Catching fraudsters with natural language processing and deep learning

How do you prevent users from sharing contact information? This is a desirable feature in many online applications due to security, safety, and even monetary issues; this is also a challenging problem because users employ creative strategies to circumvent traditional filters. For example, they will spell out phone numbers, use Unicode symbols resembling ASCII characters, invent new words that don’t exist in the dictionary, and communicate using foreign languages. Regular expressions and bag-of-words classifiers simply don’t work in this scenario, but deep learning does! We present character-level deep learning model that can solve this problem effectively and in a language-agnostic manner.

Manage a team of data scientists and engineers.