Use NLTK to filter feature place names

[mikedillion]

Disclaimer: this does take about 5 mins to run through the ___ rows

Just ran through the notebook! Though I'd throw in this suggestion using NLTK to filter the feature place names without mutating the original name and adding in the other words like 'valentine'

Aside: I've been watching your work in maptimelex! Awesome stuff! Keep maptime alive!

from nltk import Text
from nltk.tokenize import word_tokenize

def filter_words(row):
    match_words = ['Love', 'Valentine', 'Heart']
    feature_name_text = Text(word_tokenize(row['FEATURE_NAME']))
    for word in match_words:
        if feature_name_text.count(word) > 0:
            return True

    return False

love_df = data_in[data_in.apply(filter_words, axis=1)]]

[rgdonohue]

Yes, thanks Mike! Glad to have you still "here."

[rgdonohue]

Python packages still always a treat after all these years ... 😛

https://stackoverflow.com/questions/30822131/nltk-package-errors-punkt-and-pickle

[mikedillion]

Python packages still always a treat after all these years ... 😛

https://stackoverflow.com/questions/30822131/nltk-package-errors-punkt-and-pickle

Aw, jeez I always forget about that and take it for granted.

[rgdonohue]

Worked fine after installing the punkt model. Returned way fewer results ... Looks like it's not returning places like "Loveland" ...

Also thinking about that altitude attribute of the places ... could channel Steve Winwood "Higher Love" and lose the younger audience completely lol.