1 of 2 on Human Computer Interaction at U. of Maryland ---------------------------------------------------------------------- >From fox Sat Mar 22 14:26 EST 1986 Subject: alternate routing The CSNET host machine at Virginia Tech has been down since last Sunday. Mail will get through in spite of other claims - CSNET is holding our messages. Meanwhile, IRList can be reached through other addresses such as those shown below. Thank you for your patience. - Ed Fox (BITNET[cheapest]:foxea@vtvax3 or foxea%vtvax3.bitnet@wiscvm.arpa; CSNET:fox@vt;Internet:fox%vtisr1.uucp@seismo.css.gov;UUCP:seismo!vtisr1!fox) ------------------------------ >From fox Sat Mar 22 14:33 EST 1986 Subject: call for news! Dear IRList Subscribers: You may be wondering why IRList issues are settling down to a once per week, or sometimes once per 1 1/2 week schedule. The main reason is that each of you are being derelict in sending in materials! There have been numerous questions sent to the list that NOONE has replied to. Please do not be shy! If some item in IRList perks your interest, or says something you agree with, or runs counter to your viewpoint, please send back a short commentary. If you are involved in theory, practice, or related fields to information retrieval, please send in at least a yearly summary of work going on. If you prepare a technical report, please send in the title and abstract. I look forward to more active participation from all of you! Regards, Ed Fox ------------------------------ Date: 11 March 1986, 00:16:55 EST From: James H. Coombs JAZBO at BROWNVM Greetings! Please add me to the IRLIST. I have a Ph.D. in English and an A.M. in Linguistics. Am currently Mellon Postdoctoral Fellow in English at Brown. I have developed an Information Management System for Scholars, currently in beta testing, and am especially interested in the problems of full-text retrieval and maintaining keyword consistency. Thanks. --Jim [Jim: Please send more details about your system! - Ed] ------------------------------ Date: Mon, 17 Mar 86 10:39:55 EST From: Jim Davis Subject: Re: IRList Digest V2 #13 I prefer to not see .troff commands in the IRList. [Note: since you are the only person to give an opinion, I will follow your suggestion. Thanks for the feedback. - Ed] ------------------------------ >From fox Sat Mar 22 12:31 EST 1986 Subject: Inquiry from Zurich H.P. Giger recently sent a letter with the following comments. Please send replies to him (or me or IRList so I can forward responses). (If anyone know how to reach ETH through electronic mail, please explain so we can speed this process up! - Thanks, Ed ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ ... I am working in an Information Retrieval (IR) group at the Computer Science department of the Swiss Institute of Technology. My basic interest is in coupling the power of a workstation (e.g. SUN) with an existing boolean based IR system in order to improve the end-user interface (e.g. coordinate level matching with a generalized Boolean Operator ranging from "OR" to "ADJ" using combinations of the Operators). ... Some ideas presented by Jamieson at the SIGIR Forum of 1979 (The Economic Implementation of Experimental Retrieval Techniques on a Very Large Scale using an Intelligent Terminal) are of interest to me too. The weighting of the elementary logical conjuncts differs to the one used in your method. I'm looking for persons, working in the field of intelligent front ends. Could you please send me a list of persons you know working in this area? ... H.P. Giger Institut fur Informatik ETH-Zentrum CH-8092 Zurich, Switzerland ------------------------------ >From fox Sat Mar 22 12:47 EST 1986 Subject: Inquiry and abstract Following are extracts from a letter asking about IRlist. If anyone knows how to make electronic contact with GMD, please advise. - Ed +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ ... In the latest issue of SIGIR Forum I just read your announcement of IRList. This seems to me a very interesting idea and I would like to learn more about it. ... I am interested in several of the topics you mentioned, in particular in AI applications to IR expert systems for IR full-text retrieval knowledge representation probabilistic techniques statistical techniques How can I get access to your material? Right now, I cannot contact you via electronic mail. This may change when DFN (Deutsches Forschungsnetz, a German network for use by scientists) may get access to US nets. I am enclosing the abstract of a paper that will appear in Nachrichten fur Dokumentation 36 (1985) no. 6 (page numbers as yet unknown). You are free to use it for IRList. Sincerely yours, Dr. Friedrich Gebhardt GMD, Postfach 1240, SchloB Birlinghoven, D-5205 Sankt Augustin 1 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Connections between information retrieval systems and expert systems Increasingly there arise connections and mutual enhancements between information retrieval systems and expert systems. They will be analysed here on the basis of the recent literature. Theser are the connections dealt with in sections 1 to 3: (1) With their growing sizes, fact and rule bases of expert systems can no longer be held in main storage. Since they in part consist of texts, the service of information retrieval systems is of advantage. (2) The quality of search results of information retrieval systems can be improved by an expert system. (3) Some less obvious connections consist of some applications of one system in the environment of the other, in particular in expert systems assisting in the acquisition and plausibility control of data for information retrieval systems. Other proposals concern the construction of highly complex systems that comprise abilities of expert as well as of information retrieval systems. Mentioned here are a system for text retrieval in office databases, a system for utilizing statistical databases, and a vision of a very general system for problem treatment. ------------------------------ From: edu%"ben@mimsy.umd.edu" 2-FEB-1986 22:10 Date: Sun, 2 Feb 86 21:55:46 EST From: Ben Shneiderman Subject: Re: old articles [Following is part 1 of a 2 part edited extract from materials describing research at the University of Maryland. Contact B. Shneiderman for a complete reprint - Ed.] Ben Shneiderman Department of Computer Science and Human-Computer Interaction Laboratory University of Maryland College Park, MD 20742 October 1985 The Human-Computer Interaction Laboratory (HCIL) is a unit of the Center for Automation Research at the University of Maryland. HCIL is an interdisciplinary research group whose participants are faculty in the Departments of Computer Science and Psychology and the Colleges of Library and Information Services, Business, and Education. In addition, staff scientists, graduate students, and undergraduates contribute to this small, but lively community that pursues empirical studies of people using computers. Our support comes from industrial research projects, government grants, the State of Maryland, and the University of Maryland. ... Some of our efforts during the past year are described below. We have been exploring ways to attach three or more separate screens to an IBM PC and use each screen as a window. ... We have conducted observational studies and built many variations of multiple screen formats under support from IBM Federal Systems Division in Bethesda, Maryland. These studies have led us to the following concepts: Fusion: Several screens are `fused' together at the bottom/top or right/left edge to form one, continuous logical screen. ... Copying: One central screen is used like current single screens. Several other screens are used to display copies of text copied from the central screen. ... This allows the programmer to save text for later reference or, the programmers can save part of an on-line manual or a directory listing. .. Direct Selection: As the cursor moves from object to object, information about the current object is displayed on the other screen. ... Synchronized Scrolling: This is similar to fusion of the left/right edges. One file is displayed per screen. The scrolling of the files is linked; the command up-one-line moves all files up one line. ... Independent Scrolling: All screens can be controlled separately. A programmer can look at arbitrary parts of several files or several parts of one file. Independence is similar to logging on to several terminals at once. The advantage is that you can copy from one screen to another. ... These concepts can be applied in many program development situations: Input - Filter - Output Situations: Using independent screens, the input to a program, the program text, and the output of the program can be displayed at once. This makes a very powerful debugging tool. ... Integration And Reference Tasks: Many tasks require referring to some text or comparing between texts. Help screens are little help when they obliterate the lines you need help with. ... Semantically linked parts of different files can be displayed together even if there is no one-to-one correspondence of lines between the files. ... Direct selection is used in reading Pascal programs. ... Multi-Screen Editors: New editors can be designed with the extra viewing space of the multi-screen environment in mind. ... The editor presents the program text in a cognitively useful manner that does not resemble the format required by the compiler. ... Cooperative Problem Solving: Another use for multiple screens is to have programmers at different sites view the common screens, with communication arranged by a network. They can talk by phone and use cursors to point at text or graphics. ... Some Human Factors Issues For Multi-Screen Workstations Working with a multi-screen terminal introduces a number of new concepts as well as questions about the best way to human factor the system. For example, with several screens how does one refer to a particular screen to direct input to appear there or to copy information from one screen to another? ... Selection using icons and menus of actions might be powerful but confusing. How do you represent a screen using an icon and where do you put it (which screen)? ... In cases were the content of the screens implies differences in operations on the screen, some way of signifying the difference is needed. ... An obvious human factors problem is the physical separation between the screens. This can be extremely disruptive in cases where one needs to compare the contents of one screen with another, line-by-line. ... Cognitive Layout in Multi-screen Workstations The added information that can be displayed on multi-screen workstations can either be beneficial or detrimental depending on the lay out. The surface layout of the information on the screens should match the user's expectation. We use the term `cognitive layout' to indicate the user's mental picture of the information. ... On the other hand, when one screen is used as a copy screen, a short term memory or notepad layout is appropriate. ... workstations allow constant display of the temporary storage file, relieving the user of the burden of remembering the contents. A hierarchical or `blow up' layout is useful when searching for information in a database or keeping track of position in a lengthy manuscript. Putting the Concepts into Action We began our work on multi-screen workstations by building a two screen system using the monochrome and color displays on an IBM PC. This system, which supports fusion, synchronized scrolling, copying and independent scrolling, was used for several experimental studies. ... Since we could not get two useful monochrome displays on one PC we decided to link two PCs together. This gives us the two monochrome screens displaying the different information that we needed at the cost of having to get the two machines communicating. ... A copy of the same program runs on each machine. The copies communicate to keep each other in step. We needed a method to send characters from one machine to the other. We designed a simple protocol to exchange single characters across the RS-232 ports. ... The protocol is for the two machines to indicate to each other that they are running by exchanging DTR and DSR signals. ... Once we had character-at-a-time communication it was rather simple to modify existing programs to work in the new environment. ... We are conducting experimental tests of team program debugging, the importance of a telephone voice link, and different protocols for controlling cursors. We find that these new environments are appealing to programmers and most users believe that these concepts are beneficial. However, there is clearly a period of accommodation to these novel approaches before the productivity benefits accrue. We have attempted to keep the commands simple and few in number, so as to ease the learning process. During 1986 we will conduct several experiments to ascertain the effectiveness of these multi-screen concepts for programming tasks and refine the designs. There are attractive opportunities for expanding the horizon of programmers. Multi-screen workstations offer the potential of showing more relevant information concurrently. Multi-screen systems are relatively cheap to build and offer provocative and novel ways to develop software. Another effort was the use of the IBM 3290 Plasma Display Workstation. ... Seeing 120 lines of program text on the screen at once leaves a dramatic impression and changes the way programmers study programs. A second prototype was for a hierarchical browser which shows the program's modular representation in the upper half of the screen. ... There is always the danger that more information acts as a distraction and that extra commands can increase confusion. Careful attention must be paid to the user interface design. Our goal is not to produce software or hardware products, but to develop ideas and validate their effectiveness as we refine our cognitive models of human performance with computers.