HardNLPnlp-353

Interpolated Bigram Probability

Problem

Interpolated language modeling mixes bigram and unigram estimates: lam * (bigram_count / bigram_total) + (1 - lam) * (unigram_count / unigram_total).

Given lam, the counts and totals for the bigram and unigram tables, return the probability.

Hint

The mixing weight trades specificity for coverage.

case 1

in: [0.7,10,100,20,1000]

expected: 0.076

case 2

in: [0.5,0,50,5,500]

expected: 0.005

case 3

in: [1,4,8,1,100]

expected: 0.5

Starter codePython

def interpolated_bigram_probability(lam, bigram_count, bigram_total, unigram_count, unigram_total):
    # Your code here
    pass

3 test cases · run your code against them in the editor