home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!haven.umd.edu!news.umbc.edu!nobody
- From: finin@cs.umbc.edu (Timothy Finin)
- Newsgroups: comp.ai
- Subject: Re: New Problems in IJCAI Reviewing (long)
- Date: 21 Jan 1993 13:27:17 -0500
- Organization: Computer Science, University of Maryland Baltimore County
- Lines: 26
- Message-ID: <1jmpu5INN51l@cujo.cs.umbc.edu>
- NNTP-Posting-Host: cujo.cs.umbc.edu
- Summary: Let a machine do it.
- Keywords: shoemaker, IRS, human frailties, machine perfection
-
- Perhaps we are like the shoemaker who does not provide good footwear for his
- children? We clearly need to develop on automated, impartial refereeing
- agent to use in future IJCAI conferences. Authors who would like to submit
- papers to IJCAI would send a version in an agreed upon format (e.g., ASCII,
- Postscript or SGML) to the IJCAI Review Server (e.g.IRS-95@AI.MIT.EDU ).
- This system will render an immediate and unbiased decision without relying
- on the necessarily prejudiced judgement of humans.
-
- Recent work in statistical techniques for natural language processing and
- information retrieval look relevant to building such mechanical reviewing
- agents. If we can collect a large training set of past papers, both those
- accepted to and rejected from a conference, we can build a classification
- system to use on future instances of that conference. Figures, especially
- those containing images, may pose a more significant research challenge, but
- there is no principled reason why these could not be handled as well.
-
- There would be many benefits which would result from such a rapid and
- impartial reviewing process. For example, a researcher could "fine tune" a
- paper with respect to a particular conference through a process of
- submission, and modification and resubmission until the server accepted the
- paper. Alternatively, an author could submit a paper to the IJCAI server
- and, if he receives an immediate and negative reply, could subsequently
- submit the paper to the servers for AAAI, CAIA, AISB, CSCSI, etc. until a
- server is found which accepts the paper.
-
- Tim
-