In what cases (and why) does using Binary Occurrence instead of TF-IDF makes more sense?
When the vocabulary size is small, and the binary occurrence of given words are strong features
Data Preparation
What is Vector Normalization? How is that useful?
If the documents in the corpus are of varying sizes, the larger documents are more likely to have higher word counts
What is the problem with using a generic list of stop words?
The generic list of English stop words may not be appropriate if the set of documents are all related to a specific domain.
How to identify Stop Words?
Stop words are common words that appear often throughout a set of documents but add little information
What is Lemmatization?
Lemmatization is the process of reducing tokens in a document to their root form based on the context in which they appear.
What happens to new words that appear in Test dataset but are not present in Training Data?
Similar to other preprocessing techniques, it is considered best practice to fit the vectorizer on the train dataset
What are the Advantages/Disadvantages of a n-gram model
Advantages: Can provide useful information beyond just considering individual tokens
What is an n-gram model?
An n-gram model builds upon the bag of words approach by considering n consecutive tokens
What are the advantages and disadvantages of Bag-of-Words model?
Advantages: Often perform at a high level of accuracy for tasks where the frequency or occurrence of words are predictive features
What is a Bag-of-Words Model?
A Bag-of-Words model is a class of models used for text mining tasks that is based on the frequency of words