The Future of Patent Automation
At Fenix.AI, our vision is to change the way patents are drafted across the world. To understand where we are, where we have been, and where we are going, it is useful to think of patent automation in terms of different stages. For example, you may be familiar with the 5 levels of autonomous vehicles:
Each of the levels will be described in terms of the impact on claims, specification, and drawings. So, without further ado:
Levels of Patent Automation
Level 0 — No Automation
Self-explanatory. A patent lawyer drafts claims and specification from scratch, and often pays a draftsman to do drawings.
Level 1 — Templates
- Template documents are used to speed up the process of creating new claims, specification, and drawings
- Templates include legal boilerplate to clarify commonly used terms
- Templates include pre-made styles to help format different portions of the document like headers, numbered paragraphs, claim numbering, etc.
- A library of pre-written paragraphs is available for use when drafting applications in familiar technology areas
- Stencils, icons, and basic diagram elements are used to speed up the process of creating simple figures
Level 2 — Claim Propagation
- Once an initial set of claims are drafted for one class (e.g., a method class), claims for other statutory classes are generated automatically (e.g., an apparatus with a processor and memory)
- A claim summary is generated automatically based on the claims
- Claims are evaluated based for antecedent basis
- Claims, summary, and other basic elements of the specification can be dynamically inserted into a template specification
- Content from a technology library can be automatically inserted into a template specification
- Acronyms can be automatically at specified intervals
- Components in specification can be automatically numbered
- Basic flowcharts are generated based on method claims
- Block diagrams are generated based on apparatus claims (may not show nesting of elements or relationships between components)
- May be limited to use in specific technology areas
Level 3 — Dynamic Description
- Relationships between components and functions is identified using natural language processing or a dynamic user interface
- Detailed description includes content describing which apparatus components perform various functions described in the claims
- Descriptive content can be imported from publicly available information (as opposed to just a private library)
- Block diagrams and flowcharts capable of illustrating relationships between components and steps (i.e., nesting of components or branching steps)
- Sufficiently flexible to accommodate any technology area
Level 4 — Intelligent Assistance
- Components and steps are automatically identified based on the disclosure
- Dependent claims can be generated based on independent claim
- Disclosure material is automatically edited to make it appropriate for use in a patent
- Basic drawings are created automatically based on disclosure material
- Black and white line drawings for different views can be obtained automatically based on provided photos or drawings
- Claims are automatically compared to relevant prior art using context aware natural language processing
Level 5 — Full Automation
- Claims, specification, and drawings are automatically generated based on invention disclosure
- Claims are automatically compared to prior art
- Additional support not found in disclosure is provided to satisfy written description and enablement requirements
Where We Are Now
Although there are a few patent attorneys remaining that make no attempt to systematize their efforts, I believe that the vast majority of us operate somewhere within Level 1. This is what I found when I started working as a patent attorney. The system I learned on included quite sophisticated set of templates and libraries.
Still, I got bored pretty quickly by the hours of cutting and pasting that were typically required once you had drafted an initial claim set. So, within about six months I had developed my first Level 2 system. Others have been working on similar projects for years. Here is an example of a patent for an early patent automation system.
Initially, my system worked from within a MS word document. It identified the claims and generated some additional content based on them. In fact, the first version I wrote was in Visual Basic, and was affectionately referred to as “The Macro” by colleagues who knew what I was doing.
Currently, the growing patent automation industry has mostly moved from Level 2 to Level 3. One of the key things to look for in a Level 3 system is identification of a correspondence between apparatus components and steps. Providing the system with information about which parts perform which functions is the first step toward generating more substantial content for the detailed description. Then instead of just repeating the claims, new sentences can be generated describing the function of each component.
Another thing to look for in a Level 3 system is whether system takes into account relationships among components and method steps. For example, can it identify that a dependent claim is describing substeps for a particular limitation of an independent claim instead of just adding additional steps on the end?
We haven’t really reached Level 4 yet, but some of these features are within reach, and they aren’t too far off. Level 4 basically represents the application of cutting edge AI capabilities to the patent automation context (i.e., instead of off-the-shelf NLP tools).
Level 5 is probably a ways off. Drafting claims is an art, and it might just require sophisticated artificial general intelligence to match the skill of an experienced patent attorney. And since a lot of the skill of a patent attorney goes into crafting claim 1, it will be a while before computers can outperform humans in this area. Of course, for some lower tier patents, outperforming humans might be the wrong standard.
So, given this framework for understanding the past, present, and future of patent automation, it is time to ask where you are in your own practice? And where do you want to be 2 years from now? Tools for Level 3 automation are becoming widely available. Those who insist on preparing every aspect of a patent manually will soon find it difficult to remain competitive.