TF-IDF (TfidfVectorizer) in NLP - Model Metrics & Evaluation

Metrics & Evaluation - TF-IDF (TfidfVectorizer)

Which metric matters for TF-IDF and WHY

TF-IDF is a way to turn text into numbers by showing how important words are in documents. It is not a model itself but a tool to prepare data for models like classifiers. So, the metrics that matter come from the model using TF-IDF features.

For example, if you use TF-IDF with a spam detector, you want to check precision and recall of the spam classifier. Precision tells you how many emails marked as spam really are spam. Recall tells you how many spam emails you caught.

TF-IDF helps the model by giving good features, but the final evaluation depends on the model's predictions and the task.

Confusion matrix example

Imagine a spam classifier using TF-IDF features. Here is a confusion matrix:

      | Predicted Spam | Predicted Not Spam |
      |----------------|--------------------|
      | True Spam: 80  | False Negative: 20 |
      | False Positive: 10 | True Not Spam: 90  |
    

From this:

• True Positives (TP) = 80 (correct spam detected)
• False Positives (FP) = 10 (good emails marked spam)
• False Negatives (FN) = 20 (spam missed)
• True Negatives (TN) = 90 (correct good emails)

Precision vs Recall tradeoff with examples

In spam detection using TF-IDF features:

• High Precision: Few good emails are wrongly marked as spam. Good for user experience.
• High Recall: Most spam emails are caught. Good for security.

Sometimes improving recall means lowering precision and vice versa. You must choose based on what matters more.

For example, if you want to avoid missing spam, focus on recall. If you want to avoid annoying users, focus on precision.

What good vs bad metric values look like

Good metrics for a spam classifier using TF-IDF might be:

• Precision: 0.9 or higher (90% of emails marked spam are really spam)
• Recall: 0.8 or higher (80% of spam emails caught)
• F1 Score: balances precision and recall, should be high (around 0.85+)

Bad metrics might be:

• Precision below 0.5 (many good emails wrongly marked spam)
• Recall below 0.5 (many spam emails missed)
• Accuracy can be misleading if data is unbalanced (e.g., 95% accuracy but model never detects spam)

Common pitfalls with metrics when using TF-IDF

• Accuracy paradox: If spam is rare, a model that always says "not spam" can have high accuracy but is useless.
• Data leakage: If test data leaks into training, metrics look too good but model fails in real life.
• Overfitting: Model fits training data perfectly but performs poorly on new data. Check metrics on separate test data.
• Ignoring class imbalance: If spam is rare, use precision, recall, and F1 instead of accuracy.

Self-check question

Your spam model using TF-IDF features has 98% accuracy but only 12% recall on spam emails. Is it good for production? Why or why not?

Answer: No, it is not good. The model misses 88% of spam emails (low recall), so many spam messages get through. High accuracy is misleading because most emails are not spam, so the model guesses "not spam" too often. You need to improve recall to catch more spam.

Key Result

TF-IDF is a feature tool; model metrics like precision and recall on tasks using TF-IDF features show true performance.

Practice

1. What does the TfidfVectorizer primarily do in text processing?

easy

A. It converts text into numbers reflecting word importance.

B. It translates text into another language.

C. It removes all punctuation from the text.

D. It counts the total number of characters in text.

Solution

1. Step 1: Understand the purpose of TfidfVectorizer

   TfidfVectorizer transforms text data into numerical values that represent how important each word is in the text.

2. Step 2: Compare options with this purpose

   Only It converts text into numbers reflecting word importance. describes converting text into numbers that reflect word importance, which matches the function of TfidfVectorizer.

3. Final Answer:

   It converts text into numbers reflecting word importance. -> Option A

4. Quick Check:

   TF-IDF = word importance numbers [OK]

Hint: TF-IDF = numbers showing word importance in text [OK]

Common Mistakes:

• Confusing TF-IDF with translation or punctuation removal
• Thinking TF-IDF counts characters instead of words
• Assuming TF-IDF just counts word frequency without weighting

2. Which of the following is the correct way to import TfidfVectorizer from scikit-learn?

easy

A. from sklearn.feature_extraction.text import TfidfVectorizer

B. import TfidfVectorizer from sklearn.text

C. from sklearn.text import TfidfVectorizer

D. import TfidfVectorizer from sklearn.feature_extraction

Solution

1. Step 1: Recall the correct module for TfidfVectorizer

   TfidfVectorizer is located in sklearn.feature_extraction.text module.

2. Step 2: Match the correct import syntax

   The correct Python import syntax is: from sklearn.feature_extraction.text import TfidfVectorizer, which matches from sklearn.feature_extraction.text import TfidfVectorizer.

3. Final Answer:

   from sklearn.feature_extraction.text import TfidfVectorizer -> Option A

4. Quick Check:

   Correct import path = from sklearn.feature_extraction.text import TfidfVectorizer [OK]

Hint: Remember sklearn.feature_extraction.text for TfidfVectorizer import [OK]

Common Mistakes:

• Using wrong module names like sklearn.text
• Incorrect import syntax order
• Trying to import from sklearn.feature_extraction without .text

3. What will be the shape of the output matrix after applying TfidfVectorizer on 3 documents with 5 unique words total?

medium

A. (5, 5)

B. (5, 3)

C. (3, 3)

D. (3, 5)

Solution

1. Step 1: Understand TfidfVectorizer output shape

   The output is a matrix where rows represent documents and columns represent unique words (features).

2. Step 2: Apply to given numbers

   With 3 documents and 5 unique words, the shape is (3, 5) -- 3 rows and 5 columns.

3. Final Answer:

   (3, 5) -> Option D

4. Quick Check:

   Output shape = (documents, unique words) = (3, 5) [OK]

Hint: Rows = documents, columns = unique words in TF-IDF matrix [OK]

Common Mistakes:

• Swapping rows and columns in output shape
• Confusing number of documents with number of words
• Assuming square matrix regardless of input

4. Given this code snippet, what is the error?

from sklearn.feature_extraction.text import TfidfVectorizer
texts = ['apple orange', 'orange banana', 'banana apple']
vectorizer = TfidfVectorizer()
X = vectorizer.fit_transform(texts)
print(X.shape)
print(vectorizer.get_feature_names())

medium

A. fit_transform() requires a list of integers, not strings

B. get_feature_names() is deprecated; should use get_feature_names_out()

C. TfidfVectorizer() needs a parameter specifying language

D. print(X.shape) will cause an error because X is not defined

Solution

1. Step 1: Check method usage for feature names

   In recent scikit-learn versions, get_feature_names() is deprecated and replaced by get_feature_names_out().

2. Step 2: Verify other code parts

   fit_transform() accepts list of strings, TfidfVectorizer() works without language parameter, and X is defined correctly.

3. Final Answer:

   get_feature_names() is deprecated; should use get_feature_names_out() -> Option B

4. Quick Check:

   Use get_feature_names_out() instead of deprecated get_feature_names() [OK]

Hint: Use get_feature_names_out() for feature names in new sklearn versions [OK]

Common Mistakes:

• Using deprecated get_feature_names() causing warnings or errors
• Thinking fit_transform() needs numeric input
• Assuming language parameter is mandatory

5. You want to ignore very common words like 'the' and 'is' when using TfidfVectorizer. Which parameter helps you do this effectively?

hard

A. lowercase=false

B. max_features=1000

C. stop_words='english'

D. norm=null

Solution

1. Step 1: Identify parameter for ignoring common words

   The stop_words parameter removes common words (stop words) like 'the', 'is', 'and'. Setting stop_words='english' removes English stop words.

2. Step 2: Check other parameters

   max_features limits number of features but doesn't remove stop words; lowercase controls case; norm controls normalization, none remove stop words.

3. Final Answer:

   stop_words='english' -> Option C

4. Quick Check:

   stop_words='english' removes common words [OK]

Hint: Use stop_words='english' to skip common words [OK]

Common Mistakes:

• Confusing max_features with stop words removal
• Not using stop_words parameter at all
• Thinking lowercase removes stop words