00100	                         MACHINE TRANSLATION
00200	
00300		We  plan  to undertake a small effort in machine translation.
00400	The first effort will be to  create  programs  to  translate  from  a
00500	restricted  formal  language  RFL  into both English and French.  The
00600	idea is that RFL will be used to express the semantic content of  the
00700	sentences   independent   of   grammar   and  without  the  syntactic
00800	ambiguities.  There are two views  of  semantic  content  of  natural
00900	language held in the project and elsewhere, and both will be explored
01000	probably to the extent of making two translators.
01100	
01200		The first view (the linguists in the project are  betting  on
01300	this  one) is that semantic content (at least to the extent necessary
01400	for translation)  can  adequately  be  expressed  by  something  like
01500	Chomsky's deep structure.
01600	
01700		The  second  view  (held  by  AI  people  like  McCarthy  and
01800	Sandewall) is that the semantics of natural language will have to  be
01900	developed  along  lines similar to those taken in mathematical logic,
02000	i.e. the notion of denotation for phrases and  sentences  of  natural
02100	language  will  have  to  be formalized. From this point of view, the
02200	first cut at RFL should be based on the  predicate  calculus,  and  a
02300	major  effort should go into devising predicates that will enable the
02400	content of a wide class  of  sentences  of  natural  language  to  be
02500	expressed.
02600	
02700		We shall explore whether manuals written in PIMA will provide
02800	a suitable body of text.
02900	
03000		Any attempt to carry out  research  in  language  translation
03100	must  react  to  the  negative  conclusions of the ALPAK report.  Our
03200	reaction is twofold: 1. We have some new ideas (not  fully  expressed
03300	above).  2. We have adequate computer, file, and programming language
03400	facilities to get right to the problems without  major  tool-building
03500	efforts. 3. ALPAK was almost all wet anyway.
03600	
03700		From  the linguistic end, Yorick Wilks and Roger Schenk will
03800	lead the work aided 1/4 time by Dr. A. F. Parker-Rhodes and with  Dr.
03900	Margaret  Masterman as a consultant.  From the AI direction, McCarthy
04000	and perhaps Patrick Hayes will take part. Several research assistants
04100	will also be involved.