# Jarrod Kahn

• ## Part-Of-Speech Tagging

Jan. 28, 2018

A Part-of-Speech (POS) tagger is a tool that assigns parts of speech to each word of a sentence, such as verb, noun, adjective, etc. This post will summarize a POS tagger introduced in 1996 by Adwait Ratnaparkhi in his seminal paper A Maximum Entropy Model for Part-Of-Speech Tagging, which you can find here.

This method of POS tagging uses a probabilistic model that is optimized using maximum likelihood estimation. The model is defined on a set of word and tag contexts $$\mathcal{H}$$ and the set of possible tags $$\mathcal{T}$$. We want to represent a probability distribution over all ...

• ## The Viterbi Algorithm

Jan. 26, 2018

The Viterbi algorithm is a dynamic programming algorithm that efficiently computes the the most likely states of the latent variables of a Hidden Markov Model (HMM), given an observed sequence of emissions. The co-founder of Qualcomm and University of Southern California alum Andrew Viterbi (go Trojans!) proposed the algorithm in 1967. This post will gradually work from theory to a Python implementation.

### An Introduction to Hidden Markov Models

Hidden Markov Models (HMMs) model a system of discrete temporal unobserved (hidden) variables and discrete temporal observed variables. The observed and unobserved variables are related through emission probabilities.

The diagram above depicts ...

• ## Ötztal Alps, Spring 1932

Dec. 31, 2017

A post shared by Jarrod Kahn (@thekahnrad) on