17. What do we know thus far?
NN, NNS are nouns.
The most common nouns
together sound like it’s a
text related to something
technical: 'web', 'story',
'lot', 'site', ‘data', 'cycle',
'order', 'data', 'icons',
'users', 'dashboards'
The verbs that we pulled
(VB, VBG) sound typical by
themselves but are useful.
Most common verbs:
'have', 'presenting', 'be',
'pulling', 'want', 'know',
'tracking'
23. Push, Pull, Present
Abstract text was pushed
to PyGotham and stored
More realistic example
would be collecting click
tracking data from a web
frontend.
Abstract text was pulled
once, but constantly
pulled from to manipulate,
extract, and sift through.
Typical jobs like data
cleansing and sorting
would be in the pull
process.
Matplotlib and standard
Python library utils were
used to present
Whether presenting slides
or a Jupyter Notebook, the
medium to present does
not matter.
27. ◉ Searches are all education-focused
◉ Sellers want to rank higher to sell better
◉ We can’t make search changes to please everyone,
but we listen carefully to our community
◉ We A/B test a lot
TpT Search Notes
30. An example of exclusion search
◉ “apples bananas -pumpkins”
◉ Searches are education focused; suffixes are searched for (-
ed, -ing, -ly, etc.)
◉ Looking through historical searches of 519,512,359 we
found only a small fraction of searches that included a dash
34. Problem
Users are looking to click on the large file icon for a link, but it is just an
image and not a link.
Hypothesis
If we change out the large file icon,
users will click more on the Preview
button instead and have a better
user experience.
46. References
◉ Bird, Steven, Edward Loper and Ewan Klein
(2009), Natural Language Processing with Python.
O’Reilly Media Inc.
◉ stravalib: Python Strava API client
https://github.com/hozn/stravalib
◉ Examples seen in this talk http://bit.ly/29zQDBb
(https://github.com/drincruz/PyGotham-2016)