1_RegEX in NLP.md

Regular Expressions (RegEx) are a sequence of characters that form a search pattern. In NLP, before feeding text to complex machine learning models, you must clean and preprocess the raw data. RegEx is your primary tool for pattern matching, text extraction, and text cleaning.

Implementing RegEx

Using \d

By itself, \d only matches one single digit at a time.

import re

text = "I bought 3 apples and 12 bananas."

# Find every single digit individually
single_digits = re.findall(r'\d', text)
print(single_digits)
# Output: ['3', '1', '2']

import re

text = "Patient ID 8472 was admitted in 2023 and discharged in 2026."

# \d+ matches ONE OR MORE digits in a row (extracts whole numbers)
all_numbers = re.findall(r'\d+', text)
print(all_numbers)
# Output: ['8472', '2023', '2026']

# \d{4} matches EXACTLY FOUR digits in a row (great for years)
years = re.findall(r'\d{4}', text)
print(years)
# Output: ['8472', '2023', '2026']

---

Using ()

The () is used to match everything present in it It has matched all instances containing 3 digits ![[RegEx_Img2.png]]

---

Using \ (Escape Character)

RegEx has a lot of "meta-characters" that perform special functions: . * + ? [ ] ( ) { } ^ $. But what if you actually want to find a literal question mark in a sentence? Or a literal period? If you just type ? or ., the RegEx engine will try to execute a command. To tell the engine "treat this as a normal punctuation mark," you escape it by putting a \ right in front of it. ![[RegEx_Img3.png]]

• Incorrect Pattern: (\d{3})-\d{3}-\d{4}

  • Result: Captures "123" in a group, but expects the text to just say "123-567-8912". It will fail to match (123).

• Correct Pattern: \(\d{3}\)-\d{3}-\d{4}

  • How it works: \( looks for a literal opening parenthesis. \d{3} looks for 3 numbers. \) looks for a literal closing parenthesis.

---

Matching a range of characters using []

![[RegEx_Img4.png]]

---

Using the ^ operator

![[RegEx_Img5.png]]

---