How Patent Filtering Works

Understanding the Filter System

Our filtering system helps you focus on the most relevant patents by analyzing:

How Matching Works

Filter Settings

The filter settings include the keywords and the weights as well as the exclusion words, you set them both yourself:

  • When you assign a keyword a weight(number) it is used in the equation (Keyword Weight x Match Multiplier)
  • The exclusion words are used to exclude patents that include those words that you set.
  • Match Multiplier are the amount of points that will be multiplied to your keyword weight that was set for that particular word.

Example:

Equation = Keyword Weight x Match Multiplier

Keyword: "Blockchain" → Weight = 4 (Max weight that a keyword can be set to is 5)

Exclusion Words(Multiple words must be seperated using a comma): Animal, Fish, Livestock, etc

Score = 23 (This is what you'll see in the results of a filter at the end of a patent, it is a relevance score and the higher it is the more relevant a patent is)

Match Multiplier: Exact Matches (Double Points)

An exact match is worth 2 points it occurs when:

  • Your keyword appears as a complete word in the patent title
  • It's not part of a longer word
  • Matches are case-insensitive ("AI" = "ai")

Example: Searching for "AI" (Weight: 3)

Title: "AI for medical diagnosis" → Exact match (3 × 2 = 6 points)

Title: "Painting with ai brushes" → Exact match (3 × 2 = 6 points)

Match Multiplier: Partial Matches (Standard Points)

A partial match is worth 1 point it occurs when:

  • Your keyword appears inside another word
  • It's not a standalone word
  • Partial matches prevents missed opportunities. It catches patents that use your keywords in compound words or with prefixes/suffixes
  • The advantage to this is that it provides a broader search to find patents you might miss however it can also result in a lot of irrelevant patents being shown so thats why they are only worth 1 point to ensure exact matches still rank higher. If you dont want this you can just put quotes between the keywords to be checked for only exact matches in the patents.

Example: Searching for "AI" (Weight: 3)

Title: "ai-enhanced → Partial match (3 × 1 = 3 points)

Exclusion Terms

Patents are completely removed if they contain:

  • Any of your excluded words (case-insensitive)
  • Matches anywhere in the title

Example: Excluding "drug"

Title: "AI for drug discovery" → Removed

Title: "AI for medical diagnosis" → Kept

How Results Are Ranked

  1. Filter out patents with excluded terms
  2. Score remaining patents based on your keywords
  3. Sort by highest score to lowest
  4. Show in three groups: Top 15, Top 30, and Top 50

Example Output:

  • "Artificial intelligence (AI)/machine learning (ML) tensor processor (ID: 12204456) - 2025-01-21 - Score: 33"
  • "Artificial intelligence (AI) building emergency guidance and advisement system (ID: 12205447) - 2025-01-21 - Score: 27"
  • "Enhanced AI image segmentation (ID: 12229967) - 2025-02-18 - Score: 15"

Pro Tips

For Precise Results

  • Use quotes for phrases: "machine learning" only matches the exact phrase
  • Higher weights (4-5) on your most important terms
  • List a diverse range of keywords depending on your requirements, the recommended is around 10. By doing this a wider range of scores can be shown and the most relevant patents can be much easily identified. If you do not do this the results of your filtering may cause all the patents to have similar scores which will in turn make it harder fro you to identify the most relevant patents.

For Broad Searches

  • Single words without quotes catch more variations
  • Moderate weights (2-3) give balanced results