Google Professional Machine Learning Engineer
Get started today
Ultimate access to all questions.
You are part of a machine learning team at a gaming company with millions of global customers. Each game features a real-time chat function that supports over 20 languages, with messages translated using the Cloud Translation API. The task is to build an ML system to moderate these chat messages instantly and ensure uniform performance across all languages, without altering the current serving infrastructure. Your initial model was trained with an in-house word2vec for embedding chat messages translated by the Cloud Translation API, but it showed significant performance variation across different languages. What should be your next step to improve the model's performance consistency across all languages?
You are part of a machine learning team at a gaming company with millions of global customers. Each game features a real-time chat function that supports over 20 languages, with messages translated using the Cloud Translation API. The task is to build an ML system to moderate these chat messages instantly and ensure uniform performance across all languages, without altering the current serving infrastructure. Your initial model was trained with an in-house word2vec for embedding chat messages translated by the Cloud Translation API, but it showed significant performance variation across different languages. What should be your next step to improve the model's performance consistency across all languages?
Explanation:
The model shows significant performance differences across various languages, indicating that the translation process might be introducing noise or inconsistent quality. Training a classifier using the chat messages in their original language can address this issue more effectively. This approach allows the model to understand the nuances and context of each language directly, improving the overall performance consistency across different languages. Additionally, this method avoids reliance on the Cloud Translation API, which may be the source of variability.