November 18, 2004


by Andrew Stern · , 11:58 am

UCSD’s Experimental Game Lab (I wasn’t aware they had one, until reading this Gamasutra blurb) has received some support from Sammy Studios, a subsidiary of Sega Japan. The support includes a $290K donation for research in MMOG technology, character animation and rendering, as well as free use by UCSD students of Sammy Studios’ proprietary game engine SCORE, for making student projects. Cool.

One of the UCSD EGL projects is a piece called What I Did Last Summer, generated by bl0gb0t with Alex Dragulescu. bl0gb0t “is a software agent in development that generates experimental graphic novels based on text harvested from web logs.”

6 Responses to “UCSD EGL”

  1. Fox Harrell Says:

    Since you mentioned UCSD…

    Yes, with the money coming in from the new California Institute of Telecommunications and Information Technology [Cal-(IT)2] more and more art and technology initiatives are being started at UCSD. The new undergraduate college is dedicated to these issues. It all seems to be in the early stages though.

    The Meaning and Computation Lab at UCSD has been devloping some work in improvisational poetry generation based upon conceputal blending theory from cognitive linguistics and algebraic semiotics theories. You may find our work interesting. This work is currently being developed into an interactive version.

    Best regards!

  2. Michael Says:

    Great to see you posting here, Fox! Thanks for including a link to your work.

  3. Fox Says:

    Thanks for the welcome. This is a very nice blog you have set up here. I hope someone finds the work interesting, in particular I offer this recent paper (from the fall AAAI symposium) on style, conceptual blending, and text generation:
    Style as a Choice of Blending Principles

    As for the UCSD EGL — I don’t know why they chose to copy the name, hopefully it was accidental!

  4. Fox Says:

    That link should be:
    Style as a Choice of Blending Principles

    Poetry generation is used as an early case study for principles for concetual metaphor generation.

    A sample poem generated with this technology is:
    The Girl with Skin of Haints and Seraphs

    her tale began when she was infected with scaled-being first-borntis
    female oppressed vapor steamed from her pores when she rode her bicycle
    death was better
    she fears only female spectres
    she loves only black ghosts
    they inspire her
    when she was no longer a child, Exu skin marks streaked her thighs
    her lips danced with love and pride.
    it was no laughing matter
    love and pride no longer concerned her when she was elderly
    her charcoal-girl soul life saddened her
    so she no longer flies with evil shame
    she only sings out that evil pride devours and alternates-with hope pride.
    Ok, I’m done with the plug. Thanks for playing.

  5. andrew Says:

    Hi Fox, in case you haven’t seen some of the discussions we’ve had on this blog about poetry generation, specifically about the Gnoetry system, here’s some links: Unconscious Thinking, and The Dublin of Dr Moreau, which also includes links to 4 previous posts about Gnoetry.

  6. Fox Says:

    Thank you for the links Andrew. I remember meeting you back at the narrative intelligence AAAI in 1999.

    I realize that I should probably explain my angle on poetry generation.

    I am not interested in Turing-competence or automated text generation.

    The artwork involved is not the individual output of one execution of the system. Rather, it is the system itself, and what it means that every poem that it generates revolves around the theme of a girl with skin of angels and demons, and comments on typically binary views of identity (black/white, male/female, etc.). The differences and consistencies are, along with the principles used to generate these, what make it interesting. Any such system is going to have to draw upon a database at some level, though the most fluid systems will have a high level of granularity. We don’t just use random selection, however.

    What I am interested in is generating narrative content on the fly, so this system is a first step toward being able to maintain a simple narrative structure, general theme, but vary content. It is also a test of using blending theory as a basis to do this. I don’t shy away from the fact that choosing a database is a part of the artist’s role in the system, I think that is great, but what principles can be used to select novel, but coherent, new content through combining elements from this database. And how can the differences in generated content add up to an overall theme (e.g. dynamic and contingent racial or other identities vs. a simplistic static and binary notion of identity).

    Probably you have all discussed some of these issues, but I thought I would come out and state my orientation to this problem outright, since it is different than “classical” poetry generation work.

    I’ll read some of your previous discussions and maybe I can add something there.

Powered by WordPress