Table of Natural Language Generation Systems

Discover a comprehensive, curated overview of major natural language generation (NLG) systems — from early template-based engines to modern neural architectures — in our “Table of NLG Systems”. This resource gives researchers, practitioners and educators a single reference point for exploring the evolution, variety and application of NLG technology.

What you’ll find in the table

A structured list of NLG systems: including system names, key authors, start and end years, operative domains, and salient characteristics.
Coverage of both academic and commercial systems, spanning decades of NLG research and deployment.
Links or references (where available) to further documentation, source code, or publications for each system.
A format designed to support comparative analysis: enabling users to track trends such as from rule-based to statistical to neural NLG, domain-specific to open-domain, and monolithic to pipeline architectures.

Why this table matters

Historical perspective – See how NLG system design has evolved, how research emphasis shifted from templates and grammars to data-driven and neural methods.
Survey and benchmarking – For researchers planning to build or evaluate new NLG systems, the table provides a ready reference of prior work and established systems.
Teaching and course material – Useful for instructors who are designing modules on NLG: they can assign students to explore specific systems listed in the table and compare architectures, input/output formats and evaluation approaches.
Knowledge-base for system selection – Practitioners who need to select or evaluate NLG technologies can use the list to identify candidate systems, their domain fit, and maturity.

Key sections and features to highlight

System name & citation key: Provides the canonical name of each system, including bibliographic key for further lookup (e.g., via BibTeX).
Authors / Principal Investigators: Names of the researchers or organisations responsible for the system’s development.
Operational years: Start and end year (or ongoing) to help understand the system’s lifecycle and currency.
Domain / application area: Specifies whether the system was used for weather reports, business summarisation, dialogue generation, controlled language, ontology verbalisation, etc.
Input representations & architecture style: Whether the system used templates, grammar-based rules, statistical methods or neural network architectures.
Availability / links: Where applicable, pointers to source code, documentation, or demos.

How to use this table in your workflow

Filter by domain: If you’re working on e.g. data-to-text reporting or ontology verbalisation, scan the “Domain” column for matching systems and then review their architecture and output style.
Compare architectural trends: Observe shifts in the “Input representation & architecture” field to inform design decisions (for example: rule-based vs neural).
Identify maturity and stability: Systems with longer operational lifespans and multiple authors tend to have more robust documentation — useful if you are choosing a system to adapt or extend.
Integrate into teaching: Use the table as a baseline reading for students: ask them to pick one system, summarise its architecture and critique its applicability today.