accelerated-text/awesome-nlg

What it solves

Natural language generation (NLG) is the task of automatically producing human-readable text from data, dialogue, or other structured input. It is used in chatbots, story generators, data descriptions, and more, but the field is fragmented across many techniques and toolkits. This project collects the best available resources into one place so practitioners can find datasets, libraries, evaluation metrics, research papers, and ready-made products without searching across dozens of scattered sources.

How it works

The project is an 'awesome list' — a curated, community-maintained collection of links and descriptions. It is organized into categories such as Datasets, Dialog, Evaluation, Grammar, Libraries, Narrative Generation, Neural Natural Language Generation, Papers and Articles, Products, Realizers, Templating Languages, and Videos. Each entry links directly to the resource and includes a short description of what it does. The list is maintained by the open-source community, so it stays current as new tools and research appear.

Who it's for

This list is useful for researchers, developers, and students who need to build or evaluate NLG systems. It provides a quick way to discover online datasets (like the E2E or WebNLG benchmarks), generation libraries (such as SimpleNLG or Tracery), neural frameworks (like Transformers or textgenrnn), and evaluation tools (e.g., BLEURT or NLG-eval). It also connects to commercial products and educational videos, making it a one-stop entry point for anyone entering the field or seeking to expand their toolkit.

Highlights

• Covers the full NLG pipeline: datasets, training, generation, evaluation, and real-world applications.
• Includes both classical symbolic techniques (grammar-based realizers, templating languages) and modern neural approaches (GPT-2 based tools, Transformers).
• Offers links to open-source libraries and commercial products, showing the breadth of NLG in practice.
• Curates research papers and videos, giving a deep learning path from introductory talks to state-of-the-art methods.