How Popular Patent Search Databases Rank their Search Results

Pete Fenner, December 5 2023 - Patent Search

Unlike other patent search engines, Express-V4 Premium Search & Compare Databases use our novel Query-Focused, Scored Sentences method of ranking and summarizing each results document based on your highest Confidence Value Keywords that appear in that search document resulting from using those Keywords to search Google or your Input Test to search PQAI.

Our Premium Search & Compare feature allows you to see the highest Total Score patent documents containing the most high scoring sentences which contain multiple Keywords from your query. To understand the differences between our Express-V4 Results Rankings and how each Patent Search Database Ranks their search results, we offer this summary overview of Express-V4 and major public and subscription Patent Search services:

A. Search Real Fast Express-V4 Premium Search Results Ranking Process:

Our goal is to highlight your current Patent search results documents with high scoring sentences containing multiple of your search Keywords. Our method is inspired by the techniques of Query-Focused, Extractive Document Summarization. This means we identify the most important sentences that contain multiple Keywords in each patent document that matched your query. To accomplish this goal:

Each Express-V4 Premium Search & Compare Database downloads and scores all the sentences of up to 100 current Premium Google Patents and PQAI search results.
We then sum only those sentences that have a score above 100 to compute a document’s Total Score.
Based on each document’s Total Score, we sorts the resulting documents into Total Score Ranking order that usually is different from the order that they were returned from the Google or PQAI Databases.
The Scored Sentences of each document are displayed in that document’s Key Sentences display in the left column below the Ranked display of URL results documents. You Input Text’s scored Key Sentences are displayed in the right column beside the selected Key Sentences display.
You may then adjust that Sentence Score Threshold value to display a manageable number of the highest scoring sentences in a document for your analysis and comparison with your Scored Input Text.

Since Google Patents and PQAI and most other public Patent Search Databases do give the details of their Ranking Algorithms, we have queried both Google Bard and Bing Copilot to gain some understanding of each of their processes. We include both answers in cases where Bard and Bing Copilot contribute additional or complimentary information. Since we do not know the details of any one database’s ranking methods, these answers are presented to give general concepts.

Generative AI services are known to potentially provide inaccurate answers and Search Real Fast cannot verify the accuracy of these Generative AI answers:

B. What process does patents.google.com use to rank the order of their patent search results?

[Bard answer:] Patents.google.com uses a variety of factors to rank the order of its patent search results. These factors include:

Relevancy: Patents are ranked higher if they are more relevant to the search query. Relevancy is determined by the title, abstract, and full text of the patent, as well as the search terms used in the query.
Citation count: Patents with a higher citation count are ranked higher. A citation count is the number of other patents that have cited the patent in question. Citation count is a measure of the influence of a patent on subsequent inventions.• Publication date: Newer patents are ranked higher than older patents. This is because newer patents are more likely to be relevant to current technological developments.
Other factors that may be considered in ranking patents include the applicant, assignee, and classification code of the patent.
The specific algorithm used to rank patents on patents.google.com is not publicly disclosed. However, it is known that the algorithm is constantly being updated to improve the accuracy and relevance of the search results.
Here is a simplified overview of the patent ranking process on patents.google.com:
1. The Google Patents search engine receives a search query from the user.
2. The search engine identifies a set of patents that are relevant to the search query.
3. The search engine calculates a ranking score for each patent. The ranking score is based on the factors listed above, such as relevancy, citation count, and publication date.
4. The search engine sorts the patents in order of decreasing ranking score.
5. The search engine returns the top-ranked patents to the user.

It is important to note that the patent ranking process is complex and constantly evolving. There is no single factor that determines the ranking of a patent. Instead, the ranking is based on a combination of factors, including the factors listed above.

In an October 2019 Google Pandu Nayak wrote an “Understanding searches better than ever before” https://blog.google/products/search/search-language-understanding-bert/ which describe how Google is employing their open-source neural network-based technique for natural language processing (NLP) pre-training called Bidirectional Encoder Representations from Transformers, or as we call it–BERT, for short.

This breakthrough was the result of Google research on transformers: models that process words in relation to all the other words in a sentence, rather than one-by-one in order. BERT models can therefore consider the full context of a word by looking at the words that come before and after it—particularly useful for understanding the intent behind search queries.

Google has never announced or said they use BERT in processing Google Patent Search queries, but one can only assume they would employ their best search technology to parents as well as general search.

C. How does PQAI rank their patent search results?

[Bard answer:] PQAI patent search utilizes a multi-stage ranking process to effectively prioritize search results and deliver the most relevant patents to users. This process involves a combination of text-based similarity measures, patent citation analysis, and AI-powered relevance scoring.

Initial Ranking:

BM25 (Bag-of-Words Ranking): PQAI initially ranks search results using the BM25 algorithm, a widely used text retrieval method. BM25 calculates the relevance of each patent based on the frequency of search terms appearing in the patent’s title, abstract, and claims.
Patent Citation Analysis: PQAI incorporates patent citation data to further refine the ranking. Patents that cite frequently cited patents or are cited by numerous other patents are considered more influential and are given higher rankings.

AI-Powered Re-ranking:

BERT Embedding: PQAI converts both the generated patent text and the retrieved patent documents into BERT embeddings, which are vector representations of the text that capture semantic meaning.
AI-Based Relevance Scoring: PQAI utilizes a deep learning ranking system to compare the BERT embeddings of the generated patent text and the retrieved patent documents. The ranking system assigns a relevance score to each patent based on the similarity between its embedding and the embedding of the generated patent text.
PQAI has a web page that outlines a significant number of scholarly articles on using AI and Machine Learning for Patent search and analysis: https://projectpq.ai/top-ai-based-research-papers-on-prior-art-search/

Final Ranking: PQAI combines the BM25 scores, patent citation scores, and AI-based relevance scores to produce the final ranking of search results. This multi-stage process ensures that the most relevant patents are presented at the top of the search results, providing users with efficient access to valuable prior art information.

D. How does freepatentsonline.com rank their patent search results?

[Bard answer:] FreePatentsOnline.com ranks its patent search results using a combination of factors, including:

Relevancy: The system looks for patents that contain the search terms in the title, abstract, claims, and other important fields.
Date: The system also considers the date of the patent, with newer patents generally ranking higher than older ones. This is because newer patents are more likely to be relevant to current technology and trends.
Popularity: FreePatentsOnline.com also takes into account how popular a patent is, based on how often it has been viewed and shared. This helps to ensure that the most popular and relevant patents are ranked higher.
In addition to these factors, FreePatentsOnline.com also uses a variety of other signals to rank its search results, such as the number of citations a patent has received and the quality of the patent’s inventors and assignees.
FreePatentsOnline.com also offers an “Expert Search” feature that allows users to specify additional criteria for their search results, such as the patent’s classification, examiner name, and assignee address. This can be helpful for users who are looking for very specific patents.
Overall, FreePatentsOnline.com uses a sophisticated algorithm to rank its patent search results in order to provide users with the most relevant and useful information.

E. How does LENS.org rank their patent search results?

[Bard answer:] LENS.org ranks its patent search results using a relevance ranking algorithm that takes into account a variety of factors, including:

The number of times the search terms appear in the patent document

The location of the search terms in the patent document (e.g., title, abstract, claims, description.) The weight of the different fields in the patent document (e.g., claims are generally weighted more heavily than the description)
The number of other patents that cite the patent
The quality of the patent’s inventors and assignees. LENS.org also uses a variety of other signals to rank its search results, such as the patent’s classification and examiner name.

In addition to the relevance ranking, LENS.org also offers a variety of sorting options, such as by date published, filing date, and number of citations. This allows users to customize their search results to their specific needs. LENS.org relevance ranking algorithm is constantly being updated and improved to ensure that users are presented with the most relevant and useful patent search results. Here are some additional details about the In4M metric that LENS.org uses to rank patent search results:

In4M stands for “International Innovation and Industry influence Mapping metric”.
It is a new metric that tracks the influence of scholarship on industry and innovation.
It is based on the knowledge of the scholarly works cited in the global patent literature and the signaled value of the citing patents as perceived by the patent applicants.
The In4M metric is used to rank institutions based on their influence on industry and innovation. It is also used to rank patent search results, with patents that are cited by more influential institutions ranking higher.
LENS.org use of the In4M metric is unique and helps to ensure that its patent search results are ranked based on their relevance and impact.

F. How does PatSnap rank their Patent Search results?

[Bard answer:] PatSnap ranks its patent search results using a combination of factors, including:

Relevancy: PatSnap uses a variety of natural language processing (NLP) techniques to identify the most relevant patents, including matching search terms to the title, abstract, claims, and description of each patent.
Citations: PatSnap also considers the number of citations a patent has received, as this is a good indicator of its importance and influence.
Freshness: PatSnap also ranks newer patents higher than older ones, as they are more likely to be relevant to current technology and trends.
Expert curation: PatSnap also uses a team of experts to manually curate and review its patent search results, ensuring that the most relevant and important patents are ranked at the top.
In addition to these factors, PatSnap also offers a variety of other features that allow users to customize their search results, such as:
Advanced search operators: PatSnap offers a variety of advanced search operators that allow users to refine their searches and find the most relevant patents.
Filters: PatSnap also allows users to filter their search results by a variety of criteria, such as patent type, jurisdiction, and publication date.
Clustering: PatSnap can also cluster search results into groups of similar patents, making it easier for users to identify and understand relevant patents.
Overall, PatSnap uses a sophisticated algorithm to rank its patent search results in order to provide users with the most relevant and useful information.
Additional details about PatSnap’s ranking algorithm:
PatSnap’s ranking algorithm is constantly being updated and improved to ensure that users are presented with the most relevant and useful patent search results.
PatSnap also uses a variety of machine learning techniques to improve the accuracy of its ranking algorithm.
PatSnap also takes into account user feedback when ranking its patent search results.

[Bing Copilot answer:] PatSnap, a leading provider of research and development (R&D) analytics, employs a combination of AI-driven techniques and proprietary algorithms to rank patent search results. Here are some key aspects of their ranking process:

Company Ranking Algorithm: PatSnap’s Discovery platform utilizes a company ranking algorithm to understand the players in a market. It benchmarks companies based on their technological strength.This algorithm helps identify market leaders and provides insights into their competitive positioning¹.
Semantic Search and Similarity Score: PatSnap’s semantic search algorithm allows users to compare patent claims and descriptions.
It calculates a similarity score between patent applications of the same family. By analyzing abstracts, images, and PDFs, it helps users assess the extent of similarity between patents² ³.
Valuation and Index Analysis: PatSnap uses AI and machine learning to create unique index analyses for every patent an

G. How does espacenet Patent Search rank results patents?

Espacenet, a valuable resource for patent search, provides free access to millions of patent documents. Here’s how it operates:

Coverage and Features:

Worldwide Database: Espacenet covers information about inventions and technical developments from 1782 to today.
Search Features: It offers powerful search capabilities to explore patent data.

Relevance Ranking:

Espacenet does not explicitly disclose a specific ranking algorithm for search results.
By default, results are sorted by relevance. The most relevant patents related to your query appear first.

Bibliographic Data View:

When you find relevant patent documents, you can go back to the Bibliographic data view.
Select the corresponding CPC (Cooperative Patent Classification) and/or IPC (International Patent Classification) classification¹.

Grouped per Family:

The result list is grouped per patent family.
This grouping helps you navigate related patents efficiently² ³.

Access Options:

Espacenet offers both the classic and open versions for exploring patent data⁴.
Remember that while the exact ranking criteria remain undisclosed, Espacenet relevance-based sorting ensures you access relevant patent information for your search needs.

Learn more: 1. epo.org2. worldwide.espacenet.com3. worldwide.espacenet.com4. epo.org5. worldwide.espacenet.com+2 more

I hope this article has given you some ideas as to how various Patent Search Databases rank the order of their search results which they return to you, their users.

Pete Fenner, President