Download presentation

Presentation is loading. Please wait.

Published byRafe Welch Modified over 3 years ago

1
POS Tagging & Chunking Sambhav Jain LTRC, IIIT Hyderabad

2
POS Tagging - Introduction Need for POS Tag ? – Assigning category which are < Vocab – Essential for systems – Has grammatical information 2POS Tagging and Chunking

3
Building a POS Tagger Rule Based Approaches – Come up with rules – Eg. Substitution: The {sad,green,fat,old} one in the garden. Statistical Learning / Machine Learning – Make machine learn automatically from annotated instances – Unsupervised - Baum Welch Pros/Cons of each approach 3POS Tagging and Chunking

4
Machine Learning ---- ??? How can machine learn from examples (on Board) 4POS Tagging and Chunking

5
Problem Statement Given W = w0 w1 w2 w3 w4 w5 w6 w7 Find T = t1 t2 t3 t4 t5 t6 t7 OR That T for which P(T/W) should be maximum 5POS Tagging and Chunking

6
Hidden Markov Models P(T/W) = P(W/T)*P(T)/P(W) T= Argmax T P(W/T)*P(T)/P(W) = Argmax T P(W/T)*P(T) P(W/T) = P(w0 w1 w2 w3 w4 w5 w6 w7 / t1 t2 t3 t4 t5 t6 t7 ) (On Board) – Chain Rule Markov Assumption 6POS Tagging and Chunking

7
Hidden Markov Models How can we learn these values form annotated corpus. Emission Matrix Transition Matrix Example (on Board) 7POS Tagging and Chunking

8
Hidden Markov Models Given Emission/Transition Matrix – Tag a new sequence – Complexity Viterbi Algorithm - Decoding Example (On Board) 8POS Tagging and Chunking

9
Viterbi Algorithm (Decoding) Most probable tag sequence given text: T*= arg max T P λ (T | W) = arg max T P λ (W | T) P λ (T) / P λ (W) (Bayes’ Theorem) = arg max T P λ (W | T) P λ (T) (W is constant for all T) = arg max T i [ a(t i-1 t i ) b(w i | t i ) ] = arg max T i log [ a(t i-1 t i ) b(w i | t i ) ] 9POS Tagging and Chunking

10
t1t1 t2t2 t3t3 w1w1 t1t1 t2t2 t3t3 w2w2 t1t1 t2t2 t3t3 w3w3 t0t0 A(,) t1t1 t2t2 t3t3 t 0 0.0050.020.1 t 1 0.020.10.005 t 2 0.50.0005 t 3 0.05 0.005 B(,) w1w1 w2w2 w3w3 t1t1 0.20.005 t2t2 0.020.20.0005 t3t3 0.02 0.05 10POS Tagging and Chunking

11
-log A t1t1 t2t2 t3t3 t 0 2.31.71 t 1 1.712.3 t 2 0.33.3 t 3 1.3 2.3 -log B w1w1 w2w2 w3w3 t1t1 0.72.3 t2t2 1.70.73.3 t3t3 1.7 1.3 t1t1 t2t2 t3t3 w1w1 t1t1 t2t2 t3t3 w2w2 t1t1 t2t2 t3t3 w3w3 t0t0 -1.7 -0.3 -1.3 -3 -3.4 -2.7 -2.3 -1.7 -6 -4.7 -6.7 -1.7 -0.3 -1.3 -7.3 -9.3 -10.3 11POS Tagging and Chunking

12
Tools for POS Tagging It is a Sequence Labeling Task Tools – HMM based – TNT Tagger (http://www.coli.uni-saarland.de/~thorsten/tnt/)http://www.coli.uni-saarland.de/~thorsten/tnt/ – CRF based – CRF++ (http://crfpp.googlecode.com/svn/trunk/doc/index.html)http://crfpp.googlecode.com/svn/trunk/doc/index.html 12POS Tagging and Chunking

13
CRF for Chunking (On Board) Tools – CRF++ (http://crfpp.googlecode.com/svn/trunk/doc/index.html)http://crfpp.googlecode.com/svn/trunk/doc/index.html 13POS Tagging and Chunking

14
Understanding CRF Template # Unigram U00:%x[-2,0] U01:%x[-1,0] U02:%x[0,0] U03:%x[1,0] U04:%x[2,0] U05:%x[-1,0]/%x[0,0] U06:%x[0,0]/%x[1,0] U10:%x[-2,1] U11:%x[-1,1] U12:%x[0,1] U13:%x[1,1] U14:%x[2,1] U15:%x[-2,1]/%x[-1,1] U16:%x[-1,1]/%x[0,1] U17:%x[0,1]/%x[1,1] U18:%x[1,1]/%x[2,1] U20:%x[-2,1]/%x[-1,1]/%x[0,1] U21:%x[-1,1]/%x[0,1]/%x[1,1] U22:%x[0,1]/%x[1,1]/%x[2,1] # Bigram B 14POS Tagging and Chunking

Similar presentations

OK

Dishonest Casino Let’s take a look at a casino that uses a fair die most of the time, but occasionally changes it to a loaded die. This model is hidden.

Dishonest Casino Let’s take a look at a casino that uses a fair die most of the time, but occasionally changes it to a loaded die. This model is hidden.

© 2019 SlidePlayer.com Inc.

All rights reserved.

To make this website work, we log user data and share it with processors. To use this website, you must agree to our Privacy Policy, including cookie policy.

Ads by Google