NLPatent's Product Features FAQ

Are there ways to send alerts to new patents being issued?

No, this feature is not available yet. However stay tuned for new product developments.

Does removing a result using the "Hide" button train the AI to refine results?

No, it does not. The button just removes the patent from your search and hides it from the list, but it doesn’t learn from the removal. 

Can you manually add prior art? How about non-patent literature (NPL)?  

Yes, both can be added to your results in NLPatent + be used to refine results. Please check the Manually Adding Prior Art article.

Can we just do keyword search?

Yes, when you start a project, simply select the “Keyword Only” option. 

Can I search for patents by excluding concepts?

Excluding specific inventive concepts is not a function of NLPatent at the moment. However, you can use the NOT function in the keyword filter to exclude specific keywords from your list.

Which patent jurisdictions do you support? 

We currently support WIPO, USPTO, Canada, EPO, Germany, UK, France, China, Japan, Korea, India and Australia.

Do you support different languages?

We display the non-English language text in the platform, but NLPatent only supports english language searching, so the natural language AI search and the boolean keyword search will only search within the english language specification of patent publications or the machine translations where patents were not published with an official english version.

Why are there so many patents from X country?

For specific technology areas, certain jurisdictions tend to publish more. (Ex Germany in Automotive, Japan with dyes, China in semiconductors.) NLPatent does not put any bias for any jurisdiction over another, so if you are seeing  a bunch of patents from a specific country in the results, it is that those patents are the most relevant as determined by our AI. If you are not interested in a specific jurisdiction, you can always remove them in the Jurisdiction Filters. 

How do filters work?

When filters are applied (whether they be bulk filters or keyword filters), it limits the patent universe that the AI searches. I.e. it will only search patents that meet a specific criteria such as being part of the USPTO, and containing the word “Semiconductor” anywhere in the document. Hence, when filters are applied it is normal to see the “number of searched patents” to shrink. For a more detailed explanation, check the Understanding your results page.

How do I include synonyms?

No need! Our AI understands natural language in a way that a human does- i.e. looks for synonyms when searching. However, ensure that you give the AI enough context when using a Natural Language Search- see our Best Practices for Natural Language Search Article.

How does relevance analysis work? Does it Hallucinate? 

Relevance analysis compares the query at hand to a patent document of interest. For patent number searches, it compares the documents feature by feature. For Natural language search, it will point out what features are similar to the query and what features are not. Relevance analysis uses a generative AI, which is different from our Search AI. We use a technique called retrieval augmented generation (RAG) to effectively eliminate hallucinations.

Can I manage users and who gets access to this tool in my organization?

Yes, you can, but only if you are the admin of your organization's NLPatent account. You can manage users in settings if you are the admin. Here's the help article for managing users.

Will the AI learn my preferences for my account or for each search project?

Everytime you start a new project, the AI is a blank slate. Any refinements that you do within the project will be kept constricted within that project. There is no communication from project to project both within a user’s account but also between users. 

Is there a word limit for the invention disclosure query? 

Yes, there is a word limit of 600 words for text prompts in NLPatent. This limitation is associated with the capacity of our model. To achieve the best outcomes, please read our best practices for Natural Language query article. 

Can we bypass the 600-word query limit?

The 600-word limit is a technical limitation related to the model's token capacity. Please refer to our best practices for Natural Language search article.

One workaround is you can add manually more free-form text data to your saved results to help the AI refine your search using the “Add Non-Patent Literature” feature.

Why can entire patents be used as inputs, but there’s a 600-word limit for description queries?

Inputting an entire patent as a query is a different process from using free-form text. NLPatent's database already contains representations of each patent's conceptual "meaning." When processing a patent document, the AI processes individual sections and combines the outputs to generate a comprehensive representation of the entire patent.

This enables efficient retrieval of relevant patent documents to any query. It's important to understand that this process is distinct from using free-text queries, where the 600-word limit applies. For patent-specific searches, users can provide patent numbers, allowing NLPatent to analyze and retrieve relevant documents in a different manner.