Which Machine Learning algorithm should I use for gender tagging of words in NLP? I have a dataset of about 26,000 words along with their gender tags [m, f or any]. 
Which ML algorithm should I use for Gender Tagging/Classification purpose. How should I go about it? What should be the input features? 
Given a word , it's POS tags and in some cases the singularity I am trying to predict the gender of the word. 
 A: I would use one-hot encoding to map the gender of the word as follows:
male --> [1, 0, 0]
female --> [0, 1, 0]
other --> [0, 0, 1]
This tells your logistic regression that the input maps to 100% of that class of the output.  Once you train your logistic regression algorithm, your predictions will be of the following form after passing through the final Softmax function:
[0.1, 0.9, 0.1]
Indicating that the word is female with 90% probability.
A: I would start with domain knowledge.  What do you know about what patterns might be present in the language that a human might use to form a guess about the gender of a particular word?
For instance, Spanish has some simple rules: words that end in 'o' are often masculine, and words that end in 'a' are often feminine.  But not always; there are exceptions.  So, in Spanish, a good feature to use would be the last letter of the word.  You could also try using the last two letters as a feature.  
As you can see, this is likely to be language-specific.  You could try to do something language-independent, but I expect that language-specific features that take into account some domain knowledge about the language will enable your ML algorithm to perform a lot better.
