Generally speaking there are two approaches to using text for searching patents; traditionalists are accustomed to using Boolean operators, and collections of synonyms, often referred to as hedges, to conduct what’s called a keyword search. In the past decade, natural language processing systems have been applied to searching patents, and a variety of information retrieval systems that incorporate written text have made appearances in the patent space. Recently, I was made aware of a new machine intelligence based systems for searching patents called AI Patents.
With more than one million patent applications filed every year, searching for prior-art has become a daunting task. This constitutes an important challenge for technology companies and their legal representatives, as the value of its assets depends on their ability to demonstrate the novelty of their inventive efforts. Failing to identify prior-art makes it difficult for companies to patent their inventions and exposes them to costly litigation. The breadth and complexity of the IP space makes it very difficult, and time intensive to search for prior-art without the help of machine-based intelligence that identifies relationships between a new invention and those described in millions of patent documents.
Existing solutions fail to take into account that companies often, and are strongly motivated to, use different words to describe similar inventions. This makes search efforts based on the similarities between words prone to miss relevant prior-art. Additionally, existing techniques do not account for temporal changes in the terminology used to describe particular inventions. This is not a trivial omission as, by definition, the search for prior-art requires comparing an invention with other produced at different points in time.
AI Patents developed a conceptual search engine that addresses the challenge that the same idea can be described in multiple ways. For example, a coffee cup can be described as a mug, or as a container that holds liquid. The traditional way of searching forces the user to select small number of keywords, and to deeply understand the alternative keywords that could have been used to describe the same idea. Often times, inventors describe their inventions using “novel” terms to mitigate textual overlap with past invention and reduce the likelihood that patent application is rejected by the patent examiner. To address this challenge, AI Patents developed a search process that “learns” from thousands of patent examination decisions. These decisions determine that two distinct inventions describe the same scientific idea, even though their exact wording differs.
Based on this learning algorithm, AI Patents allows its users to compare patent documents based on their underlying scientific ideas, and not merely based on their textual overlap. For example, when users input the word “encode” into their search query, they are offered additional keywords which other inventors have used in the past when describing identical ideas (whereas, identical ideas are determined by the patent examiners). In the image below the synonyms that are offered include the words “gene” and “protein”.
This happens because in the learning process they do not limit and compare documents from the same International Patent Classification (IPC) only.
Another example would be using their database to identify acronyms. If the user runs the query DDR*, the system will ask their approval for attributing DDR to Double Data Rate. Once accepted this relationship will find additional useful terms associated with DDR. This is illustrated in the image below:
How to use AI Patents?
AI Patents offers a free form text search, you can copy the claims, or abstract of a patent, an invention disclosure, ideas written in your own words, or products description into the search box. Based on this input, the system will generate a collection of results sorted by relevancy. Results will present multiple paths to the user, since the belief is that no machine can replace human judgement. The system enables the user to do this in a quick, but cautious way. The user, by going over the first 20-50 titles, will be able to understand the different paths, and can filter the ones that are not relevant to them. This part in the process is also important – for making sure that critical patents are not missed.
The system provides the user with a tool that allows them to:
- Reduce the time of running the search and retrieving interesting prior art significantly
- Uncover relevant patents that she will miss using key words search (whither because these patents come from a different IPC, thus using different terminology, or since it was just described differently)
In the image below an example of traditional searching compared to what can be done with AI Patents is provided.
In summary, taking advantage of the experiences of ex-patent examiners AI Patents has built a systems for using their search engine to a get good understanding on what exists in the collection, and reveals patents that others miss. Users can get to this point by using one of the following approaches:
- Using the first claim of the patent, or multiple claims
- Using the abstract of the patent
- Using the invention disclosure
- Writing the invention in their own words, as if you described it to another person
These iterations are very quick, and the user doesn’t need to break ideas into keywords and complex Boolean, and proximity operators to convey an idea. The user can just copy the text as is. Results are sorted by relevancy thus reducing the complexity, and need to potentially read thousands of patents.
This approach, in a very short time, can assist the user to better understand the space of the invention.
Interested users can contact AI Patents by visiting their website at aipatents.com.