At Riahi Patents, we constantly follow the technological unfoldings of the industry so that we can deliver state-of-the-art services and exceptional experiences to our customers. To that end, we have developed a range of tools and automation solutions to streamline and optimize intellectual property protection processes. In addition to this, we are actively collaborating with patent search tool developers in order to help improve and advance their tools, and in turn, the industry as a whole. By working together, we hope to drive progress and innovation in the field of patent search tools, and continue to provide the highest level of service to our clients.
Given the current stage of tech development, there have been countless efforts to incorporate new technologies into patent searching, mainly through the use of artificial intelligence, machine learning, and neural networks. However, further analysis is fundamental, since one of the biggest challenges when it comes to adopting AI tools is to compare their efficiency – in terms of accuracy and time-efficiency – with some benchmark data provided by the non-AI techniques.
Two areas of the primary focus of our R&D efforts:
- Internal Process Improvements
- Patent Search Tools testing
Internal Process Improvements
At Riahi patents, we are dedicated to improving our internal processes and developing new tools and technologies to protect our customers’ data and ease the process for innovators and entrepreneurs. Our fully automated system allows for safe data collection and reporting, and we are constantly working to improve our systems to provide the most advanced and secure intellectual property protection experience.
Our online idea submission portal and search report submission portal make the process convenient and efficient, and our automatic report generation and cloud-based document management ensure that all necessary information is organized and easily accessible from the idea disclosure stage to the granting of patents. We prioritize our customers and their data above all else.
- Fully online Idea (Patent/Trademark) submission portal – allows innovators and entrepreneurs to easily and securely submit their ideas for patents and trademarks from the convenience of their time.
- Fully online Search report submission portal – enables analysts to easily and securely submit their search and analysis convenience of their time.
- Automatic Report generation – streamlines the process of gathering and organizing information, allowing for efficient and accurate reporting so that analysts can focus more on analysis rather than preparing the report.
- Cloud based document management system – organizes and stores all necessary documents and information from the idea disclosure stage to the granting of patents, making it easy for users to access and track the progress of their patent filings.
Patent Search Tools testing
Literature search and particularly patent searching has been used for more than five decades for patentability searches, validity searches, and freedom-to-operate. Like any industry, patent searching has been revolutionized over the past five decades.
Many traditional patent searchers talk about the days that they had to go through printed copies of the patents in the drawers called “shoeboxes”. These were very similar to the index cards being used at libraries. These boxes were organized by the art units (disciplines) and the patent searchers had to be experts to know where they could find the particular patent that they were looking for. When it came to mechanical parts, however, they needed to rely on their photographic memory. The searchers kept asking themselves the same questions – “I remember seeing something similar to this a few weeks ago, but I don’t remember where.”
When we are hundred percent relying on a mechanical procedure, there is no way to repeat the exact same process. However, let’s say the searchers were able to repeat the search procedure. Since it was being performed by a human being (from A to Z), it might contain some errors and some discrepancies were introduced between the two searches performed by the same patent analyst.
The first revolution happened when computers were introduced to almost every industry. The traditional databases (index cards, shoe boxes, etc.) were replaced by electronic databases. Everyone was amazed by how much space was saved by transitioning from the old-fashioned databases to computer-based databases. Within a few years, all the libraries adopted the new system and technology. Not only did they save space by this transition, but they also minimized the errors (duplicates, topics, etc.).
The next step for all the databases was to search through all the data using the computers. Immediately after, the entire community, which was relying on using different databases, found out how much time can be saved by using machine power. Moreover, significant amounts of error were eliminated using computers. Just imagine for a patent searcher who was just relying on traditional searching methods (going through the shoeboxes and index cards) all day long, chances of making a mistake were much higher than utilizing machine power for the mechanical work.
In the next couple of decades, patent databases were equipped with text searching, using Boolean functions, classification searching, and other searching techniques. More importantly, all the additional tools (such as highlighting tools, machine translation) helped the intellectual property (IP) community to access the prior art much faster. It brought such significant accuracy that made it much easier to search for all the relevant documents. The patent searchers were no longer limited to a particular geographical location and, within a few minutes, could access a prior art written in Japanese or German and could access the English translation in just a matter of seconds.
As the IP community got used to the new method of computer-based searching, something peculiar happened. Numerous search engines were developed around the world with different capabilities and each search engine was being used by a specific group of patent analysts and scientists based on their demands. Many IP firms were using multiple tools based on what kind of project they were working on. Based on experience, IP firms had found out that search tool “A” is powerful when it comes to searching foreign patents or search tool “B” does a tremendous job when validity searches are needed.
As we have noticed over the past decade, the amount of data we need to digest, when it comes to research and development projects, is beyond our expectations. More importantly, accuracy plays a huge role in narrowing down your search. Figuring out if an idea is novel or if you might have followed another inventor’s or assignee’s footsteps (unpurposefully) can put you in a delicate position. Even after the comprehensive search, if a document is missed, no one is going to accept that prior art was missed and slipped through the cracks. The response to missing documents is: “Times have changed. With all the access you had to the patent and non-patent databases, how come you have missed these documents?”
A significant part of patent search and analysis is performed by human beings. With all the patent search engines relying on text searching and Boolean functions, all the concepts (from the patent claims) need to be dissected into words (texts) and the text will be plugged into a search engine and the results need to be narrowed (and sorted) down using other search terms. Meanwhile, the relevant documents are being read, analyzed by the patent analyst, and all the found features and limitations need to be pieced together to find out if they satisfy the requirements of the proposed concept.
There is a great deal of work in the aforementioned process that is being repeated. Given the current stage of tech development, it is possible they can be performed by computers after some training, so they generate the needed information. This concept has been utilized in many fields for over ten years using Artificial Intelligence (AI) and its subsets, such as Machine learning, and Neural Networks.
By adopting mentioned technologies, the industry will be eliminating all the repetitive and routine work being handled by the analyst for dissecting and piecing the elements together and put that time and effort into training the algorithms associated with the patent searching. The analyst’s effort and expertise will be redirected to analyze the relevant documents and those documents will be used to train the search algorithms.
- One of the biggest challenges when it comes to adopting AI tools is to compare their efficiency – in terms of accuracy and time-efficiency – with some benchmark data provided by the non-AI techniques. Some of the questions that we are facing when analyzing the efficiency of the AI-powered patent search tools are as follows:
- Do you save time using AI tools?
- Do you increase the level of accuracy when using these tools compared to traditional and mechanical searching?
- Don’t you think that during the time you are training the AI tool, you may introduce some noise to the system and run into some irrelevant prior art? If yes, how do you avoid that?
- Can you use both text searching (using Boolean functions) and searching tools that are AI-based?
- Does this combination help you find better prior art?
- What are the risks associated with using AI tools as the primary search tools?
In order to provide answers to the aforementioned questions, detailed and comprehensive research work needs to be performed. Based on some preliminary research work performed on the AI patent search tools, it was revealed that AI tools can contribute significantly to the patent analysis field.
In the first phase of research and testing which was performed in the context of validating the use of AI-powered tools, we noticed, the average performance of AI platforms was similar to the performance of traditional patent search tools, when used as a stand-alone tool. However, in a posterior data analysis effort, we noticed that combining the results of traditional and AI-powered tools increased the accuracy and efficiency of the overall patent search report.
As shown previously, when used alone, the AI-powered tools performed slightly better than the traditional approach – i.e. using keywords in Boolean functions. However, the differences between the analyzed values were not significant – between the traditional tools and AI-powered tools – and didn’t have enough robustness to sustain our hypothesis.
However, the combination of the results coming from two different approaches (AI and traditional) and from different patent analysts provided better results – 20.85% – on average, than stand-alone tools. Three different combinations of tools (platforms) were used, as follows: traditional “A” + AI “B”; traditional “A” + AI “C”; and traditional “A” + AI “B” + AI “C”. After combining the result, the five most relevant references were selected. The results for the combination of the first ten cases can be requested from here
In continuation of the first phase of the research and testing to better understand the impact and effectiveness of AI-powered search tools in patent searching and analysis, we advanced our research in next phase.
The initial hypothesis of Riahi Patents was that standalone AI-powered search tools could potentially return better results, in prior-art search, than conventional searching (based on keyword searching and Boolean functions). However, it was previously found that, on average, the performance of AI platforms was similar to the performance of traditional search tools, when used as stand-alone tools. Also, the combination of the results improved the accuracy and efficiency of the patent search by 20.85% (on average) than stand-alone tools. Report available on request
From a scientific standpoint, it was not clear if this significant impact (20.85%) was due to the combination of search tools or due to the combination of inputs coming from different patent analysts. In order to verify this ambiguity, ten different cases and 40 prior art searches were carried on. During this new scientific effort, each case was searched and analyzed four times – testing all the possible combinations of tools, as described in the second section.
The second phase of the research, thus, intends to provide continuity to the first phase and answer the remaining doubts in addition to deepening the understanding regarding the power of combining the tools and crowdsourcing effect on patent searching. Therefore, the main goal of the second phase of our research effort was to analyze the data and verify if the good results were derived from (i) using multiple analysts, (ii) using a combination of tools, or (iii) combining both methods (i) and (ii). Additionally, this phase of the research provided some insights on IP analysts’ perception on using multiple patent search tools and optimizing the results while getting the patent search completed within the time frame provided to them.
Artificial intelligence has been introduced to patent searching and analysis for about a decade. Although numerous algorithms have been developed and utilized, there are multiple unknowns and intellectual property professionals look at AI (artificial intelligence) as a hit or miss tool for patent searching. There has been a cloud over the entire concept of utilizing AI methods for patent searching and the level of uncertainty has been at the same level.
We, at Riahi Patents, conducted two phases of detailed and comprehensive research on three patent search engines (one being conventional and two being AI based), and, during each phase of the research, many of the unknowns were resolved about the patent searching techniques and methods using AI tools. Moreover, when the results coming from each analyst were superimposed, the combination provided even higher ranking references. Granted that each AI-based patent search engine uses a different algorithm, the aforementioned results can not be extended to all the search tools.
Although our results have shown the best set of results come out of a combination of multiple methods and techniques, each patent search engine needs to be tested separately and the results can be analyzed to find out the potential shortcomings and shed the spotlight on the strength of the patent search engine.
In addition to all the aforementioned results coming out of the research performed by Riahi Patents, Inc. it needs to be brought to IP professional’s attention that using the best patent search engines can made a huge difference on the outcome of the search, but a unique method and system for patent search and how to utilize the conventional and AI-based tools will be an effective method to increase the accuracy of the patent searching at the same time that other relevant parameters are being optimized.
At Riahi Patents, we are dedicated to constantly improving and innovating our patent search process. To that end, we are currently working with a new AI-based patent search tool, which we are evaluating for efficiency using established methods. In addition to this, we are actively exploring new and advanced experimentation methods to further optimize and streamline our search process in the future. By constantly striving to improve and evolve our patent search capabilities, we hope to better serve the needs of our clients and partners.
If you are a developer of patent search tools, whether utilizing traditional Boolean methods or AI technology, we would love to have a conversation with you. As a company that is always looking for ways to improve and advance our patent search process, we are interested in hearing about your experiences and insights as a patent search developer. By collaborating and sharing knowledge, we hope to learn from one another and continue to push the boundaries of what is possible in the realm of patent research and analysis.