Text processing with regular expressions

Building Chatbots in Python

Alan Nichol

Co-founder and CTO, Rasa

Regular expressions

Match messages against known patterns
Extract key phrases
Transform sentences grammatically

The regex behind ELIZA

USER: "do you remember when you ate strawberries in the garden?"

ELIZA: "How could I forget when I ate strawberries in the garden?"

Pattern matching

import re

pattern = "do you remember .*"

message = "do you remember when you ate strawberries 
            in the garden"

match = re.search(pattern, message)

if match:
    print("string matches!")

string matches!

Extracting key phrases

import re
pattern = "if (.*)"

message = "what would happen if bots took over the world"

match = re.search(pattern, message)

match.group(0)

'what would happen if bots took over the world'

match.group(1)

'bots took over the world'

Grammatical transformation

import re

def swap_pronouns(phrase):
    if 'I' in phrase:
        return re.sub('I', 'you', phrase)
    if 'my' in phrase:
        return re.sub('my', 'your', phrase)

    else:
        return phrase

swap_pronouns("I walk my dog")

'You walk your dog'

Putting it all together

pattern = 'do you remember (.*)'
message = 'do you remember when you ate strawberries 
            in the garden'

phrase = pattern.search(pattern, message).group(1)
phrase

'when you ate strawberries in the garden'

response = choose_response(pattern)
response

'how could I forget {}'

Putting it all together

phrase = swap_pronouns(phrase)
phrase

'when I ate strawberries in the garden'

response.format(phrase)

'how could I forget when I ate strawberries in the garden'

Let's practice!

Building Chatbots in Python