![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
toujours gai
i have been playing with the gender genie
which is at
http colon //www.bookblog.net/gender/genie.html
which i gacked from notapipe
and it gets my gender right most of the time but interestingly
if i put in stories i ve written it gives the reverse of the narrator s gender
but the most fun is that it makes your writing
look like archy s of
archy and mehitabel fame
so i must write that way too now
you can tell i m procrastinating from writing my paper
also i just realized that fiendslist filters work
even if you aren t a paying member
the things you learn
i will stop now before this becomes too annoying
strange when archy writes this way it s literature
but when other people do it s
netspeak
oh well toujours gai my friends toujours gai
this can t be worse than the dierks bentley songfic
which was the other alternative anyway
which is at
http colon //www.bookblog.net/gender/genie.html
which i gacked from notapipe
and it gets my gender right most of the time but interestingly
if i put in stories i ve written it gives the reverse of the narrator s gender
but the most fun is that it makes your writing
look like archy s of
archy and mehitabel fame
so i must write that way too now
you can tell i m procrastinating from writing my paper
also i just realized that fiendslist filters work
even if you aren t a paying member
the things you learn
i will stop now before this becomes too annoying
strange when archy writes this way it s literature
but when other people do it s
netspeak
oh well toujours gai my friends toujours gai
this can t be worse than the dierks bentley songfic
which was the other alternative anyway
no subject
--C
no subject
i shall have to fix that
sorry
but i think the site just went down anyway
no subject
--C
no subject
putting up fences are you
i thought i was the specialist at that
it is fixed i can t get anything right today it seems
also i should be at school half an hour ago
wotthehell
autobogotification
It was wrong on 3 out of 4 of my journal entries that I tried.
According to the reports on bookblog, it's doing worse than random guessing right now. Sad llama.
I'd like to read Koppel's original article since this is right up my natural-language processing / machine-learning/classification alley. In particular the nature article doesn't say how they came up with those weights or words (I'm guessing they pulled the words out of a hat and then did some kind of training to assign the weights) - I wonder if they did something stupid like test their model on the same or substantially similar texts (e.g. by the same authors) as they used to train it, thereby producing something that only works on texts by a particular author, which (as I understand) are fairly easy to classify anyway.
It'd be interesting to make a simple unigram or bigram model and see how well that works - i.e. take a bunch of text written by men and a bunch written by women, look at the frequency of individual words (or pairs of words) and then based on that information predict the probability of some other unrelated text being written by a man or a woman (you know approximately how likely each word or pair of words is in a text, so just multiply those probabilities together.. your guess at the end is the one with the higher probability) - I could probably hack this together in a few minutes but I'm likely not interested enough to do so - It may be that what they're doing is based on a unigram model like this, but it seems overly simplistic to me.
Unfortunately College Park doesn't seem to get Literary and Linguistic Computing, so I have no way of seeing if what they did is totally bogus or not.
btw, writing in the style of said archy is exquisitely annoying. keep it up if you want to get thwapped by my new catfish;-)
Re: autobogotification
you know
but i think it s probably less accurate on lj entries
since it weights you and we and such as female
which are far more likely to occur in lj than print
net writing in general has different style standards than outside
which could account for the putrid accuracy score
i m sure they could get you the article on loan
i did wonder why it seemed to take such a major effort
since i can think of several simple implementations
and i know very little
also why two hebrew professors did their study on english
and why they didn t start with a wider spectrum of samples
mehitabel is very fond of fish.
no subject
Re: autobogotification
Re: autobogotification
i am standing in the seventh floor of mckeldin
that is the joy of the internet
where did you think i was
anyway i would be on my way to class now if someone wasn t commenting
in my journal
so there
no subject
but i was having more fun with stories
so i didn t try too many
makes sense with the idea that less chatty entries are more likely male
the stories were backward fairly consistantly
don t know what that says about my characterization
Re: autobogotification
did i really explain it that badly? if i had a whiteboard i could probably explain it to you in about 5 minutes, you're pretty smart. i wouldn't be too surprised if your simple implementations aren't too different from what i suggested.
what you said about lj writing did cross my mind, though one would hope that a real classifier would be more robust - it's a /little/ more complicated than it initially appears, since although you get positive and negative points for things, NOT using those things biases you the other way - so if you use an alternative to "it" (e.g. any other pronoun including you or I, since those are perfectly good things to talk about instead of "it"s") I guess the alternative to using determiners is using more proper nouns (and mass nouns like 'snow', 'water', 'ice' that can be used without a determiner) and possessives - women talk more about things with respect to other things and men just talk about things? (gee, that's as comprehensible as pterry's quote about things that try to be like things being more like things than the real things) Likewise I'm assuming that the inclusion of 'with' or 'for' means they expect most prepositional phrases headed by 'with' and 'for' to have a person as the noun? I guess "Bears like eating raw, bloody meat with their bare hands" is pretty feminine despite not talking about any people. (Most any classification scheme will be easy enough to foil with contrived examples, though) - I don't know what the significance of 'not' and 'n't' is, though, or what the alternative to using a negation is..
(time to go home, I wasted enough time thinking about this and then changing my mind several times)
Re: autobogotification
it used some technical jargon i would have had to figure out
and i was trying to think in sociology jargon at the time
so it was over my head
sociologists are allergic to logical scientific methods
i woud say the alternative to using negation
is to write in terms of positives
the idea of the pronouns
is that men use they rather than we or it or he
i think
they qualifies as masculine
similarly prepositional phrases rather than posessives
but i think a lot of the reasoning in the nature article
was just sexist stereotypes anyway
so i seriously doubt it will cross cultures very well
it is hard to get more feminine than bears eating raw bloody meat with their bare hands
i think most women would agree with me on this
you have stopped using capital letters
why
mehitabel would sell her soul for a mess of fi
Re: autobogotification
no subject
no subject
i wouldn t worry
it does not seem very accurate for net writing
that s why i said the reformatting
was the fun part
look i can t stop
writing archy poems
no subject
The main point is that most LJ entries are about the writers life/work/school/friends, leading an overwhelming use of possesive speech (the word "My" springs to mind), which the Genie interprets as a "female" score.
Or not.
no subject
Re: autobogotification
their problem was that they assumed the ridiculous stereotype was true and tried to find a model that reinforced it. i'd rather find a better model that doesn't use those assumptions and then see if there's actually any significant difference.
Re: autobogotification
silly
i m just not very good at it yet
my sociology prof actually told our class
that it s better to go into the field with a theory
and only look for evidence to support that theory
than the other way around
and they call it science
Re: autobogotification
now that your class is done you should finish my sandman stuff;-) when do you move back to college park? i'm guessing the 30th or 31st?