Bucketize Patent documents for Landscape analysis in minutes
Patent landscape analysis is always a challenge to an IP analyst right from commencement of forming search strings to the preparation of report with insights. If we break down the task into different sections and measure them in terms of time consumed for understanding the technology, searching, and bucketizing/categorizing, patent landscape analysis will take most of the time allotted for the project. When we are performing a patent landscape analysis, we always want to have an equation where we should be getting all relevant patents for the technology. However, we should not miss any relevant patent, and at the same time we must keep in mind that there should not have many non-relevant results either. The balance between good recall and precision is a sign of good patent analyst.
After the patent analyst has finished the search string, he along with his team, will be ready to take up the individual set of results for the initial screening of the patents. Next task is to keep the patent in relevant bucket, so if there are around 5000 patent references to be checked, at least a five-member team will be required to complete the task in approximately 10 days. Once the bucketizing/categorization of patents is completed, real task of generating insights by preparing different graphs must be done, which can be completed by any Power BI or Tableau. In fact, these days many tools are available to provide you with insights.
Imagine, if you can do the task of categorizing the patent documents in minutes instead of waiting for 10 days, just think of how much time will you be saving? Which you can utilize to have better insights, and to do more projects. Moreover, you need not to worry about how many patent documents are relevant or non-relevant. So, is it possible by using AI to categorize the patents? The answer is ‘Yes’.
Categorizer is an AI based tool provided by Relecura which is used to bucketize the patents into different categories. The important point to notice here is that, if the patent is not relevant then it will place the patent into another category which will indicate it doesn’t belong to any relevant category.
AI based categorizer uses several machine learning algorithms and tune them according to the training set provided by the users. Now, there are few questions here: does AI based categorizer take care of appropriate keywords, classification codes, and relevancy of patents? The answer is certainly yes. Another question may arise to IP analyst that how he will be sure each time that the selected training set is the best, this part can be answered with the help of K-factor value which defines the relevancy of training set. K-factor helps us decide whether our training set is as relevant as we expected.
Suppose you are working on landscape of electric vehicles, which have around 5000 patent references and they need to be categorized or bucketized. The steps are simple, you must select some relevant patents or control patents in each category, and your client is interested in patents which belong to (a) Lithium-ion batteries (b) charging stations (c) Hybrid vehicles (d) electric traction motor, and (e) thermal cooling categories. Select some 15 to 20 relevant patents in each category which will be termed as training set for the categorizer and 5000 patent documents will be termed as patents to categorize. First, make an AI model by using training set and once the model is created use it to categorize the 5000 patent documents to be categorized in six different categories set by you. The patent documents which do not belong to these five categories will be categorized in the sixth one and will be termed as others.
The task which takes approximately 10 days to complete manually can be completed in minutes. So, you can focus on more important tasks of generating insights and providing better reports to your clients.
#Relecura #categorizer #patent #patent landscape