Whatsapp Chat Analysis

A Jupyter Notebook with the analysis for a Whatsapp Chat using several techniques of data wrangling, EDA and Sentiment Analysis

🔰 How does it work?

Data is downloaded as a .txt file from Whatsapp App Without Media and loaded in the Jupyer Notebook

Data Wrangling

• Chat data is organized into three columns:
  • Complete Date
  • User
  • Message
• A Regex Function is used for several operations and create a Message_Modified column:
  • Lowercase the messages
  • Delete messages with URL
  • Remove numbers
  • Remove messages with spaces
  • Remove special characters
  • Remove repeated letters
• An Emoji column is created to identify all the emojis used in messages
• Complete Date information is used to create the following columns.
  • Date [YYYY-MM-DD]
  • Month (Name)
  • Day (Number 1-31)
  • Day (Week Name)
  • Hour This is what the final dataset looks like:

Exploratory Data Analysis (EDA)

Several questions were answered using the dataset information.

Who has sent the highest number of messages?

Messages interactions through time

Use of Emojis

Top Emojis User A [('😘', 7403), ('🥺', 1058), ('😬', 411), ('🤪', 277), ('❤', 266)]

Top Emojis User L [('🥺', 1378), ('😘', 1111), ('🥰', 912), ('☺', 622), ('😊', 587)]

Count of Words

WordCloud

Both Users

Both Users with Spanish Stop Words

A User

L User

Sentiment Analysis

Spanish Analysis

Sentiment score {Negative [0 : 0.33] | Neutral [0.33 : 0.66] ! Positive [0.66 : 1]}

Positive Words

Negative Words

The Spanish Sentiment Analysis didn't get good results. Now we want to try to perform a Sentiment Analysis with the Messages Translations to English

Sentiment Analysis Using Text Blob

Sentiment score {Negative [-1 : -0.33] | Neutral [-0.33 : 0.33] ! Positive [0.33 : 1]}

Positive Words

Negative Words

Sentiment Analysis Using NLTK

Sentiment score {Negative [0 : -0.05] | Neutral [-0.05 : 0.05] ! Positive [0.05 : 1]}

Positive Words

Negative Words