Brought to you by EarthWeb
IT Library Logo

Click Here!
Click Here!

Search the site:
 
EXPERT SEARCH -----
Programming Languages
Databases
Security
Web Services
Network Services
Middleware
Components
Operating Systems
User Interfaces
Groupware & Collaboration
Content Management
Productivity Applications
Hardware
Fun & Games

EarthWeb Direct EarthWeb Direct Fatbrain Auctions Support Source Answers

EarthWeb sites
Crossnodes
Datamation
Developer.com
DICE
EarthWeb.com
EarthWeb Direct
ERP Hub
Gamelan
GoCertify.com
HTMLGoodies
Intranet Journal
IT Knowledge
IT Library
JavaGoodies
JARS
JavaScripts.com
open source IT
RoadCoders
Y2K Info

Previous Table of Contents Next


Even if the type of internal representation adopted in LUNAR was supposed to favor the independence between the database and the query modules, the design of the last one was, to some extent, specific to the project's needs, and portability was not assured.

SHRDLU was a computer program that carried on a dialog via teletype with a user: in this dialog, the system simulated the behaviour of a robot's arm in a tabletop world of toy objects (blocks, pyramids, spheres, etc.). SHRDLU could answer questions about the state of the world ("Is the red block clear?"), carry out commands ("Put the red pyramid on the red block."), and add new facts about the world ("The pyramid is on the table."). The simulated world was displayed on a CRT, and it was possible, therefore, to follow the activities carried out by SHRDLU when moving the objects around. The system was composed, basically, of three modules: for syntax, semantics (to convert a sentence into a sequence of commands for the arm or into a query of the system's database), and problem-solving. The modules were physically separated, but they were able to interact all the time: this was particularly new for that period, when all the systems were fundamentally sequential systems. Thanks to the naturalness of the dialog, and to its apparent reasoning ability, SHRDLU has been one of the most popular AI programs ever built up, and it is still one of the most frequently mentioned, especially in introductory texts on AI.

We can mention here briefly some important technical features of SHRDLU:

  1. In this system, the procedural approach was systematically followed, and realized by using the deductive properties of the MICRO-PLANNER programming language. For example, instead of representing a linguistic notion saying that a sentence is composed of a noun phrase and a verb phrase under the form of a rule in a grammar, "S —> NP VP," see subsection 3.4.1, this notion was directly represented as a MICRO-PLANNER procedure that called other procedures under the form of (PARSE NP) and (PARSE VP).
  2. Coherently with the above, SHRDLU was characterized by an "imperative" conception of NL, derived from the "speech act" theories. According to this, the meaning of an utterance is not represented declaratively as a fact about the world, but as a command addressed to the system in order to do something. For example, a simple assertion like "the pyramid is on the table," is translated as a command (a program) for adding information to the database.
  3. The problem-solver component of the system (i.e., the modules that know about how to accomplish tasks in the block world) contained some form of explicit representation of the cognitive context. For example, SHRDLU has embedded the notion of "focus" (see, below, Section 3.4.4.2) that, in front of an assertion relating to "the red block," is able to link this assertion with one of all the possible red blocks of the world, the block which is more in focus of the others because of having been mentioned or acted upon recently.

SHRDLU represents a significant step forward in NLP research because of its successful attempt to implement both a "serious" linguistic analysis and some "realistic" reasoning methods. The system had, of course, many limitations, mainly due to the fact of dealing with a simple, logical, and closed domain: this allowed one, e.g., to avoid handling many of the more complex features of English.

Thanks to the successes, real or (at least partly) emphasized, of these first AI query-answering systems, the domain of NL interfaces to database became very popular from the late 1970s onward. A complete overview of the domain is beyond, of course, the possibilities of this section, and we will limit ourselves to indicate here some systems that have been mentioned intensively in the literature; see Androutsopoulos et al. (1995) for an exhaustive bibliography.

In the late 1970s LADDER, by Hendrix and colleagues, and PLANES, by Waltz, are examples of systems based on the "semantic grammars" approach, see subsection 3.5.2.1: semantic grammars are linguistic tools, very effective from an engineering point of view, where the domain knowledge is hard-wired into the grammar. In the early 1980s, some systems were developed that, in the absence of the required data, tried to avoid the ambiguous "silence" of traditional databases by informing the user that his assumptions about the contents and the structure of the base might be wrong. The most well-known example of this type of systems is CO-OP, by Kaplan. CHAT-80, by Pereira and Warren, is an example of the PROLOG system, where the English questions were translated into PROLOG expressions, and then evaluated against a PROLOG database. In the mid-1980s, TEAM, by Grosz and colleagues, was designed with the aim of being easily configurable by database administrators with no knowledge of NLP. ASK, by Thomson and Thomson, and, in the late-1980s, JANUS, by Bobrow, Resnick, and Weischedel, were characterized, inter alia, by the presence of an NL interface capable of interacting with multiple external databases, electronic mail programs, and other computer applications. Examples of contemporary, commercial NL interface systems are INTELLECT (Trinzic), Q&A (Symantec), and NATURAL LANGUAGE (Natural Language Inc.). We can remark here that, as it was already the case with SHRLDU, etc., current interfaces to databases can only cope with limited subsets of NL, and this can explain why they have not reached wide acceptance on the commercial market, where other alternatives (e.g., graphical and form-based interfaces) are still preferred.


Previous Table of Contents Next

footer nav
Use of this site is subject certain Terms & Conditions.
Copyright (c) 1996-1999 EarthWeb, Inc.. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of EarthWeb is prohibited. Please read our privacy policy for details.