Genderizer is a language independent module which tries to detect gender by looking at given first names and/or analyzing sample texts.
Data-driven predictions can succeed --and they can fail. It is when we deny our role in the process that the odds of failure rise. Before we demand more of our data, we need to demand more of ourselves. -- The Signal and The Noise by Nate Silver
You can install this package using the following
$ sudo pip install genderizer
from genderizer.genderizer import Genderizer print Genderizer.detect(firstName = 'John') # >>> male print Genderizer.detect(firstName = 'Marry') # >>> female print Genderizer.detect(firstName = 'mustafa') # >>> male print Genderizer.detect(firstName = 'fatma') # >>> female print Genderizer.detect(firstName = 'fikret') # >>> None print Genderizer.detect(firstName = 'fikret', text='galatasary maçı kaçmaz') # >>> male print Genderizer.detect(firstName = 'fikret', text='annemlerle yemek yedik') # >>> female print Genderizer.detect(text='askerlik yoklamasını kaçırdım mk') # >>> male print Genderizer.detect(text='bana çiçek alan erkek için canım feda') # >>> female print Genderizer.detect(firstName='fatma', text='askerlik yoklamasını kaçırdım mk') # >>> female print Genderizer.detect(text='futbol sevgi') # >>> None print Genderizer.detect(text='lan bi siktir git') # >>> male
Note: You may claim that women can say
lan bi siktir git, of course. But the probability of being female is less than the probability of being male according to the trained data of the classifier.
So it is obvious that the success of detection depends on the trained data.
By the way, in Turkish saying 'lan bi siktir git' makes you quite rude.
How It Works
Genderizer is a module which tries to detect gender by looking given first names and/or analyzing sample text of a person.
If a first name is definitely used for only one gender, the system will accept this gender and will not make any further analysis. For example, while 'Mustafa', 'Osman', 'Hasan' is used in Turkish only for male; 'Fatma', 'Ceyda', 'Elif' only for female.
When looking at first names does not infer any gender for sure, the system will make text analysis if it is given. For example; 'Ekim', 'Meric', or 'Tümay' is used for both male and female.
The text analysis is the classification of sample texts. It simply try to compute the probability of being male or female mining the sample text. In this system Naive Bayesian Classification is adopted and naiveBayesClassifier is used.
How To Improve It
TODO: write a step by step guide
##Preparing Language Dependent Training Sets TODO: give a few examples
Customization and Optimization
Using Memcached For Speed
""" Under heavy usage, for example tens of thousands detection request in a few seconds the default configuration could not meet the demand. By the default configurations, genderize will load necessary data from files and this is well known to be slow. Instead of each time loading data into memory, doing this one time will be clever approach. One of the best way of this approach is to use memcached. For more information have a look at the documentation of memcached. Genderizer provides a memcached interface to store first names in memory. To active this interface, you need to instantiate memcachedNamesCollection interface and pass it to genderizer while initializing it. """ from genderizer.memcachedNamesCollection import MemcachedNamesCollection #For memcached, do not forget to setup the memcached server. MemcachedNamesCollection.memcacheHost = '127.0.0.1:11211' Genderizer.init( namesCollection=MemcachedNamesCollection ) print Genderizer.detect(firstName = 'John')
""" If you want to use Genderize on Mongodb for arbitrary reasons, the MongoNamesCollection first names collection interface will do much of the necessary works for you. """ from genderizer.mongoNamesCollection import MongoNamesCollection MongoNamesCollection.mongodbURL = 'mongodb://192.168.1.170' Genderizer.init( namesCollection=MongoNamesCollection ) print Genderizer.detect(firstName = 'Marry')
Custom Text Classifier
""" NaiveBayesClassifier is adopted as the default classifier. But you can use another, entirely different classifier; as long as the classifier has a 'classify' method taking text as a parameter. For more information please have a look at the naiveBayesClassifier project's documentation. https://github.com/muatik/naive-bayes-classifier """ from naiveBayesClassifier import tokenizer from naiveBayesClassifier.classifier import Classifier from cachedModel import CachedModel Genderizer.init( lang='en', classifier=Classifier(CachedModel.get('en'), tokenizer) ) print Genderizer.detect(firstName = 'fikret', text='annemle kahve keyfi')
- inline docs
- Mustafa Atik @muatik
- Nejdet Yucesoy @nejdetckenobi