prosper.datareader.utils

General utilities for additional data management. Useful functions and classes for getting even more out of generic data sources.

NLTK Utilities

NLTK provides a platform for building NLP tools; granting easier access to cutting edge plain-language parsing.

NOTE: requires pip install prosperdatareader[nltk] to enable. Not downloaded by default to trim down optional libraries.

vader_sentiment()

Designed originally for use with stocks.news, vader_sentiment() allows for basic application of the Vader sentiment analyzer. This provides a basic breakdown of english speech on a [-1.0, 1.0] scale:

• positive
• negative
• neutral
• compound

import prosper.datareader.stocks as stocks
import prosper.datareader.utils as utils

## load a pandas.DataFrame with today's market news
## Expected columns: TODO
market_news_df = stocks.news.market_news_google(pretty=True)

## grade article titles
## Expected columns: TODO
market_news_df = utils.vader_sentiment(
    market_news_df, #DataFrame full of data
    'title'         #which column to grade
)

worst_article = market_news_df[max(market_news_df['neg'])]
bleakest_article = market_news_df[min(market_news_df['compound'])]
brightest_article = market_news_df[max(market_news_df['compound'])]
best_article = market_news_df[max(market_news_df['pos'])]

NOTE: for running vader_sentiment() over the same DataFrame twice, note that vader_columns is available. Requires strict equivalence for column count and order.

COLUMN_NAMES = ['neu', 'pos', 'compound', 'neg']

market_news_df = utils.vader_sentiment(
    market_news_df,
    'title'
    vader_columns=['title_neu', 'title_pos', 'title_compound', 'title_neg']
)

market_news_df = utils.vader_sentiment(
    market_news_df,
    'blurb',
    vader_columns=['blurb_neu', 'blurb_pos', 'blurb_compound', 'blurb_neg']
)