I still have the grope source. It starts with /************************************************** * GROPE - spelling correction - Bill Taylor - Dept 43231 * * The information contained herein is for the use of BELL LABORATORIES * and is not for publication. (See GEI 13.9-3) ***************************************************/ but I have been told by people I trust that the algorithm is really due to Tom Szymanski. I have used the algorithm for matching city and street names from the US Postal Service, and it did very well, quite a bit better than the Levenshtein algorithm, for that application. It worked well for "ordinary" dictionary matching, but would have been ghastly for things like approximate matching of DNA base pairs. Don't get me started about automating directory lookup. That was my first BTL project in the Business Information System division in the early 70s. It was obviously superior to paper records, and we had no trouble pitching up the corporate ladder until we got to the executive director level. In a blaze of insight, he asserted that microfilm had stolen the market, whereupon our director executed a light-speed 180 about supporting the project, and it died. The executive director later became Vice President of Electronic Information Systems at Western Electric. On Mon, Jan 4, 2021 at 9:13 AM Clem Cole wrote: > I also remember a program that was kicking around WH and MH called > grope(1) that IIRC used the algorithms in the code for the 411 operators. > Was that related? I do remember is that had a separate dictionary from > spell and ispell that was stored in /usr/lib/grope/[a-z]/mumble > ᐧ > > On Mon, Jan 4, 2021 at 4:08 AM Rob Pike wrote: > >> That was done by Tom Duff, I believe before he came to Bell Labs. I might >> have brought the idea with me from Toronto. The code, or at least a simple >> version of it, is in The Unix Programming Environment starting around page >> 208. We credit Tom in the endnotes for the chapter. >> >> -rob >> >> >> On Mon, Jan 4, 2021 at 8:00 PM Ed Bradford wrote: >> >>> Thank you for responding. My recollection is that one of your >>> folks put the spelling corrector into the shell so when I typed >>> the wrong letters for a directory or file, the spelling correct >>> would help. It was particularly noticible in the "chdir - cd" shell >>> command. Do you recall any such person and if so, did he (and it >>> was a he) use Peter's work? >>> >>> There was a distance algorithm that was far better than anything I've >>> seen >>> since. Yes, please send me Peter's contact information. >>> >>> I am >>> >>> Ed Bradford, Ph.D. Physics, retired from IBM >>> Pflugerville,TX >>> egbegb2@gmail.com >>> >>> PS: We chatted sometime in 1980 or so about >>> adding database capabilities to the interactive >>> environment. I was interested in adding it to >>> the Bourne Shell at the time. >>> >>> On Sun, Jan 3, 2021 at 2:23 PM M Douglas McIlroy < >>> m.douglas.mcilroy@dartmouth.edu> wrote: >>> >>>> > I was a BTL person for 8 years between 1976 and 1984. During >>>> > that time there was a spelling corrector that was better than >>>> > anything I see today. There was a concept of "spelling distance" >>>> > that corrected a whole bunch of stuff that even today cannot be > >>>> corrected. >>>> >>>> > Who in that era worked on spelling correction at BTL. I was at >>>> > Columbus BTL (1976-1979) and Whippany BTL (1979-1984). >>>> >>>> Peter Nelson made an interface to spell(1) that showed putative errors >>>> in context. I believe it could suggest corrections. I remember the project; >>>> I installed hooks for it in spell(1). I don't remember the date, but it >>>> would probably not have been early enough for you to have used it in >>>> Columbus. >>>> >>>> If there's a chance that Peter's program is the one you remember >>>> and you'd like to get in touch with him, I can give you his >>>> email address. >>>> >>>> Doug >>>> >>> >>> >>> -- >>> Advice is judged by results, not by intentions. >>> Cicero >>> >>>