ISFDB:RPG Cleanup

The RPG Cleanup page is used to keep track of rogue RPG (Role Playing Games) related items that have made their way into the ISFDB except RPG related fiction. Once found, they will be deleted from the database.

A search on "title contains module" comes back with a couple dozen hits, most of them RPGs. Looking for a heuristic to find and kill them all. Ahasuerus 13:53, 2 May 2006 (CDT)

• If you do find a heuristic, I'll implement it in Dissembler. It doesn't need to be simple - it can be a combination of names plus keywords. Alvonruff 05:38, 21 May 2006 (CDT)
  • Well, for starters, any publication with "Wizards of the Coast" or "TSR" in the Publisher field and "accessory"/"rule[book]"/"source[book]"/"figure"/"reference"/"supplement"/"miniature"/"figurine" in the title field is highly likely to be non-SF. Ahasuerus 16:05, 21 May 2006 (CDT)

Heuristics

I just started a Dissembler run, and noticed I didn't get the RPG heuristics in. Since you've been looking at lot of these, are the following rules correct?:

 if isRpgPublisher(publisher) and hasRpgKeyword(title):
     workIsRpg = TRUE
 elif isRpgPublisher(publisher) and isRpgAuthor(author):
     workIsRpg = TRUE
 elif isRpgAuthor(author) and hasRpgKeyword(title):
     workIsRpg = TRUE
 else:
     workIsRpg = FALSE

Publishers and Authors will have a lot of overlap since TSR and White Wolf have what amounts to RPG factories complete with "house authors". They also churn out a fair amount of fiction as by their most popular RPG module creators. The idea here is that the customers who enjoyed playing these RPG modules would be more likely to buy books bylined by the people who created their favorite modules. I would guess that 70% of the "target" Author/Publisher combinations will be non-SF, but even 20-30% of their voluminous output is quite a bit of fiction that we would be missing.

Title keywords are tricky. We really need two separate subcategories here. Some are series names, e.g. "Dark Sun", and include both RPG paraphernalia and SF novels, so they don't necessarily tell us anything up front and require human intervention to categorize. Others are clear indications that the record in question is non-SF, e.g. "accessor", "module", "rulebook", "sourcebook", etc. The latter type will give us a 98% success rate, which is pretty good. I will create two subcategories accordingly shortly. Ahasuerus 08:50, 30 May 2006 (CDT)

Done. At this point if isRpgPublisher(publisher) and hasRpgKeyword(title) should have a very high success rate. Hopefully :) Ahasuerus 08:59, 30 May 2006 (CDT)

What about if isRpgAuthor(author) and hasRpgKeyword(title)? (Just trying to find a way to use the author list...) Alvonruff 09:03, 30 May 2006 (CDT)

Hm, that may work too, but there is a caveat. Given that Authors can (and often do) do non-RPG work when working for non-RPG publishers, there will be cases when an otherwise "RPG author" will write "Jim's Adventures in the Solar System", an SF title, that we will miss because of the keyword "system". I think it's somewhat safer to go by the publisher/keyword, but there is no harm in trying both approaches and checking what kind of overlap we will get. Ahasuerus 09:44, 30 May 2006 (CDT)

Targeted Publishers

Note: Substring search on the Publisher field withi Publications is currently broken.

• Fast Forward Entertainment
• Green Ronin Publishing
• Mongoose Publishing
• TSR
• White Wolf
• Wizards of the Coast

Series names (may include works of fiction)

• dark sun
• everquest
• forgotten realms
• ravenloft
• spelljammer

Targeted Keywords

The following is a list of RPG-related keywords that suggest (but don't guarantee) that a Work with this keyword in its title may be an RPG-related item and not a work of fiction. Once you have confirmed that it is indeed an RPG-related item, make sure to follow its Author(s)' links to find and kill other Publications first.

• accessory
• accessories
• advanced
• booster
• campaign
• d20
• d