Building a Vietnamese Query Completion System Using Statistical Language Model

Automatically completing queries in the search function is considered particularly important in every ecommerce system. By automatically returning the top completed queries for users to choose from, it helps them quickly find information about the products that they want. However, the challenge for each system is the time to automatically complete each query and must be performed in real-time. Many previous studies using neural network language models lead to very high training costs, and the time to calculate the completion of queries is also a problem with these models. In this paper, we propose a simpler and more effective method for completing Vietnamese word level and query level based on statistical language models. In our experiments, we use dataset of queries in search logs and product titles of Vietnamese ecommerce platforms.