PDF] Unsupervised Synonym Extraction for Document Enhancement in E

A two-phase unsupervised synonym discovery framework is proposed for extracting synonym rules from the search log data and it is demonstrated that 85.5% of the extracted rules are high quality pairs. The vocabulary gap between search queries and product descriptions is an important problem in modern e-commerce search engines. Most of the existing methods deal with the vocabulary gap issues by rewriting user-input queries. In this work, we describe another way to address vocabulary gap issues in the e-commerce search systems. In particular, we propose an unsupervised synonym extraction framework for document enhancement. Comparing to existing methods, the purposed framework has two main differences: 1) instead of extending and rewriting user-input queries, we make the synonym tokens searchable by adding them into the text descriptions of products; 2) the whole process is unsupervised and doesn’t require any human labels. A two-phase unsupervised synonym discovery framework is proposed for extracting synonym rules from the search log data. We demonstrate the effectiveness of our approach by two experiments: 1) we do online A/B testing experiments in multiple countries, which show significant improvements in key business metrics; 2) we conduct a human audit to evaluate the quality of the extracted synonym rules, which indicates that 85.5% of the extracted rules are high quality pairs.

Applied Sciences, Free Full-Text

PDF) SynoExtractor: A Novel Pipeline for Arabic Synonym Extraction Using Word2Vec Word Embeddings

Opinion mining for national security: techniques, domain applications, challenges and research opportunities, Journal of Big Data

Extraction and normalization of IR indexing terms and phrases in a highly inflectional language in: Journal of Greek Linguistics Volume 23 Issue 1 (2023)

Applied Sciences, Free Full-Text

General-use unsupervised keyword extraction model for keyword analysis - ScienceDirect

Dataset construction method of cross-lingual summarization based on filtering and text augmentation [PeerJ]

A detailed library perspective on nearly unsupervised information extraction workflows in digital libraries

Future Internet, Free Full-Text

Machine Learning: New Ideas and Tools in Environmental Science and Engineering

Unsupervised synonym Extraction - Speaker Deck

Multi Page Document Classification using NLP and ML, Doc2Vec