How Amplified projects work
This article is a complete overview of Amplified projects. You can watch the video below for a demo or scroll through the page if you prefer screenshots and reading.
Searching in Amplified is organized into Projects. The left side contains all of your inputs. These define the focus of the project. Search results appear on the right.
How search works
Searching in Projects works by combining AI similarity sorting with Boolean filtering.
This is important to give you clear control over results. Simple keyword search is easy to drive but too noisy. Amplified's AI sorting prioritizes similar patents without ever excluding results. Combining the two makes search simple, intuitive, and highly effective.
Project Scope settings
Click Edit to change the Project Scope settings. The scope statement will update automatically.
Controlling similarity for ranking
Amplified understands both conceptual and semantic similarity. Mark results as relevant to find more conceptually similar patents. Turn learning off to prefer semantic similarity to your text only.
Using keywords and other filters to refine and focus results
Amplified includes filters for class codes, assignees, inventors, and more but the most important are keywords. Keywords serve two purposes.
With filters off, keywords help you quickly skim and understand result sets by automatically extracting relevant passages from each result's full text and highlighting terms.
With filters on, keywords are used for search. This lets you steer your result list by specifying critical aspects that must be included.
The best way to use keyword filters in Amplified
Start with precise keywords and don't worry too much about being comprehensive. It's much more efficient to find a relevant patent first and then expand your search. Amplified includes tools like similar term suggestions and learning from relevant patents that make this super easy.
Controlling AI similarity sorting and learning from relevant patents
How Amplified's AI works: Results are ranked by overall similarity. Amplified's AI uses a custom-built language model that we designed and trained specifically to learn patent conceptual similarity. Our model is trained using citations, classification codes, and the full text of global patents in 29 languages.
Searching with text only: When you search with text only Amplified will strongly prefer patents that use similar terms. This doesn't exclude patents which don't have similar terms but they may be ranked lower.
Learning from relevant results: Saving relevant patents helps Amplified also find conceptually similar patents even if they don't use similar terms. You click under Learning from to turn this learning off and go back to text-only similarity.
How results are displayed
The search results page is designed for quickly scanning results to help you decide which are worth diving into and iteratively search.
Patent result cards
Results are organized into cards that contain selected information to help you judge relevance. You can group results by publication or simple family. The default is family grouping, so you'll see information from a representative family member.
Each card shows the patent's similarity rank, title, patent number, priority date, publication date, assignee, and inventor. In addition, Amplified has a number of value-added fields and badges.
You can toggle additional information in these cards with the checkboxes at the top of the screen. The Relevant text option uses your keywords to automatically find and display relevant sections of text from the patent's description and claims. Click on the patent's title to open it's full text page in a new tab.
Try adding a keyword with Relevant text on!
Relevant text and keyword counts update in real-time so when you add a new keyword, you'll instantly see which results have that keyword and where it shows up in each patent.
How to review and organize results
Keep track of your review by marking results with the Relevant, Flagged, and Hidden buttons. You can change this after making a selection. There's also a Status button which can be assigned various colors when more nuanced review is needed. Reviewed status information is unique to each project.
You can also add notes, tags, and annotations to results. This information will be visible in future projects as part of your knowledge base.
If a patent has a note, the notes button will be gold and include a link to the project where the note was added so you can easily see the past context. Annotations can be added from the result's full text page.
Patent full text page and annotations
When you open a patent's full text page in a project, the header will have navigation controls so you can move to the next or previous patent in your result list.
Navigation hotkeys
Use the arrow keys on your keyboard or click on the arrow buttons to navigate to the previous < or next > patent in your list. You can also mark patents as Saved z , Flagged x , or Hidden c . Drawings in the side-by-side view have keyboard shortcuts as well.
On the full text page you can highlight text to add annotations, copy text to the project chat, or ask our AI to explain what things mean. Annotations with a brain icon mean that it was generated by AI not added by a human.
Projects have a Share button to invite collaborators and Discuss button to chat with them. All Amplified teams allow for free Viewer users who add annotations and join discussions in the projects they are invited to.
Everything is integrated with the project and patent information to make it easy to share context and collaborate. When you @mention a user they'll get a notification.
