Metadata is also a way to detect/correct/mitigate algorithmic biases...! 

Bohyun 

-----Original Message-----
From: Code for Libraries <[log in to unmask]> on behalf of "Xiao, Jingshan" <[log in to unmask]>
Reply-To: Code for Libraries <[log in to unmask]>
Date: Tuesday, December 11, 2018 at 12:01 PM
To: <[log in to unmask]>
Subject: Re: [CODE4LIB] [External - Possible Spam: 11%] Re: [CODE4LIB] ai in libraries

    Good information and discussion.  I agree with Tim.  Metadata is indispensable at the stage of machine learning in library and information science disciplines. 
    
    Thanks,
    
    Jingshan
    
    -----Original Message-----
    From: Code for Libraries <[log in to unmask]> On Behalf Of Tim Thompson
    Sent: Tuesday, December 11, 2018 10:50 AM
    To: [log in to unmask]
    Subject: [External - Possible Spam: 11%] Re: [CODE4LIB] ai in libraries
    
    Thank you, Eric, for sharing this. At least for supervised machine learning algorithms, a training set is needed--and the labels applied to it are certainly a form of metadata. In fact, I think it makes sense for libraries to start devoting staff time to developing robust training sets to support AI-driven data processing and analysis workflows. This seems like an area in which Technical Services units could contribute.
    
    Also, if metadata is not necessary, why would we use machine learning to create and manage it? :)
    
    Best regards,
    Tim
    
    --
    Tim A. Thompson
    Discovery Metadata Librarian
    Yale University Library
    
    On Mon, Dec 10, 2018 at 4:53 PM Eric Lease Morgan <[log in to unmask]> wrote:
    
    > Last week I attended an artificial intelligence (AI) in libraries 
    > conference, and I've written the briefest of travelogues. [1] Some of 
    > my take-aways include:
    >
    >   1. Machine learning is simply the latest incarnation of AI, and
    >      machine learning algorithms are only as unbiased as the data used
    >      to create them. Be forewarned.
    >
    >   2. We can do this. We have the technology.
    >
    >   3. There is too much content to process, and AI in libraries can
    >      used to do some of the more mechanical tasks. The creation and
    >      maintenance of metadata is a good example. But again, be
    >      forewarned. We were told this same thing with the advent of word
    >      processors, and in the end, we didn’t go home early because we
    >      got our work done. Instead we output more letters.
    >
    >   4. Metadata is not necessary. Well, that was sort of a debate,
    >      and (more or less) deemed untrue.
    >
    > If you want to participate in AI for libraries-like discussions, then 
    > consider subscribing to ai4lib.
    >
    > [1] travelogue - 
    > https://sites.nd.edu/emorgan/2018/12/fantastic-futures/
    >
    > [2] ai4lib – https://groups.google.com/forum/#!forum/ai4lib
    >
    > --
    > Eric Lease Morgan
    > University of Notre Dame
    >