curate_tweets.py

import pandas as pd
import numpy as np

def curate_tweets(filename, col_id, col_tweet, num_rows, out_filename):
    df_in = pd.read_csv(filename, encoding='latin-1')

    rows = np.random.choice(df_in.index.values, num_rows)
    df_out = df_in.ix[rows]
    df_out = df_out[[col_id, col_tweet]]
    assert 'class' not in df_out.columns
    df_out['class'] = np.nan
    print('{} records, {} with unique tweet text'.format(len(df_out), len(df_out[col_tweet].unique())))
    df_out.reset_index(drop=True, inplace=True)
    
    i = 0
    end_i = len(df_out) - 1

    print("[0-9] for classes, b for back, s for save, x for exit")

    while i < end_i:
        try:
            print(df_out[col_tweet].iloc[i])
            resp = input("class --> ")
            if resp == "x":
                i = end_i
                tosave = input("Save? (y/n): ")
                if tosave in ["y", "Y", "yes", "YES", "Yes"]:
                    df_out.to_csv(out_filename+'.csv')     
            elif resp == 'b':
                i -= 1
                df_out.loc[i, 'class'] = np.nan
            elif resp == 's':
                df_out.to_csv(out_filename+'.csv') 
            elif 0 <= int(resp) <= 9:
                df_out.loc[i, 'class'] = int(resp)
                i += 1
      
            else:
                print ('*** INVALID ENTRY, TRY AGAIN ***')
        except:
            pass