![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
Femslash Week In Review feature request poll
Open to: Registered Users, detailed results viewable to: All, participants: 10
High-level, what kinds of changes are the most interesting to you?
More accurate and comprehensive data about F/F fic on AO3
1 (10.0%)
More/better analysis and presentation of the data and the relevant trends and features
5 (50.0%)
More/better recs of specific works to check out this week
4 (40.0%)
Data accuracy is my bugbear, because I feel like I am telling you lies whenever I have known bugs in my dataset. Here's the current list of things I'm tracking that I know are wrong or broken. Which of these do you think are important to fix (pick as many as you want --- if everybody picks all of them, I'll take it as justification to keep obsessing about data inaccuracy bugs):
The dataset is missing many works from revealed exchanges (e.g. Yuletide, Chocolate Box) because of the way "last updated" times work
9 (90.0%)
Fandoms where there are multiple approximately-equally-popular tags for the same set of ships are not handled cleanly (e.g. "Doctor Who" vs "Doctor Who & Related Fandoms"), making changes to the top 20 hard to understand
2 (20.0%)
I don't audit character tags for duplicates, so many (anecdotally) works which should be in the same ship are misclassified as being different ships, making ships appear smaller and more numerous than they actually are
1 (10.0%)
In the process of my ship audit, I've found a number of ship tag problems on AO3, and haven't prioritized getting those back to the tag wranglers so they can be fixed on actual AO3 (anyone know a tag wrangler who wants to work with me on these?)
3 (30.0%)
Analysis and presentation is where I feel like the stats have been falling down on the job. In part, thinking up interesting metrics is challenging; in part, there's a certain amount of work to mechanically crafting graphs, and any time I want to do something that changes how the data visually looks, I have to make it work in both Tumblr and Dreamwidth HTML; in part, defining a new metric almost always entails discovering a raft of new data accuracy bugs that make the metric not work as well as I want, which takes me back to focusing on data accuracy. That said, presenting data should be why this project is here. Which of these would most interest you as a reader (again, pick as many as you want, and please do use the comments to mention specific things you'd like to see, if you have interest in that):
Graphs of additional metrics for the top 20 fandoms (using stuff that's available on AO3 like works by ship, comments/kudos/hits, percentage of WIP works, etc)
3 (30.0%)
Regularly displaying more than one graph per week (splitting this out from the above because of the technical/integration work needed to make it feasible with both Tumblr and Dreamwidth as noted)
1 (10.0%)
Improve the "top 20 fandoms" table (change how existing columns are displayed or what data they contain, add new columns, e.g. replace "rating" with something else; replace "F/F:M/M ratio" with "percent of top 10 F/F ships which are actually F/F", etc)
4 (40.0%)
Increasing the table from 20 fandoms to a larger number like 50 or 100
3 (30.0%)
Adding graphs or tables for "top ships" in addition to the existing top fandoms
3 (30.0%)
Picking a fandom of the week and making some detailed graphs about the historical/current-week breakdown of works within that fandom (e.g. top ships, fandom activity over time)
9 (90.0%)
Writing more intelligent words about trends visible in the data, possibly including doing some side research week-by-week to explain what we're looking at, or looking or comparing my conclusions to those of other stats posts (which ones?)
7 (70.0%)
Recs: I know from anecdotal evidence that at least 1-2 of my 6 regular readers think the recs are the most interesting part of the post. I find recs difficult, because I feel like my own preferences are my own problem. Say I categorically don't like trope X. If I read a story with trope X, then either (a) I don't like the story, and other people might like it, but I can't tell because I didn't, or (b) I like the story, and I could say "this is such a great story that you'll enjoy it even if you don't like trope X," but that's implicitly obnoxious to lovers of trope X, for no good reason. Also, I don't want readers to get totally bored with my favorite fandoms and ships, so I at least try to be significantly pickier about things I know are in my personal wheelhouse (maybe after a single-week "new hotness" grace period). So I tend both to say as little as possible about the things I rec, and to rec significantly fewer things than I read. How would you like to see this be different?
More recs: list a higher percentage of stuff I read every week
1 (10.0%)
More metadata: say more works about why I'm reccing particular things (what kind of information is helpful here?)
3 (30.0%)
Better classification: make up a system for what kinds of recs things are, e.g. "new hotness", "Letzan's personal favorite ships", "cool thought experiment", and use those for all recs so you can gravitate towards areas where you agree with me
5 (50.0%)
Start recommending WIPs (I don't currently)
0 (0.0%)
Start recommending older F/F works which I happen to read during the current week for unrelated reasons (e.g. I read a book and go find four great three-year-old fics about that book. Do you want to see those in the recs?)
8 (80.0%)
Try to recommend more stuff from fandoms you personally like (tell me which ones!)
0 (0.0%)
More recs from fandoms which are currently very active/on the rise
0 (0.0%)
Fewer recs from fandoms which are currently very active/on the rise
1 (10.0%)
More recs from and/or information about exchanges
4 (40.0%)
Anything else at all you thought of while reading this, let me know here or in the comments. Thanks again for following the week in review enough to have encountered this post, and for reading this post!
no subject
My main issue is that I'm not in the main F/F fandoms, so recs from them aren't that useful to me. Mostly I'm interested in seeing trends and odd things that stick out.
no subject
Re your first comment: cool!!! So, the thing I have right now this second is a list of ship tag names and URLs for which:
* The two ship tags contain an identical set of parent character tags
* Both tags are what I think is called "canonical" (neither one has been merged with the other, or with anything else)
Here's an example:
* Gilbert Blythe & Anne Shirley https://archiveofourown.org/tags/Gilbert%20Blythe%20&%20Anne%20Shirley
* Anne Shirley & Gilbert Blythe https://archiveofourown.org/tags/Anne%20Shirley%20&%20Gilbert%20Blythe
As far as I, a non-wrangler, know, this is always a bug (unless one tag is a "subtag" of the other? But most of them aren't, and I can check for this if it's likely to be common) --- one tag should be merged with the other.
Any thoughts on the most useful thing to do with this data?
no subject
Yeah, there shouldn't be two canonicals for the same relationship (and the current guideline is alphabetic order by last name, so "Gilbert Blythe & Anne Shirley" is 'correct' and the other should be synned to it. (Anne of Green Gables looks like a pretty complex fandom because there are so many versions, books, TV shows, and movies, so it's unsurprising things fall through the cracks.) If you give me a list of similar pairs of tags (with links, like you so kindly did before!) I can comment on the 'wrong' tags to give the wranglers for that fandom a heads-up that a user is having problems compiling statistics because of the duplicate tags. I can't fix them myself (unless they're for fandoms I wrangle).
You can email me at isiscolo@gmail.com or contact me through a Dreamwidth DM.
no subject
(I'm not actually having problems compiling statistics --- because of the obviousness of the issue, this is something where I can and have scripted cleaning up the duplicates in my own dataset --- so don't guilt the wranglers into any heroic effort on my behalf. But other tag-scraping users, as well as people searching on the actual website, might have the same problem.)