Gender Identification of Unfamiliar Names

1
Gender Identification of Unfamiliar Names

• Lia McClanahan

2
He looked at the dragon thoughtfully. "You
need a name.  I heard some interesting ones
today perhaps you'll like one."  He
mentally ran through the list Brom had given him
until he found two names that struck him as
heroic, noble, and pleasing to the ear.  "What do
you think about Vanilor or his successor,
Eridor?  Both were great dragons."  No,
said the dragon.  It sounded amused with his
efforts.  Eragon. "That's my name you
can't have it," he said, rubbing his chin. 
"Well, if you don't like those, there are
others."  He continued through the list, but the
dragon rejected every one he proposed.  It seemed
to be laughing at something Eragon did not
understand, but he ignored it and kept suggesting
names.  "There was Ingothold, he slew the..."  A
revelation stopped him.  That's the problem! 
I've been choosing male names.  You are a she!
Yes.  The dragon folded her wings smugly. 

3
Now that he knew what to look for, he came
up with half a dozen names.  He toyed with
Miremel, but that did not fit -- after all, it
was the name of a brown dragon.  Opheila and
Lenora were also discarded.  He was about to give
up when he remembered the last name Brom had
muttered.  Eragon liked it, but would the
dragon?  He asked. "Are you Saphira?"  She
looked at him with intelligent eyes.  Deep in his
mind he felt her satisfaction. Yes, 
Something clicked in his head and her voice
echoed, as if from a great distance.  He grinned
in response.  Saphira started humming.  
Paolini, Christopher.  Eragon.  Alfred A.
Knopf New York, 2002. 
4
Do these names sound like boy names and girl
names?

• Valinor
• Eridor
• Ingothold

• Miremel
• Ophelia
• Saphira

5
Previous Work

• Slater and Feinman, 1985, found sex-associated
  features in the structural characteristics of
  names such as manifesting more or fewer sounds,
  syllables, and word-final sonorants.
• Cutler, 1990, noticed differences in length, in
  proportion of weak initial syllables, in vowel
  sounds, and in similarity to common nouns.
• Barry and Harper, 1995, suggested a causative
  link with cultural and environmental conditions
• Hough, 2000, posits an explanation that deals
  mostly with the origin of the names the
  differences result from a range of contributing
  factors, including gender-based structures in the
  source languages, differences between the
  proportions of the various source languages
  represented in the masculine and feminine name
  stock, and the higher frequency of suffixes in
  feminine names.

6
What I Want to Know

• Do speakers use the phonology of a name to tag
  its gender?
• How to test this ask speakers to tag completely
  unfamiliar names

7
Methodology

• An e-mail survey with 50 names
• Classmates,This is a survey I'm conducting for
  my ling 490 paper. If you'd like to participate,
  just reply to this e-mail with your responses
  typed in. You're to answer the question, "Does
  this sound like a boy's name or a girl's name?"
  Most of the names will be unfamiliar to you.
  Please say them out loud before you decide.Type
  "M" for male or "F" for female. You may also type
  "C" for "could go either way."I'm analyzing the
  data Monday night, if you could reply by then. If
  you have friends that would like to take the
  survey, the more the merrier, statistically
  speaking. Thanks.Does this sound like a boy's
  name or a girl's name?SandoRomelincRosepTi
  monegiJaqueloNikertetc.

8
Methodology

• Where did the names come from?
• The Social Security Administration has a website
  containing name data from every U.S. Census from
  1879 to 2006.
• http//www.ssa.gov/OACT/babynames/

9
Where did the names come from?

• A computer program used that data to train, for
  each gender, a Markov Model a probabilistic
  model that can generate a character that is
  probably based on given characters. In our case,
  we used the previous two characters.
• Example
• In the male names model, the most common first
  letter is J. Given J, the next most common
  character is a. Given Ja, the most probable
  character is m.
• Frequency and sequence

10
MethodologyExample lists Can you tell which
model is which?

• Jayle
• Kenah
• Megan
• Hophi
• Jessi
• Desle
• Janahi
• Alahin
• Halize
• Magann
• Marenn
• Kianna
• Sanahle
• Camiann
• Maleahl
• Justime
• Dessayl
• Frislen
• Laiyanya

• Ezell
• Alenn
• Abdua
• Trenn
• Anton
• Justo
• Domiah
• Howano
• Kienny
• Ranavi
• Darone
• Ralius
• Frayler
• Aliusto
• Cobersh
• Julinne
• Santerl
• Joseffi
• Treenzo

11
Results

• 92 respondents
• An average of about 70 percent answered the
  gender that according to the Markov Model is
  statistically more likely

12
(No Transcript)
13
(No Transcript)
14
Conclusion

• It seems that speakers have in their minds a
  mental inventory of sound sequences and
  frequencies that helps them decide what gender a
  name is. This inventory allows them to do 40
  percent better than random guessing.

15
Future Work

• Ling 420 paper Effect of Orthography?