I completely relate to your struggle with Hidden Markov Models (HMMs). It is definitely a challenging topic. In fact, the most confusing topic for me in this course is the ...
I completely relate to your struggle with Hidden Markov Models (HMMs). It is definitely a challenging topic. In fact, the most confusing topic for me in this course is the use of mathematical matrices and algorithms, specifically the HMM and the Viterbi algorithm used for Part of Speech tagging. Coming from a traditional linguistics background, I am used to analyzing language through standard grammar rules and human context. However, these advanced computational methods approach language completely differently. Instead of reading words, the machine converts sentences into complex probability matrices to calculate the most likely grammatical tag for each word.
Trying to understand how a computer multiplies 'transition probabilities' and 'emission probabilities' in a giant matrix just to identify a simple noun or verb is extremely overwhelming for me. Because I do not have a strong foundation in IT or advanced mathematics, visualizing how these abstract numbers and mathematical formulas actually process real human language is my biggest challenge. Like you, I realize I will need a lot more time and simple practical examples to bridge this gap between linguistic theory and machine calculation.
Trying to understand how a computer multiplies 'transition probabilities' and 'emission probabilities' in a giant matrix just to identify a simple noun or verb is extremely overwhelming for me. Because I do not have a strong foundation in IT or advanced mathematics, visualizing how these abstract numbers and mathematical formulas actually process real human language is my biggest challenge. Like you, I realize I will need a lot more time and simple practical examples to bridge this gap between linguistic theory and machine calculation.