Harnessing Frequency and Language Features for Keyword Extraction on E-commerce Platforms

3rd International Conference on Automation, Control and Robotics Engineering

工业技术;计算机科学;无线电电子学

Khine, Chit^1 ; Nongpong, K.^1

Intelligent Systems Laboratory (ISL), Vincent Mary School of Science and Technology, Assumption University of Thailand, Bangkok

10240, Thailand^1

关键词: Document frequency;    E-commerce systems;    Internet resources;    Keyword extraction;    Language features;    Noun phrase;    Product descriptions;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/428/1/012021/pdf
DOI  :  10.1088/1757-899X/428/1/012021

来源: IOP

【 摘 要 】

Automatic keyword extraction has become essential with the growing number of internet resources. This method aims to extract quality keywords that are relevant to products in e-commerce platforms. The problem with e-commerce products is that there are wide ranges of categories in which a product can be in it. We propose a technique to improve the extraction of keywords from products information for the e-commerce systems by applying both frequency-based and language-based features. For frequency-based features, we consider the fact that products in the same category may have popular keywords which are different from other frequency-based features. For language-based features, different types of noun phrases are extracted and ranked accordingly. In this work, the proposed category-based document frequency is combined with the traditional TFIDF and noun phrases ranking. The approach is evaluated using product descriptions from Amazon. The results show that our approach performs better than the traditional TFIDF and RAKE by at least 10 percent on various categories of e-commerce products.