AML Intensity
Project Link on Github
Goal
to compute intensity of AML words per company
Imagine you web search a company and you wish to count the number of times blacklist,sanction,ofac,embargo appeared - irrespective of their variations like pluralized (blacklists), superlative form, or variations (ban/restrictions).
This mini pipeline could assist you.
On a high level it does the following
- Gets bing search result of a company
- Removes proper nouns/companies/people
- remove stopwords like a, an, um
- remove punctuation
- lemmatize best, better to equivalent words like good
- searches and increments counts for similar words
Input output example
Input: EMARAT KAVKAZ
Output: {'blacklist': 24, 'sanction': 12, 'ofac': 0, 'embargo': 11}
Tech used
- NLTK
> Stopwords
> Tokenization
> Part of speech tagging
- Textblob
- Gensim with google negative wordtovec text embeddings pretrained model
