Predict Next Word from Data | Dataford Interview Questions

Given a sample data set of sentences, write a function to predict next word.

Asked in the Round 3 stage. Build a frequency-based language model using trigrams, with fallback to bigrams and then overall word frequency.

Input and Output

sentences is a list of non-empty lowercase sentence strings. prefix is a non-empty string of lowercase words. Return the most frequent predicted word, breaking ties lexicographically, or None if the training data contains no words.

Example: sentences = ["i like tea", "i like coffee", "i like tea"], prefix = "i like" returns "tea".

Constraints

1 <= len(sentences) <= 1000
Each sentence contains at least one whitespace-separated word
The total number of words across all sentences is at most 10^5
Words contain lowercase letters and are separated by whitespace
Return None only when the training data contains no words

Function Signature

def predict_next_word(sentences, prefix):

Your question is Predict Next Word from Data. Start with the requirements in the Question tab.

Run and submit as often as you like. When you're ready, talk me through your approach or go straight to the code.

Problem