Go to file
Debdut Karmakar 6f361fa122
Update README.md
2020-08-19 11:58:28 +05:30
.github Working Py 2020-08-19 11:56:16 +05:30
assets Working Py 2020-08-19 11:56:16 +05:30
generation Working Py 2020-08-19 11:56:16 +05:30
names_dataset Working Py 2020-08-19 11:56:16 +05:30
.gitignore Working Py 2020-08-19 11:56:16 +05:30
LICENSE Working Py 2020-08-19 11:56:16 +05:30
README.md Update README.md 2020-08-19 11:58:28 +05:30
evaluate.py Working Py 2020-08-19 11:56:16 +05:30
main.py Working Py 2020-08-19 11:56:16 +05:30
setup.py Working Py 2020-08-19 11:56:16 +05:30

README.md

name-dataset

A First and Last Name Dataset

Features

  • ~160k first names
  • ~100k last names
  • Find Names in Texts
  • Hiigh Precision / Recall
  • Worldwide Names
echo -e "$(python main.py 'Brian is in the kitchen while Amanda is watching the TV on the sofa.\nThey are both waiting for Alfred to come.')"

How reliable is it?

Well, it depends if you are looking for a high recall or a high precision. For example, the word Rose can be either a name or a noun. If we include it in the list, then we increase the precision but we decrease the recall. And vice versa, if it's not in the list. The library checks that the word starts with a capital letter. In our case, we emphasize more on precision. So I would say the best use case here is to check whether it's a name or not based on a prior knowledge that the customer has submitted a name.

Here is an example on a (old) text: ALI BABA AND THE FORTY THIEVES.

License

I don't own the data obviously. It's fetched from the websites listed in:

https://github.com/philipperemy/name-dataset/blob/master/generation/generate.sh

So I guess the most strict software license should apply here.

Sources and References

Exhaustive list of all the possible websites. Not all are used since there is a lot of garbage in the lists.