action_detection_in_email

A heuristics-based linguistic model for detecting actionable items from the email. A rule-based model to classify sentences to actionable sentence and non-actionable sentence.

• Learns common grammer rules to catch action/transition verbs in the mail and prints if ENRON dataset mails are actionable or non-actionable.
• If mail is found actionable, it grabs the action part of the mail.

Requiremments :

• To install dependencies run : pip install -r requirements.txt
  • Spacy is using "en_core_web_sm" model so download it manually by running python -m spacy download en_core_web_sm in separate bash.

How to run the module

• Download the dataset here.
• To run to classify a sentence : python3 main.py "Sentence to classify"
• Simple run to classify emails in ENRON dataset : python3main.py /Path/to/data/file/ number_1 number_2 number_3
  • number_1 : Maximum no of chunks to run (to prevent system running out of memory)
  • number_2 : Chunk size to take for processing
  • number_3 : No of email to predict in each chunk

Needs for improvement

• Increase the no of verb action words to increase the rule based classification.
• Add 2-gram and 3-gram words to boot the accuracy of model.
• Importing NLTK wordnet model to fetch and add synonyms of transition verbss