Towards Product Attributes Extraction in Indonesian e-Commerce Platform

Muhammad Rif’at, Rahmad Mahendra, Indra Budi, Haryo Akbarianto Wibowo


Product attribute extraction is an important task in e-commerce domain. Extracting pairs of attribute label and value from free-text product descriptions can be useful for many tasks, such as product matching, product categorization, faceted product search, and product recommendation. In this paper, we present a study of attribute extraction from Indonesiane-commerce product titles. We annotate 1,721 product titles with 16 attribute labels. We apply supervised learning technique using CRF algorithm. We propose combination of lexical, word embedding, and dictionary features to learn the attribute using joint extraction model. Our model achieves F1-measure 47.30% and 68.49% respectively for full match and partial match evaluation. Based on the experiment, we find that doing attributes extraction on more various number and diverse attributes simultaneously does not necessarily give worse result compared to extraction on less number of attributes.


Attributes extraction, e-commerce, product title, named-entity recognition, indonesian language

Full Text: PDF