Thursday 15:00–15:30 in Track 3

Use of vectorized text and siamese recurrent neural networks for Allegro offers clustering

Mikołaj Sędek, Rafał Wojdan

Audience level:
Intermediate

Description

We would like to present a data science approach to business case of clustering similar objects on Allegro platform. Our solution is using text embeddings (GloVe algorithm) as input to training supervised similarity models - siamese recurrent neural networks in Tensorflow (TF and Keras API). We will present results of our models and business application of our method.

Abstract

Subscribe to Receive PyData Updates

Subscribe

Tickets

Get Now