Текст наукової роботи на тему «An intelligent os-agent interface»
?9. Gao T., Vaidya P.M., Liu C.L., A new performance-driven placement algorithm, Proc. of ICCAD, pp. 44-47, 1991.
10. Gao T., Vaidya P.M., Liu C.L., A performance-driven-macro-cell
placement algorithm, Proc. of ICCAD, pp. 44-47, 1991.
11. Tetelbaum A.Y., Wey C.L., Bickart T.A., A cell-level-based
performance-driven standard-cell placement, Technical report, Department of
Electrical Engineering, Michigan State Univenlty, June, 1993.
AN INTELLIGENT OS-AGENT INTERFACE
Paper describes designing and testing an intelligent-agent man-machine interface based on a mixture of a syntax-based approach, a memory-based approach [4,7] and an approach based on intelligent agents [1,5,6]. Communication between a human user and an operating system VAX / VMS was chosen for a test domain enabling testing of software agent technotogy .
The paper is organised as follows: Related approaches are discussed in Section 2, an intelligent operating interface IOI in Section 3, tests in Section 4, followed by the concludind discussion in Section 5.
2. RELATED APPROACHES
According to , traditional approaches to translation often use explicit rules of knowledge and are guided by rigid control rules, i.e., the syntax of programming languages. However, for domains such as translation between two natural languages it is almost impossible to obtain a complete set of rules for a given problem. Memory-based approach builds on memory as the foundation of intelligence. It is assumed that large numbers of specific events are stored in memory. New situations are first handled by recalling and comparing with previous events, and then performing similar or modified reactions.
According to , intelligent agents represent personal assistants collaborating with the user in the same environment. Agents and humans both initiate communications, monitor events and perform tasks. The assisting agent learns and modifies according to the user's interests, habits and preferences. There is a rich set of emerging views and new terms such as knowbots, knobots, softbots, userbots, taskbots, personal agents etc.
The approach by Maes is based on self-programmable agents that leam by • obsenring the user
|receiving positive and negative feedback from the user 'receiving explicit instructions from the user |experience from the environment.
The above directions were guiding also the design of the IOI system. In addition, self-programmable agents often rely on memory-based approaches, as does IOI.
3. AN INTELLIGENT OS INTERFACE IOI
We have designed and implemented an Intelligent Operating Interface (IOI) for VAX / VMS  based on the above and two additional purposes:
|to implement a rule-based and a memory-based approach and make a comparison
|to design and test an adaptable interface enabling small improvements such as correcting typing errors and providing help about approximately correct comands.
The classical rule-based approach consist of designing syntax diagrams (trees) and choosing specific keywords. When searching for a specific slightly wrong command sequence, basic commands (roots of trees) are examined first. If a sufficient match is found, the search continues one level below enabling several viewpoints. Learning in the rule-based approach in IOI is performed as adding or deleting words from specific boxes. Basic rules which correspond to the tree structure (i.e. syntax diagrams) are coded or corrected by direct commands from the user.
In the memory-based approach aH.command inputs are stored as string in a file. The program searches through all strings and sorts matching strings according to two match parameters: frequency and quality. Quality is calculated by parsing two strings and calculating the number of identical chars at identical positions, identical chare in nearby positions, the rest of the chars, and attaching specific weights. Frequency of each stored command is updated each time it is encountered by a user or OS. Besides basic commands, transformations are also stored in another table, e.g., COPPY -> COPY 4 meaning that if COPPY is in the command input, it will be automatically transformed into COPY by IOI. The number of appearances (frequency) of COPY so far is 4.
Learning is implemented as storing new words in syntactic trees in a rule-based approach, and new strings into corresponding tables in the memory-based approach. Besides storing correct commands, wrong commands are stored as patterns for future search, and their corresponding cured versions for transformations.
Currently, all versions of IOI remember commands the first litre no error is reported, and wTong and corrected commands after one confirmation by the user. Due to mistakes and garbage, commands are forgotten if they are too rarely used. In addition, a user can directly edit all tables in IOI.
An example of a learning session with a "tabula-rasa" IOI is bellow:
% DCL-W-IVKEYW, unrecognized keyword - check
validity and spelling / DEFF /
$$ SHOW DEF DU AO: | USER.WORK |
Device Device Error Volume Free Trans Mat
Name Status Count Label Blocks Count Cnt
FIBRESDUAO: Mounted 0 UD1 597588 26 1
S5SHOW DEFF SHOW DEF (1)
SHOW DEV (2)
Accept? | 1 |, 2, No, Abort, Edit: <Enter>
SHOW DEFF -> SHOW DEF PROCESS: SHOW DEF DIJAO: [USER.WORK |
At the beginning, IOI does not have any knowledge of the SHOW DEFF command string and submits it to VAX / VMS. Since VAX / VMS does not recognise it as a correct command, IOI ignores it. Then IOI remembers two correct commands SHOW DEF and SHOW DEV. When the first command SHOW DEFF is repeated IOI proposes two close matches. A user chooses (I) and from there on IOI knows that SHOW DEFF means SHOW DEF. In future appearances IOI automatically transforms SHOW DEFF into SHOW DEF.
The above version of IOI did not have any a priory knowledge about any VAX / VMS command. Learning was performed by the memory-based part. The rule-based part is not able to start learning from the scratch. If in any form knowledge about SHOW DEF and SHOW DEV already existed before the first command, IOI would provide two or more close solutions in the first reply.
Learning in IOI was studied with different amounts of a priory knowledge: a) with no a priory knowledge, b) partial knowledge, and c) all possible knowledge about OS and the user (meaning the LOGIN.COM file was given to IOI in advance ). Tests with the last version are presented in the not Section.
IOI is capable of learning and forgetting. The main motivation for forgetting is cleaning the garbage which inevitably and constantly appears in human communication. For example, if a user wrongly types a command of an application program that reports an error directly to the user without informing VAX / VMS, IOI accept the command as correct. Since such commands are remembered as correct they can be later offered as possible solutions. Unwanted commands can be discarded by direct commands from the user, however, it is much better that IOI cleans them itself. Forgetting is implemented as deleting commands with the lowest relative frequency.
IOI is implemented as a 2000-line program in Pascal with parts of it written in the VAX / XMS command language. Time delays due to IOI are meaningless. The tested users used IOI and filled the ISO evaluation forms.
The most positive and negative aspects of IOI are analysed. According to the tested users, a) IOI is easy to use, b) it does not demand specific knowledge, c) it is easy to leam and use. The tested users were also reasonably satisfied with d) transparency of IOI's current activities.
The four most unfavourable properties of IOI as evaluated by the testing users are: a) around average guide for further work, b) no HELP about IOI, c) bad error messages of IOI itself, and d) no adjusting of the screen.
IOI has been successful in showing that a memory-based intelligent-agents approach ofTers substantial advantages in a shorter programming time, greater flexibility and adaptability to a specific user and a programming system. Man-machine communication can be improved in the sence of greater flexibility and understanding especially in mundane tasks. The unanswered question is if human users are ready to accept current "Intelligent" systems especially in difficult and responsible tasks.
1. AIM-15-3 (1994), Consenting Agents, AI magazine, 15, 3.
2. Etzioni, O. (1993), Intelligence without Robots: A Reply to Brooks, AI Magazine, pp. 7-13.
3. Hribovsek, B. (1994), Intelligent Interface for VAX / VMS, M.Sc.Thesis.
4. Kitano, H. (1993), Challenges of Massive Parallelism, Proc. of the 13th Int. Joint Conf. on Art. Int., Pp. 813-834.
5. Maes, P. (1994), Agent that Reduce Work and Information Overload, Communications of the ACM, 37, pp. 31-40.
6. Negroponte, N. (1970), The Architecture Machine; Towards a more Human Environment, MIT Press.
7. Stanfil, C. and Walts, D. (1988), The Memory-Based Reasoning Paradigm, Communications of the ACM.
Завантажити оригінал статті: