Questions for Self-monitoring: Sequence labeling with HMM

• What's the difference between pos tagging and named entitity recognition?
• How can a segment labeling task be transformed into a sequence labeling task?
• What's the difference between a generative and a discriminating model?
• Can a sequence labeling problem be solved by means of a logistic regression classifier?
• Why this solution would not be an optimal one?
• How can a global decision on the optimal label sequence be achieved?

• What are the components of an HMM?
• How are tokens and labels mapped to the components of an HMM?
• How are the different probabilities combined in decoding?
• Which simplifying assumptions an HMM is based upon?
• Why can we drop the denominator in Eq. 8.13?
• What's the relationship between the algorithm for determining the minimum edit distance and the Viterbi-algorithm?