Path9.Mod1.d - Selecting Text Analyics and Recommender Algorithms for Azure ML

Question 1

When you want to Extract information from text after having performed cleaning operations like Stop-Word Removal and Case Normalization

Answer 1

Preprocess Text (Know this for the exam!)

Question 2

When you want to Extract information from text by converting words to values used by NLP tasks like Recommenders, NER and Machine Translation

Answer 2

Word2Vector…naturally since most all NLP tasks require words be translated to decimal vectors for plotting…

Question 3

When you want to Extract information from text by grouping similar text through Unsupervised Topic Modeling

Answer 3

Latent Dirichlet Allocation

Pronouced “Dur-ish-lay”

Question 4

When you want to Extract information from text using N-Grams

What N-Grams are…

Answer 4

Extract N-Gram Features From Text

N-Grams: Continous symbols or words, or tokens in documents. the N is the number of tokens in the series or grouping:

N | Name | Sample
1 |Unigram | [“green”,”eggs”,”and”,”ham”]
2 | Bigram | [“green eggs”,”eggs and”,”and ham”]
3 | Trigram |	[“green eggs and”, “eggs and ham”]

Question 5

When you want to Extract information from text using the Vowpal Wabbit library to convert text to Integer-Encoded Features

Answer 5

Feature Hashing

Question 6

When you want to Generate Recommendations based on a collaborative filtering and content approach, using a hybrid recommender

Answer 6

Wide & Deep Recommender

Question 7

When you want to Generate Recommendations using collaborative filtering, better performance and with lower cost through reducing dimensionality

Explain the algorithm this Recommender is based on…

Answer 7

SVD Recommender

SVD: Singular Value Decomposition. Basically a matrix factorization technique, which decomposes any matrix into 3 generic and familiar matrices.