Some guys came up with an algorithm that supposedly does a decent job of predicting gender based on writing samples. Check it out.
It did a bang-up job calling out my (academic) writing, though it should be noted that most of that is in the fields of business, econ, linguistics, and science. The only academic papers I ran through it that came up with false positives were sociology. Interesting stuff. Apparently "the" is a masculine word.
Posted by Megan (Member # 5290) on :
Apparently, based on the introduction to my dissertation proposal, I am male.
Posted by ElJay (Member # 6358) on :
Based on a long post from here I am just barely female, but it was a relationship advice post, so that makes sense.
Posted by kojabu (Member # 8042) on :
Mine are all split pretty much evenly, whether it's fiction or blog entries.
Posted by Fusiachi (Member # 7376) on :
quote:Originally posted by Megan: Apparently, based on the introduction to my dissertation proposal, I am male.
Out of curiousity, what is your area of study? Some fields seem to exhibit distinct writing styles--perhaps there is a connection?
Posted by Phanto (Member # 5897) on :
Or maybe this is a stupid program.
Question the axioms!
Posted by Shigosei (Member # 3831) on :
I ran a bunch of stuff through, including papers I've written for class, blog posts, Hatrack posts, and some old yearbook articles. Apparently, my writing style is male. This seems mostly to be based on my excessive use of the word "the." Also, the technical science paper writing style probably matches the "male" stereotype better. However, even my blog posts came out male. I had to dig out stuff from high school to get a female score. Even then, most of it registered male.
Posted by Phanto (Member # 5897) on :
That's because only men use the word the. Articles are so masculine.
Posted by ElJay (Member # 6358) on :
I put through another post and came out male. It's not quite interesting enough to keep trying. Posted by Hitoshi (Member # 8218) on :
It got my gender correctly on my fictional writing, but I think that's because it skewed the results by using "as" and "the" as masculine words, which of course there were many of.
On my blog writing though, it guessed female. I don't find that surprising though.
Posted by PrometheusBound (Member # 10020) on :
Right every time.... Posted by Fusiachi (Member # 7376) on :
quote:Originally posted by Phanto: Or maybe this is a stupid program.
Question the axioms!
Undoubtedly it is far from perfect. In fact, the site even indicates that it is using a simplified version of the algorithm. All the same, it is an interesting concept.
I put in excerpts from a few famous authors to see what might come up.
Dostoyevsky's The Idiot - Male Crime and Punishment - Male
J.K. Rowling - Female
Gabriel Garcia Marquez - Male in translation by Edith Grossman - Female
Pushkin, Eugene Onegin (in translation) - Male
Orson Scott Card, Hatrack River - Female.
------
Ok, so the site doesn't really say much that's... useful. Just mildly interesting.
Posted by PrometheusBound (Member # 10020) on :
Victor Hugo (in translation): Strongly male Hamlet's soliloquy : Male, but barely. Portia's speech: strongly male.
Ha, ha Mr. Shakespeare.
Posted by Celaeno (Member # 8562) on :
My existentialism final is apparently male, but my blog entries are female. It kind of makes sense.
I wonder what sort of results would turn up with the unsimplified version of the algorithm.
Posted by katharina (Member # 827) on :
Academic paper: male Personal statement: male Journal entry about relationships: female, by a hair Journal entry about an event: male, barely
Possible conclusions: 1. The algorithm is faulty, being a better judge of so-called male and female topics rather than male and female writing.
2. Women are forced to mimic men's writing to be taken seriously an academic world.
3. My writing is more male than female.
My completely-unscientific vote is for 1 & 2.
Theory about the why "the" is a masculine word: It's definitive. Calling something "the" is placing it as first in its group. Not "a" genius of the twentieth century, but "THE" genius of the twentieth century. Women are often seen as more conciliatory, and hence the algorithm expects them to place more ambiguity in their speech to allow others wiggle room in the interpretations.
Posted by Celaeno (Member # 8562) on :
I tried a few more things. My religious studies thesis was male, a work of fiction was female, and a relationship rant was female.
I'm going to have to throw my unscientific vote in with Katharina's.
Posted by kojabu (Member # 8042) on :
All of my academic papers are marked as male (religious studies courses). I don't think it's incredibly scientific either.
Posted by Euripides (Member # 9315) on :
The algorithm predicted I was male based on most of my short stories. I noticed a particularly wide margin between the male and female scores when I pasted in my first attempt at science fiction. Another short story I wrote which takes place in Nazi Germany and is interspersed with telegrams/letters in pseudo-official lingo also yielded a noticeable gap. Casual observation suggests the algorithm roughly equates 'personal, subjective' with female and 'third person, impersonal, objective' with male.
I'm definitely with katharina on her second point. The first point could be a consequence of the second.
Still, this is very interesting. Thanks for sharing it Fusiachi.
Posted by Hitoshi (Member # 8218) on :
I agree with Euripides. It seems to take expressions of feelings to be more feminine than masculine, just by going by the scores I got.
Posted by Shanna (Member # 7900) on :
Ooh, this is fun!
Parts from my last two academic papers (Shakespeare and Palahniuk/Plato) came back as male. So did my thesis proposal on Satan.
My blog entries were split. In one dealing mainly with my relationship, they came back as female. The ones about the rest of my life came back as male.
My fiction, the most interesting test for me, was also split. I was kind of excited that the first one came back male since it was written from the perspective of a male character. The second one though was also a male narrative but came back as female. (Oddly enough, the first piece dealt with a male homosexual affair and the second from a male relating his heterosexual interests)
A poem which recently won me some money came out with an exactly split score. The writer was of "unknown gender."
Posted by Phanto (Member # 5897) on :
Obviously your poem is a gender-transcending work .
I can understand the loose linguistic thinking behind this test. Yes, I bet you there *are* statisical differences betwenn the word choices that men + women use. But what are they, how meaningful are they, and to what degree are they influenced by your close society? (I.e., is it an inherent feature of English that, say, women will say XXXOOOAM! more than guys, or is it an inherent feature of women living on the East coast.)
But I doubt that this test is based on anything serious and of lasting interest beyond that, of course, of random amusement in a boring world .
Posted by ricree101 (Member # 7749) on :
I put about half a dozen papers through it, and it has been 100% accurate for me.
Judging by the posts here, it seems to have a harder time picking up female writing. Then again, this seems to be a pretty simplistic algorithm, so it may just be that the sort of people who frequent hatrack write in a way that is harder for the program to detect.
Posted by Uprooted (Member # 8353) on :
It was about 50-50 for me. It thought I was a man as often as it thought I was a woman in various kinds of writing. I also put a sample of male writing that came out female.
Posted by ReikoDemosthenes (Member # 6218) on :
It was very accurate for me when it looked at my academic writing and stories. But when I put in lj entries it read me as being definitely female.
Posted by Eaquae Legit (Member # 3063) on :
My academic writing comes out overwhelmingly male, even the essay I wrote about a women in Judaism book. My lj is definitely female, and the one short story I submitted was just barely female.
Posted by The Pixiest (Member # 1863) on :
it's pretty positive I'm a woman
Posted by Kwea (Member # 2199) on :
It is sure I am as well.
Too bad it is wrong.
Posted by Noemon (Member # 1115) on :
Okay, this is frustrating--I can get to the site, but when I click the "submit" (or whatever) button my work's firewall blocks is. Could someone do me a favor and enter in the text of the James Tiptree Jr. short story The Women Men Don't See? I'm curious to find out if this thing is any better at recognizing that she's a woman than Silverberg was.
Posted by Phanto (Member # 5897) on :
quote: Too bad it is wrong.
Hmm, I do wonder what this means exactly, Kwea ;-).
Posted by katharina (Member # 827) on :
Noemon, I tried it, and this is what I got.
quote: Female Score: 13565 Male Score: 15357
The Gender Genie thinks the author of this passage is: male!
So, male, but it was close.
Posted by Noemon (Member # 1115) on :
Interesting. Thanks kat!
Posted by Stephan (Member # 7549) on :
My landmark says I am a female.
Posted by Uprooted (Member # 8353) on :
I thought it was pretty funny that a piece of flash fiction that I, a woman, wrote--a parody of bad romance writing done in first person from a woman's POV--came out as male.
After coming out male so often on my journal entries I'm beginning to wonder if it's not just my neutral screenname that causes people on Hatrack to assume I'm male so often!
Posted by Icarus (Member # 3162) on :
I copied in all three of my landmarks, one at a time, and each came up female. Then I put in two short stories by me, and they both came up female.
This is why I always get worked up about those "this is how men are and this is how women are" threads--because according to those generalizations, I am a woman.
Posted by Noemon (Member # 1115) on :
Did you take that "brain gender" quiz from the BBC that I posted a year or two ago, Icarus? It'd be interesting to see how the results of that quiz correspond with the guesses made by this software.
Posted by Stephan (Member # 7549) on :
Using the Ender's Game short story OSC is officially male, but it is actually close with that many words used.
Words: 15579
(NOTE: The genie works best on texts of more than 500 words.)
Female Score: 17466 Male Score: 19335
The Gender Genie thinks the author of this passage is: male!
Posted by Noemon (Member # 1115) on :
I think that this is the quiz I'm thinking of, but because my work's firewall is blocking it I can't be sure.
Posted by Icarus (Member # 3162) on :
I've never seen that one before. That one also says I'm a woman. :-\ But I did take one that someone posted that had a bunch of random personality questions, and it's been teaching itself, kind of like the 20 Questions games online, and that one said I was male. Barely.
Posted by Architraz Warden (Member # 4285) on :
Ouch.
Female Score: 438 Male Score: 183
It's an odd conundrum trying to figure this out... Pretty much anyone who has had influence on my writing style during my life has been female. I wonder if that has influenced the results any.
Incidentally, the above passage is also female.
Posted by DaisyMae (Member # 9722) on :
Hmm. After using multiple examples I am decidedly female.
Glad that's been figured out.
Posted by James Tiberius Kirk (Member # 2832) on :
Hm. I wonder if a panel of ordinary readers would fare any better.
--j_k
Posted by ketchupqueen (Member # 6877) on :
3/4 of the things I ran through came out male. Huh.
Posted by ricree101 (Member # 7749) on :
quote:Originally posted by Noemon: I think that this is the quiz I'm thinking of, but because my work's firewall is blocking it I can't be sure.
Ouch, I came off solidly female there, so it seems that the writing test was more accurate for me. I'm not sure how good of a test this was, though. I'm curious to know how well this test matched up for other people on hatrack.
Posted by Kasie H (Member # 2120) on :
Mine are evenly split. My feature stories are all coming up female and my hard-news stories are coming up male.
Posted by Shan (Member # 4550) on :
Oh my goodness! threads dealing with gender and writing absolutely deserve to be linked to dingles Posted by pH (Member # 1350) on :
Based on a large chunk of my New Product Development final, I'm a man, baby, yeah!
-pH
Posted by Fusiachi (Member # 7376) on :
It should be noted that "dingles", in fact, are gender-neutral.
Posted by Icarus (Member # 3162) on :
Posted by pfresh85 (Member # 8085) on :
I tested both my blogs and my fiction writing. My fiction always came out as male by a decent margin (62% or above). My blogs are almost always male as well (with a few rare exceptions) but only by a slight bit (usually only 51% male or so).
Posted by Teshi (Member # 5024) on :
I put a bunch of my fiction in, and I'm almost always male, with the occaisional female- a story about a pregnant girl, for instance, ha ha. Even my very oldest stories, all about girls and twins and things, are male.
In non-fiction (school essays), I'm male.
I think that using words like "said" to indicate maleness is pretty thoughtless, but then this whole concept is pretty thoughtless.
:/
Posted by Icarus (Member # 3162) on :
I agree, but then, I would. Posted by scifibum (Member # 7625) on :
The site uses a simplified version of the algorithm huh?
I'm going to use this logic as much as possible. Well, of course that work was sloppy and useless, but I was only using a simplified version of my competence.
Posted by Fusiachi (Member # 7376) on :
There internet is not required to be useful. It is free to be as sloppy as it likes.
Furthermore, there is probably a sound reason why they didn't make an online version more robust. Hosting/bandwidth fees come to mind.
Posted by Orincoro (Member # 8854) on :
The plain fact that "feminine" triggers are words that as a rule are rare in academic papers, tells me that this whole process is basically flawed.
I submitted a paper on Mozart string quartet, and there would have been near zero ocassion to use the words "me" "myself" "we" "hers" "she" or "your." On the other hand the male words are all essential in an academic paper of any kind. This algorithm is based partly on a sexist principle, that women are morely likely in general to write in a non-academic setting. Ridiculous.
This statement from the paper is interesting though "Those features which are more prevalent in male writing are almost invariably more prevalent in non-fiction."
Edit: I must say that I find the related paper linked on the site to be much more interesting than the algorithm. The abstract references the fact that the study is based on writing for an "unseen audience," which substantiates the notion that the author is writing "to" no sex in particular. Knowing myself slightly better than I know any other writer, I think I write more with the idea of a female reader than a male (I am male). This probably has to do with, or is the cause of, my having collaborated much more with female writers on many writing projects, and having talked MUCH more with women about writing than with men. For example, I have only ever peer edited female friends' writing, and I have only ever asked women to read my work. I don't exactly know why this is, except that I grew up as the only boy in a family with three sisters.
This is something I'd like to explore further though, since the thought of who I was writing TOO has never really occurred to me before.
[ December 30, 2006, 10:44 PM: Message edited by: Orincoro ]
Posted by Tante Shvester (Member # 8202) on :
I submitted six of my short fiction pieces for analysis. The five that I liked best were female. The one that I didn't care for much was male.
Good writing may just be a feminine trait, huh?
Posted by fiddle_stix (Member # 9941) on :
apparently im a dude Posted by Tante Shvester (Member # 8202) on :
Trapped in a woman's body, huh, f'stix? Happens sometimes.
Posted by fiddle_stix (Member # 9941) on :
it's a tough life, but i'll manage
Posted by Phanto (Member # 5897) on :
quote: This algorithm is based partly on a sexist principle, that women are morely likely in general to write in a non-academic setting. Ridiculous.
Is it wrong? Who publishes more academic papers, men or women? If it is men and by a significant margin, then will you rescind your comment? Would you be surprised to hear that in medical fields at least women are vastly outpublished by men?
Are women then not less likely to write academic papers?
Posted by Dr Strangelove (Member # 8331) on :
I'm a guy through and through. Posted by Shigosei (Member # 3831) on :
I think the thing that bothers me is that "academic" writing is somehow more masculine.
Posted by Avatar300 (Member # 5108) on :
quote:Originally posted by Phanto:
quote: This algorithm is based partly on a sexist principle, that women are morely likely in general to write in a non-academic setting. Ridiculous.
Is it wrong? Who publishes more academic papers, men or women? If it is men and by a significant margin, then will you rescind your comment? Would you be surprised to hear that in medical fields at least women are vastly outpublished by men?
Are women then not less likely to write academic papers?
You forgot to quote this part:
quote:On the other hand the male words are all essential in an academic paper of any kind.
I think it's fairly obvious that Orincoro means that most academic writing will bias toward a "male" classification because the words chosen to represent "maleness" are found in academic writing far more than those chosen to represent "femaleness." The relative numbers of actual men and women writing in an academic setting then have no relationship to the results of this algorithm.
//Edit to add:
Fairly obvious unless I'm putting words in his mouth. Which, as a male, I tend to do.
[ December 31, 2006, 01:34 AM: Message edited by: Avatar300 ]
Posted by aspectre (Member # 2222) on :
"apparently im a dude " "Trapped in a woman's body, huh, f'stix? Happens sometimes." "it's a tough life, but i'll manage"
Why can't a woman be more like a man? Men are so honest, so thoroughly square; Eternally noble...
Posted by Teshi (Member # 5024) on :
quote:Is it wrong? Who publishes more academic papers, men or women? If it is men and by a significant margin, then will you rescind your comment? Would you be surprised to hear that in medical fields at least women are vastly outpublished by men?
I think you're [plural you're] missing the point. The algorithm's goal is to discern the gender of the author. If it fails, then it fails.
Concerning academic papers: Since many (if not most) universities have more women attending than men, I should think that in studentland, at least, there are more papers written by women than by men.
I think suggesting that you can generalize that academic language is male is as thoughtless as the algorithm itself.
If you were going to make this kind of algorithm work at least, let's say 80% of the time, you'd have have to take in to account a lot more than the language, and assign probabilities. For instance, you could say that if the paper is about women's rights, it's more likely to be written by a woman. If there are more male engineers, you could assign the reverse etc. Then you could take into account sentence structure- maybe. Then maybe word usage, if you could find enough of a correlation. I think just using word usage is insulting to both genders.
Hopefully, the original algorithm was far more complex than this one.
Posted by Icarus (Member # 3162) on :
Someone should find an electronic copy of "The Women Men Don't See," by Triptree, and run it through this.
Posted by Orincoro (Member # 8854) on :
quote:Originally posted by Phanto:
quote: This algorithm is based partly on a sexist principle, that women are morely likely in general to write in a non-academic setting. Ridiculous.
Is it wrong? Who publishes more academic papers, men or women? If it is men and by a significant margin, then will you rescind your comment? Would you be surprised to hear that in medical fields at least women are vastly outpublished by men?
Are women then not less likely to write academic papers?
Calculating for a specific paper, in an individual case, by assessing the greater odds of whether more or less papers are published by women is cheating. It may return a statistically accurate sampling, but it does not reliably judge individual samples.
The algorithm should judge each case individually and on its on merits. However this algorithm is flawed because it maintains assumptions about "feminine" writing, specifically that women are less likely to be the authors of "academic" papers, and therefore associating "academic" qualities with male writers. However if you read the paper, you'll find that IT maintains that academic writing, by nature makes the gender of the author ambiguous. The fact therefore, that academic writing is similar in ways to male fiction writing does not signify when you are judging an academic paper- and the results should not return as male because the paper is "academic" in nature.
If the algorithm depends on broad statistical probabilities in order to judge specific cases, such as the idea (and I do not submit that it IS a fact because I don't know) that men publish more academic papers, and therefore are more likely to be the author of any specific paper, then the algorithm is basically flawed. It is unreliable in judging specific samples, and will only (possibly) return a statistically consistent number of male over female responses. This does not make the test accurate or reliable, it makes it typical internet BS, imho. The paper is much more clear, I'd suggest reading that instead of using the algorithm.
Posted by Orincoro (Member # 8854) on :
quote:Originally posted by Shigosei: I think the thing that bothers me is that "academic" writing is somehow more masculine.
Reverse the descriptors then, it's like conceptual algebra:
Masculine writing is somehow more "academic"
This is highlighted in the paper, and it is attributed to fundamental differences in the way that male and female writers explore objects and relationships. In fiction writing, men more often describe objects and women more often describe relationships. There is also a point made in the paper about feminine fiction writing having a stronger awareness of the self. These qualities are somewhat un-academic, but when women write academically, they seem to do just as well as men, it's just that men write that way all the time (or much more).
Posted by katharina (Member # 827) on :
Perhaps it is that when men get around to writing, they write about things. Perhaps it is not that all their writing is academic, but that they only write when there is a due date.
There are so many possible reasons for the obvious failure of this algorithm that the guesses as to the reasons tell me a great deal more about the poster than about the algorithm.
I'm annoyed with some of you. *wrinkles nose*
Posted by Teshi (Member # 5024) on :
quote:In fiction writing, men more often describe objects and women more often describe relationships.
![[Wink]](wink.gif)
![[Eek!]](eek.gif)
![[Smile]](smile.gif)
.![[Dont Know]](graemlins/dontknow.gif)
![[Evil]](graemlins/evil.gif)
![[Big Grin]](biggrin.gif)