In a bigram model, how is the next word predicted?

In a bigram model, how is the next word predicted?

• A. Using the entire history of previous words.
• B. Using the next word to backward predict.
• C. It looks at the word prior to predict the probability of the word to come.
• D. It ignores previous words and uses semantic similarity.

Answer: (C)

In a bigram model, the prediction of the next word is based on the immediately preceding word. The model estimates the probability of each possible next word conditioned only on the word that came right before it, i.e., P(next word | previous word). This is typically done using counts from a training corpus: P(next | previous) = count(previous, next) divided by count(previous). So the model asks, given the word before, what are the likely words to come next? This approach focuses on the prior word to determine the next word, rather than looking at longer histories or using semantic similarity. It also doesn’t predict backwards; it predicts forward from the preceding word.

In a bigram model, the prediction of the next word is based on the immediately preceding word. The model estimates the probability of each possible next word conditioned only on the word that came right before it, i.e., P(next word | previous word). This is typically done using counts from a training corpus: P(next | previous) = count(previous, next) divided by count(previous). So the model asks, given the word before, what are the likely words to come next?

This approach focuses on the prior word to determine the next word, rather than looking at longer histories or using semantic similarity. It also doesn’t predict backwards; it predicts forward from the preceding word.