Problem Statement we would like to analyze the queries that the consumers of an FMCG company type and respond accordingly with relevant information.

1. find Various keywords that are primary concern of the consumers or seeking for information.
These can be put under various categories such as brand, product line and sub-product line,
usage, etc;
2. Examples of inputs (searched phrases) that the consumers may type in
3. Examples of formation of composite keywords (phrases) out of keywords and thus
enhancing the list of keywords (‘shoe polish’ out of ‘shoe’ and ‘polish’)
4. Examples of Synonyms
We need to create an optimal set of categories (buckets) in the range of 10 to 20 (e.g. brand,
product line, price, region, etc.) and an optimal set of sub-categories under each category (e.g.
wellness, skin care, makeup, etc under ‘product’ category; Lux, Hamam, Liril, Lakme, Lipton,
etc sub-categories under ‘brand’ category; cheap, expensive under ‘price’ category; etc.) out of
the Keywords; examples of the final attributes /keywords under a sub-category such as ‘loofah’,
‘soap’, ‘shower gel’ etc under ‘Lux’ sub-category. And you need to put all the key words in one
or many categories and/ or sub-categories. The keywords may belong to different categories or
sub-categories (e.g. polish can be under makeup ‘nail polish’ and also under accessories ‘shoe
polish’).
You may require applying more than one supervised and / or unsupervised learning
techniques for the same.

The issue names no files, tests, or entry points, so first map the repository's Python and NLP components before choosing an approach. Define the keyword, synonym, phrase, category, and sub-category outputs described in the issue, and consider how supervised and unsupervised learning would be evaluated against those requirements.

Written by the indexing model from the issue text.