Human Engineering for Applied Natural Language Processing

Human engineering features for enhancing the usabil ity of practical natural language systems a l re described. Such features include spelling correction, processing of incomplete (ell ipt ic-~I) input?, jntfrrog-t ior of th p underlying language definition through English oueries, and ?r rbil.it y for casual users to extrnd the language accepted by the system through the-use of synonyms ana peraphrases. All of 1 h* features described are incorporated in LJFER,-"n r ppl ieat ions-orj e nlf d system for 1 creating natural language j nterfaees between computer programs and casual USERS LJFER's methods for r<"v] izir? the mroe complex human enginering features ? re presented. 1 INTRODUCTION This pape r depcribes aspect r of a n applieations-oriented system for creating natural langruage interfaces between computer software and Casual users. Like the underlying researen itself, the paper is focused on the human engineering involved in designing practical rnd comfortable interfaces. This focus has lead to the investigation of some generally neglected facets of language processing, including the processing of Ireomplfte inputs, the ability to resume parsing after recovering from spelling errors and the ability for naive users to input English stat.emert s at run time that, extend and person-lize the language accepted by the system. The implementation of these features in a convenient package and their integration with other human engineering features are discussed. There has been mounting evidence that the current state of the art in natural language processing, although still relatively primitive, is sufficient for dealing with some very real problems. For example, Brown and Burton (1975) have developed a usable system for computer assisted instruction, and a number of language systems have been developed for interfacing to data bases, including the REL system developed by Thompson and Thompson (1975), the LUNAR system of Woods et al. (1972), and the PLANES system ol Walt7 (1975). The SIGART newsletter for February, 1977, contains a collection cf 5? short overviews of research efforts in the general area of natural language interfaces. Tnere has rise been a growing demand for application systems. At SRi's Artificial Irtellugene Center alone, many programs are ripe for the addition of language capabilities, Including systems for data base accessing, industrial automation, automatic programming, deduct ior, and judgmental reasoning. The appeal cf these systems to builders ana users .-'like is greatly enhanced when they are able to accept natural language inputs. B. The LIFER SYSTEM To add …