Building an AI-Powered Writing Assistant for WPS Office

What this post added

This post details the integration of Milvus into WPS Office's AI-powered writing assistant. It describes the process of extracting features from unstructured text using TFIDF and a bi-directional LSTM-CNNs-CRF deep learning model, creating sentence embeddings with Infersent, and then storing and querying these embeddings in Milvus. Specifically, it highlights the use of the IVF_FLAT index and Milvus's partitioning function to improve query performance for sentence embeddings, achieving average query times of 0.2 seconds.