Two technical interviews were conducted.

The first round did not go so well. The interviewer seemed unhappy that day and eager to leave.

The second round interview was quite nice, and the interviewer helped me think through the interview questions.

Given training data as a list of tokenized words:

• ["I", "am", "a", "person"]
• ["I", "am", "good"]
• ["a", "person"]

Write a predictor that predicts a word's most frequent bigram.

For example:

• predict("I") -> "am"
• predict("a") -> "person"
• predict("am") -> "a"