CSC 594 Home

Lecture Notes
& Schedule

Projects

Assignments

Syllabus

Resources
& Links

 

CDM COL

DePaul academic 
calendar

 

CSC 594 Topics in AI: Applied Natural Language Processing
Fall 2009/2010

(d) Video Game Reviews

[draft]
The goal of this project is to develop a search/retrieval system for video games.  If time allows at the end of the quarter, we will try a recommender system.

We envision our video game retrieval system to be something that allows the user to search video games not only by the typical categories (such as title, publisher, genre, platform, release date, popularity, user rating, ESRB rating), but also by the characteristics of the game content, such as the kind of gameplay experience (e.g. exhilarating feeling, ultra-fast action, mesmerizing sensation), graphics, sound and game design.

We have already crawled a game review site called Gamespot (http://www.gamespot.com), and extracted all reviews (critics' as well as users'; also the basic information of a game as listed above) in a text format. We will use this data as the data resource for our system.
Here is an example of "Metal Gear Solid 4: Guns of the Patriots" (a PS3 game by Konami).

Research work may include tasks and discussions as follows:

  • Previous work -- Survey of the current video game retrieval or recommender systems.  Is there any that used the characteristics of the game content?
  • Look at movie retrieval or recommender systems -- How would a video game retrieval/recommender system be different?  Also other domains 'similar' in the content/structure as video games (e.g. hotel reviews, restaurant reviews)?
  • What kind of "gameplay" experiences are there?  How are gameplays 'described' to begin with (especially by adjectives/adverbs)?
  • Are graphics/sound/game-design important as search criteria?  Are there any other good search criteria which the users would care about?
  • How could we "get at" the parts in a game review which mention those aspects?
    Also we must determine which NLP "pre-processing" tasks are needed for the data -- Part-of-speech tagging? Named entity recognition?  If so, are there any data resources such as "all game titles" and "the names of the characters in video games"?
  • Design and implement an interface of the search/retrieval system.

Since the gameplay experience seems to involve adjectives and adverbs, we may utilize the "Google n-gram data" which I have.