home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!snorkelwacker.mit.edu!ai-lab!news.ai!ilh
- From: ilh@lcs.mit.edu (Lee Hetherington)
- Newsgroups: comp.text.tex
- Subject: Re: IBM Pennant Systems survey
- Message-ID: <ILH.92Aug28162140@winnie-the-pooh.lcs.mit.edu>
- Date: 28 Aug 92 20:21:40 GMT
- References: <199208281156.AA08401@claude.cs.umb.edu>
- <l9sn7jINNnrm@utkcs2.cs.utk.edu>
- Sender: news@ai.mit.edu
- Reply-To: ilh@lcs.mit.edu
- Organization: MIT/LCS Spoken Language Systems
- Lines: 15
- In-reply-to: eijkhout@cupid.cs.utk.edu's message of 28 Aug 92 17:05:55 GMT
-
- I'd guess that they're going to use the documents to build statistical
- language models, probably for speech recognition purposes. I know
- that they've collected literally millions of words of internal memos
- and email for just that purpose. Basically, they estimate the
- probability of a word given the surrounding context and need millions
- of words to get reliable estimates for still longer strings of words.
-
- Why they don't tell you, I don't know. This is just my educated
- guess. I have nothing to do with IBM.
-
-
- --
-
- Lee Hetherington
- ilh@lcs.mit.edu
-