nlp-coronovirus-tweets-proposal.md

NLP | Project Proposal

Coronavirus Tweets

Question

• What is the question behind your analysis or model and what practical impact will your work have?
  • What were Americans tweeting about coronavirus and COVID-19 in April 2020?
• Who is your client and how will that client benefits from exploring this question or building this model/system?
  • To better inform future pandemic communication strategy, the Centers for Disease Control and Prevention want to understand what type of information spread via Twitter early on in the COVID-19 pandemic in the United States.

Data Description

• What dataset(s) do you plan to use, and how will you obtain the data?
  • Coronavirus COVID-19 Tweets:
    • early April
    • late April
• What is an individual sample/unit of analysis in this project?
  • One tweet
• What characteristics/features do you expect to work with?
  • Tweet characteristics:
    • created_at
    • screen_name
    • text
    • country_code
    • account_lang
    • verified
    • lang(uage)
  • Filtered for:
    • country_code = US
    • lang(uage) = English
• If modeling, what will you predict as your target?
  • N/A

Tools

• How do you intend to meet the tools requirement of the project?
  • Python, numpy, pandas, sklearn,
  • NLTK, VaderSentiment
  • Matplotlib, Seaborn
• Are you planning in advance to need or use additional tools beyond those required?
  • None at this time.

MVP Goal

• What would a minimum viable product (MVP) look like for this project?
  • Exploratory data analysis
  • Preprocess text data
    • Define corpus-specific stop words
    • Remove punctuation, digits, special characters, convert to lowercase
    • Stemming/lemmatization, parts of speech, named entity recognition
    • Vectorize
  • Dimension reduction/topic modeling
    • VADER Sentiment Analysis
    • Latent Semantic Analysis
    • Non-Negative Matrix Factorization