The main purpose of a Hidden Markov Model (HMM) is to model sequences of observable events that depend on hidden states.

To generate random sequences of observable events without any hidden states.

To classify data into distinct categories without considering sequences.

To predict future hidden states based on past observations.

Generative models are always more accurate than discriminative models.

Generative models require labeled data; discriminative models do not.

Generative models only classify data; discriminative models generate data.

Generative models model the joint distribution; discriminative models model the conditional distribution.

Maximum Entropy is a technique for minimizing uncertainty in predictions.

Maximum Entropy is a principle used in statistical modeling to derive the most uniform probability distribution that satisfies specific constraints.

Maximum Entropy refers to the highest possible entropy in a closed system.

Maximum Entropy is a method to maximize the likelihood of a dataset.

CRFs are discriminative and consider the entire sequence context, while HMMs are generative and rely on Markov assumptions.

HMMs consider the entire sequence context while CRFs do not.

CRFs are generative and rely on Markov assumptions.

CRFs are only used for classification tasks, unlike HMMs.

Variables, Constants, Functions, Operators

Grammar trees, Lexical analysis, Syntax analysis

Syntax rules, Semantic rules, Parsing techniques

Terminals, Non-terminals, Start symbol, Production rules

Use HMM for simple sequence labeling tasks.

Use CRF over HMM when modeling complex dependencies between output labels and when the entire observation sequence context is important.

Choose CRF when data is limited and context is irrelevant.

Prefer HMM when the output labels are independent of each other.

The training process for a Maximum Entropy model involves defining feature functions, collecting labeled training data, initializing model parameters, optimizing the likelihood using an algorithm, and validating the model.

Applying a linear regression model instead

Using only unlabelled data for training

Ignoring feature functions during optimization

The transition probability represents the initial state of the HMM.

The transition probability indicates the likelihood of moving between hidden states in an HMM.

The transition probability measures the accuracy of observations in an HMM.

The transition probability defines the output symbols in an HMM.

CFG is only used for programming languages, not natural languages.

CFG is not suitable for any form of language processing.

CFG can fully capture all aspects of natural language.

Yes, CFG can be used for natural language processing, but it has limitations.