Lowercasing and normalization in NLP - Model Metrics & Evaluation

Metrics & Evaluation - Lowercasing and normalization

Which metric matters for Lowercasing and normalization and WHY

Lowercasing and normalization help make text data consistent. This improves how well models understand words. The key metric to check is accuracy or F1 score on text classification or language tasks. Better normalization usually means higher accuracy because the model sees fewer confusing word forms.

Confusion matrix example

Imagine a text classifier before and after normalization. Here is a confusion matrix after normalization:

      | Predicted Positive | Predicted Negative
  ---------------------------------------------
  Actual Positive |        85 (TP)       |       15 (FN)
  Actual Negative |        10 (FP)       |       90 (TN)
    

From this, we calculate:

• Precision = 85 / (85 + 10) = 0.895
• Recall = 85 / (85 + 15) = 0.85
• F1 Score = 2 * (0.895 * 0.85) / (0.895 + 0.85) ≈ 0.872

Precision vs Recall tradeoff with examples

Lowercasing and normalization reduce errors from different word forms. This usually improves both precision and recall.

Example: Without normalization, the model might miss "Apple" vs "apple" as the same word, lowering recall. Or it might wrongly guess because of case differences, lowering precision.

Good normalization balances precision and recall, so the model finds most correct answers (high recall) and makes few wrong guesses (high precision).

What good vs bad metric values look like

Good: Accuracy or F1 score above 85% after normalization means the model understands text well.

Bad: Accuracy below 70% or big gaps between precision and recall show the model struggles with inconsistent text forms.

Common pitfalls in metrics

• Ignoring normalization impact: Metrics might look good on training but fail on new text with different cases or accents.
• Data leakage: If test data is normalized differently, metrics can be misleading.
• Overfitting: Model might memorize specific word forms instead of learning normalized patterns.
• Accuracy paradox: High accuracy can hide poor performance on rare words if normalization is inconsistent.

Self-check question

Your text classification model has 98% accuracy but only 12% recall on rare words after normalization. Is it good?

Answer: No. The model misses most rare words (low recall), which means it fails to recognize many important cases despite high overall accuracy. You should improve normalization or model training to catch more rare words.

Key Result

Lowercasing and normalization improve model accuracy and F1 by making text consistent, balancing precision and recall.

Practice

1. What is the main purpose of lowercasing text in Natural Language Processing?

easy

A. To translate text into another language

B. To make all letters small so words like 'Apple' and 'apple' are treated the same

C. To remove all punctuation marks from the text

D. To split sentences into words

Solution

1. Step 1: Understand what lowercasing does

   Lowercasing changes all letters in text to small letters.

2. Step 2: Understand why lowercasing is used

   This helps treat words like 'Apple' and 'apple' as the same word, improving consistency.

3. Final Answer:

   To make all letters small so words like 'Apple' and 'apple' are treated the same -> Option B

4. Quick Check:

   Lowercasing = uniform word form [OK]

Hint: Lowercase to treat same words equally [OK]

Common Mistakes:

• Confusing lowercasing with removing punctuation
• Thinking lowercasing translates text
• Believing lowercasing splits sentences

2. Which of the following Python code snippets correctly converts a string text to lowercase?

easy

A. text.lowercase()

B. lower(text)

C. text.toLowerCase()

D. text.lower()

Solution

1. Step 1: Recall Python string method for lowercasing

   Python strings have a method called lower() to convert text to lowercase.

2. Step 2: Check each option

   text.lower() uses text.lower(), which is correct. lower(text) is not a Python function. text.toLowerCase() is JavaScript style. text.lowercase() is not a valid method.

3. Final Answer:

   text.lower() -> Option D

4. Quick Check:

   Python lowercase method = lower() [OK]

Hint: Python lowercase method is .lower() [OK]

Common Mistakes:

• Using JavaScript syntax in Python
• Calling non-existent methods like lowercase()
• Trying to use a function named lower() instead of method

3. What will be the output of this Python code?

text = 'Café'
normalized = text.lower()
print(normalized)

medium

A. 'café'

B. 'cafe'

C. 'CAFÉ'

D. 'Cafe'

Solution

1. Step 1: Apply lower() method on the string 'Café'

   The lower() method converts all uppercase letters to lowercase but does not remove accents.

2. Step 2: Understand effect on accented characters

   The accented 'é' remains unchanged because lower() does not normalize accents.

3. Final Answer:

   'café' -> Option A

4. Quick Check:

   lower() keeps accents, just lowers letters [OK]

Hint: lower() changes case but keeps accents [OK]

Common Mistakes:

• Assuming accents are removed by lower()
• Expecting uppercase output
• Confusing normalization with lowercasing

4. The following code aims to lowercase and normalize text but has an error:

import unicodedata
text = 'Café'
normalized = unicodedata.normalize('NFKD', text).lower()
print(normalized)

What is the error and how to fix it?

medium

A. normalize returns a string with accents separated; fix by removing accents after normalization

B. Calling lower() before normalize; fix by swapping the calls

C. lower() returns a string; normalize expects bytes, fix by encoding first

D. No error; code works correctly

Solution

1. Step 1: Understand what normalize('NFKD') does

   It decomposes accented characters into base character plus accent marks.

2. Step 2: Check the code behavior

   After normalization, accents are separate characters, so lower() works but accents remain. To remove accents, you must filter out combining marks after normalization.

3. Final Answer:

   normalize returns a string with accents separated; fix by removing accents after normalization -> Option A

4. Quick Check:

   Normalization decomposes accents; remove them explicitly [OK]

Hint: Normalize then remove accents explicitly [OK]

Common Mistakes:

• Thinking lower() removes accents
• Swapping normalize and lower() calls incorrectly
• Assuming no extra step needed to remove accents

5. You want to preprocess text data by lowercasing and removing accents for a machine learning model. Which Python code snippet correctly does this?

hard

A. import unicodedata
text = 'Café'
text = unicodedata.normalize('NFKD', text)
print(text)

B. text = 'Café'
text = text.lower()
print(text)

C. import unicodedata
text = 'Café'
text = text.lower()
text = ''.join(c for c in unicodedata.normalize('NFKD', text) if not unicodedata.combining(c))
print(text)

D. text = 'Café'
text = text.upper()
print(text)

Solution

1. Step 1: Lowercase the text

   Use text.lower() to convert all letters to lowercase.

2. Step 2: Normalize and remove accents

   Use unicodedata.normalize('NFKD', text) to decompose accents, then remove combining characters to strip accents.

3. Step 3: Combine steps correctly

   import unicodedata text = 'Café' text = text.lower() text = ''.join(c for c in unicodedata.normalize('NFKD', text) if not unicodedata.combining(c)) print(text) does both steps properly: lowercasing first, then normalization and accent removal.

4. Final Answer:

   import unicodedata text = 'Café' text = text.lower() text = ''.join(c for c in unicodedata.normalize('NFKD', text) if not unicodedata.combining(c)) print(text) -> Option C

5. Quick Check:

   Lowercase + normalize + remove accents = clean text [OK]

Hint: Lowercase first, then normalize and remove accents [OK]

Common Mistakes:

• Skipping accent removal after normalization
• Using upper() instead of lower()
• Normalizing without removing combining characters