Skip to main content

Topic Modeling

Topic Metadata​

Pandas dataframe where each row corresponds to a unique topic. Metadata associated with each topic includes:

  • topic depth
  • a human-readable topic description (topic label)
  • identifying keywords that differentiate the topic from other topics
# Returns a Pandas df
print(map.topics.metadata)
      depth  topic_id  topic_depth_1           topic_description                                 topic_short_description  topic_depth_2  topic_depth_3
0     1     0         Women's Fashion (3)      women/tops/dress/sandals/womens/casual/shoes/p...  Women's Fashion (3)     NaN           NaN
1     1     1         Electronics (5)          USB/Bluetooth/iPhone/charging/Intel/cable/HDMI...  Electronics (5)         NaN           NaN
2     1     2         Jewelry Collection (2)   jewelry/IceCarats/Jewelry/Type/ICECARATS/Sterl... Jewelry Collection (2)   NaN           NaN
3     1     3         Phone Protector         phone/Galaxy/Samsung/dogs/Watch/protector/scre...   Phone Protector         NaN           NaN
4     1     4         Pool Supplies           Pool/pool/Floats/chair/Brand/Amazon/Lathe/floa...   Pool Supplies           NaN           NaN
...   ...   ...       ...                     ...                                                 ...                      ...           ...
605   3     507       Lighting Replacement    hose/garden/Garden/watering/Hose/ft/plants/Jet...   Garden Hose             Plumbing S... Garden Hose
606   3     508       Lighting Replacement    Rate/9930/gallons/207/months/38℃/125PSI/GPM/34...   Water Pump              Plumbing S... Water Pump
607   3     509       Lighting Replacement    NPT/¼/½/PSI/Pump/Straight/tire/pump/12V/Connec...   Tire Pump               Plumbing S... Tire Pump
608   3     510       Lighting Replacement    drain/Drain/sink/pipe/Sink/stopper/steel/toile...   Plumbing Fixtures       Plumbing S... Plumbing Fixtures
609   3     511       Lighting Replacement    shower/water/Shower/filter/solar/fountain/head...    Bathroom Essentials     Plumbing S... Bathroom Essentials
610 rows × 7 columns

Topic Hierarchy​

Learn more about your topic breakdown as a Python dictionary. What are the most general topics, and which sub-topics do they contain?

# map.topics.hierarchy is a dict

hierarchy = map.topics.hierarchy
print(f'Your depth 1 (most general) topics are: {hierarchy.keys()}')
Your depth 1 (most general) topics are: dict_keys([
  ("Women's Fashion (3)", 1), 
  ('Electronics (5)', 1), 
  ('Jewelry Collection (2)', 1), 
  ...
])

You can use higher-level topic keys to access lower-level topics in your hierarchy.

import random
# List the subtopics in a random top-level topic
random_topic_1 = random.choice(list(hierarchy.keys()))
print(f'The general topic {random_topic_1} contains subtopics {hierarchy[random_topic_1]}')
The general topic ('Footwear (14)', 2) contains subtopics [
  'Shoes (3)', 'Sandal', 'Sneaker Culture', ..., "Women's Sandals"
]

Topic Groups​

By providing a level of hierarchy, get a list of dictionaries where each item is a distinct topic at that level.

Keys for that topic include subtopics, subtopic_ids, topic_id, topic_short_description, topic_long_description, and datum_ids.

your_depth_level = 2
print(map.topics.group_by_topic(your_depth_level)[0])
{
  'subtopics': ['Miscellaneous (3)'], 
  'subtopic_ids': [87], 
  'topic_id': 16, 
  'topic_short_description': 'Audio Equipment (3)', 
  'topic_long_description': 'Bluetooth/speaker/Speaker/music/CarPlay/MP3/prevention/bluetooth/stereo/sound/karaoke/Loss/Radio/⭐/radio', 
  'datum_ids': {'61c', '/WM', 'Rsw', 'q6I', ...,  'AVjU'}
}