VizSeq is a research toolkit for natural language generation (translation, captioning, summarization, etc.). It provides a collection of accelerated scorers, visualization in Jupyter Notebook/Web App, and seamless integration with fairseq.
VizSeq supports a wide range of natural language generation tasks (e.g. machine translation, summarization, image captioning, speech recognition, and video description), where model outputs aren’t easy to inspect with the naked eye. It is designed to be a productive tool for model evaluation and error analysis at scale. Main features:
Full coverage of common n-gram-based and embedding-based metrics
Jupyter Notebook / Web App UI for all-in-one-place visualization
pip install vizseq
Use accelerated scores to to evaluate model outputs
Analyze model outputs in Jupyter Notebook / Web App with visualization and fairseq integration