August 2018
Intermediate to advanced
438 pages
12h 3m
English
We now need to choose an appropriate model-performance evaluation metric to evaluate how well our model might be performing. A relevant metric here is the Bilingual Evaluation Understudy (BLEU) score. This is an excellent algorithm for evaluating the performance of models in machine translation when they translate languages. The motivation behind BLEU is that closer the generated output is to human-level translations, the better the score. To this day, it remains one of the most popular metrics to compare model output against human-level output.
The simple principle of the BLEU algorithm is to evaluate the generated text captions against a set of reference captions (typically one caption is evaluated ...
Read now
Unlock full access