Building Patent Family Graphs From Imperfect Data

Why patent family continuity visualization is harder than it looks

If you've ever tried to map out the continuity relationships for a patent family with more than a handful of members, you know the pain. Patent family continuity visualization should be straightforward, the information exists, but it's fragmented across multiple sources, each with its own conventions and blind spots.

Patent attorneys need to understand how applications in a family relate to each other. This information drives prosecution strategy, priority date analysis, and portfolio-level decision-making.

Building patent family continuity visualization

We built &AI's patent family visualization tool to pull from multiple sources and reconcile them into a single graph. Each source has strengths and limitations worth understanding.

USPTO

The USPTO's Open Data Portal API is our primary source for US continuity relationships. It provides structured data about parent-child relationships between applications, including relationship types (continuation, divisional, CIP) and filing dates.

The limitation: coverage begins on January 1, 2001. For any application filed before that date, the ODP API has no data. Given that many active patent families trace lineage back to the late 1990s or earlier, this is a meaningful gap.

International patent office databases

For pre-2001 US applications and international filings, we pull continuity data from databases that aggregate patent information across multiple patent offices worldwide. These sources provide the most comprehensive coverage for cross-jurisdictional family data, filling the gap that the ODP API leaves for older and non-US applications.

Application Data Sheet (ADS)

The ADS is filed by the applicant and contains the applicant's own declaration of continuity relationships. We parse the ADS of the target application to extract this local continuity information.

This is where things get interesting. The ADS reflects what the applicant claimed about the application's relationships at the time of filing, but ADS filings can contain errors, like wrong parent application numbers, missing relationships, or incorrect filing dates.

Recursive graph building

The core of our approach is recursive graph construction. The user provides a single application number, and the system walks every continuity relationship it finds to build the complete family graph.

The process works like this: given an input application number, we query our data sources for all known continuity relationships involving that application. For each related application we discover, we repeat the query. This continues until we've exhausted all reachable relationships. The result is a complete continuity map with every application in the family and every relationship between them, fully linked and typed.

ADS discrepancy detection

The ADS parsing layer deserves special attention because it addresses a data accuracy problem.

When we parse the ADS and compare it against the continuity graph built from external sources, we track four types of discrepancies: relationship existence, relationship type, filing dates, and application numbers.

In one case we found, an ADS listed provisional 61/976,445 when the correct provisional was 61/976,554 — a simple error that broke the priority chain for an entire branch of the family. Every application downstream of that incorrect provisional was pointing to a filing that didn't establish the priority they thought it did.

We don't modify our continuity graph based on ADS data. Instead, we cross-reference the ADS against what we find in external sources, and we log any discrepancies. These discrepancies represent potential filing errors that attorneys should be aware of. We're building toward surfacing these as contextual alerts, flagging potential issues for the attorney to investigate, not making automated corrections.

Try it yourself. 

&AI's patent family tool is live in the platform. Enter any US application or patent number and get the full family graph — every continuation, divisional, CIP, and PCT national phase entry, pulled from USPTO and international databases, reconciled automatically. Click any node to see patent details, switch to table view for filtering and bulk operations, or use the claims analysis view to track how claims evolved across the family.