How should I practice Most-Likely Next-Word Continuation Predictor?

Start by reading the full statement, solve the visible examples, then compare the pattern against related Bloomberg LP phone screen and onsite interview problems in FastPrep.

Is Most-Likely Next-Word Continuation Predictor associated with Bloomberg LP?

FastPrep tags this as a Bloomberg LP phone screen and onsite interview coding practice problem when the problem catalog has that company and stage context.

Most-Likely Next-Word Continuation Predictor

Bloomberg LP● MediumNEW GRADPHONE SCREENONSITE INTERVIEW

Learn

Problem statement

Train a word-transition predictor from trainingData. Every adjacent pair (word, nextWord) in a training sequence contributes one occurrence to that transition.

For each word, its predicted successor is the outgoing word with the greatest transition count. Break equal-count ties by lexicographically smaller successor.

For each word in queries, construct a continuation by repeatedly appending its predicted successor and using that successor as the next current word. Stop after maxWords appended words or as soon as the current word has no learned successor. Return one continuation row per query.

Function

predictContinuations(trainingData: String[][], queries: String[], maxWords: int) → String[][]

Examples

Example 1

trainingData = [["I","am","sam"],["am","sam"]]queries = ["I","am"]maxWords = 3return = [["am","sam"],["sam"]]

I transitions to am, then am transitions to sam. No successor was learned for sam.

Example 2

trainingData = [["a","b"],["a","c"],["a","b"],["b","d"]]queries = ["a","missing"]maxWords = 2return = [["b","d"],[]]

The transition a → b appears twice while a → c appears once. The unseen query has no continuation.

Example 3

trainingData = [["x","z"],["x","y"],["y","x"]]queries = ["x"]maxWords = 4return = [["y","x","y","x"]]

x → y wins the equal-frequency tie over x → z. The learned cycle is bounded by maxWords.

Constraints

The total number of words across trainingData is at most 100000.
1 <= queries.length <= 10000.
1 <= maxWords <= 100.
Every token and query is a nonempty case-sensitive string of at most 40 characters.
A training sequence may contain one word and therefore contribute no transition.