Natural Language Processing in Google Cloud

Question 1

What are key features on NLP API?

Answer 1

• Entity extraction
• Sentiment analysis
• Entity sentiment analysis
• Text classification
• Syntax analysis

Question 2

What is stemming?

Answer 2

Stemming is a process of converting a word to its root representations ex. playing -> play

Question 3

What is normalization in NLP?

Answer 3

Normalization is a process of converting a phrase to its normal form ex. lol -> laugh out loud, wtf -> what the fuck. It is useful for social networks.

Question 4

Explain bag-of-words approach.

Answer 4

Bag-of-words represents a tecnique for converting text to numeric representation. It is similar to one-hot encoding but in this case each occurence of the word in the sentece is marked in vocabulary. You can track the occurence for each word like marking 2 for a dog or ignore number of occurence of the same word.

Question 5

What is word2vec?

Answer 5

word2vec reprents a family of model architectures and optimization that are used to learn word embeddings from a large dataset.

Question 6

What are 2 most common model architectures in word2vec?

Answer 6

CBOW (continous bag-of-words) - predicts center word by providing a surrounding context
Skip-gram - predicts surrounding context by providing a center word

For both architectures you first need to traing a neural network by creating training data.

CBOW

Question 7

What is a Tensorflow Hub?

Answer 7

Tensorflow Hub is platform that contains pretrained models that you can use to fine-tune and deploy.

Question 8

How do you decide if pretrained embeddings model should be fine-tuned or not?

Answer 8

It depends on the amount of training data. You should not fine-tune the model if you don’t have too much data to avoid overfitting.

Question 9

What is the difference between ANN (artificial neural network) and DNN (deep neural network)?

Answer 9

ANN has only one hidden layer.