RiverText: A Python Library for Training and Evaluating Incremental Word Embeddings from Text Data Streams

RiverText is a Python library designed to address the limitations of traditional word embedding models by providing a comprehensive framework for training and evaluating incremental word embeddings from data streams. These incremental words embeddings allow for dynamic updating of word representations in response to evolving language patterns in sources such as social media and the web. The library implements popular techniques like Skip-gram, Continuous Bag of Words, and Word Context Matrix, using PyTorch as the backend for efficient neural network training. In addition, RiverText includes a module that adapts static intrinsic NLP evaluation tasks to a streaming setting. The open-source library, available at https://dccuchile.github.io/rivertext/, provides detailed documentation and examples to facilitate quick and easy adoption. RiverText is expected to be a valuable resource for researchers and practitioners working with large-scale streaming text data.