General Motors Research Laboratories

and high level sources of knowledge (concepts). The function of the parser is to produce a segmentation f the sentences in a case frame structure, thus determining the meaning of prepositions, polysemous verbs, of the noun group etc. But the function of this parser is not to produce an output to be interpreted by semantic routines, but to start the parsing process and to proceed until a concept relevant to the theme of the text is recognized. Then the concept (a cluster of production rules) takes control of the parsing process overriding the low level linguistic processes. The concept continues supervising and guiding the parser until the sentence has been understood, that is, the meaning of the sentence has been mapped into the final internal representation. Thus a text is parsed directly into the final knowledge structures. Results The area that I have chosen to test the ideas expressed above has been the understanding of programming problems tated in natural language. A program, called LLULL, has been implemented following the ideas summarized above. The program has understood ten programming problems taken verbatim from introductory books to programming. It runned in the DEC20/20 of the Dept. of Computer Science (DSU) under UCI Lisp. Now, the necessary changes are being introduced to make it to run in a VAXll/780 under Franz Lisp. Future Plans I intend to continue the work at both the theoretical and experimental level. Immediate plans are to extend LLULL so that it will have capability to understand a large class of programming problems. Next, I will extend my ideas to other domains. The final theoretical goal is the development ofa theory that explains the relation between natural language understanding and problem solving. The reference below contains a summary of the basic ideas as well as the background literature