Relecura Adds a No-Code AI Patent Classification Tool to its Repository
By Mamata Saha
Being among the prominent names in the IP and innovation space, Relecura believes in harnessing the power of AI and its associated concepts through continuous and persistent research and innovation, to create products or add features to those that cater to a wide range of innovation-related concerns.
To that end, the company has added yet another tool to its repository of AI-based tools. AI Classifier, the tool name, works to arrange massive volumes of patent data sets into simple categories in minutes, thereby doing away with the painstaking process of manual data management, which takes weeks, maybe even months to organize. Powered by AI, the highlight of the tool is its flexibility to adapt to different conditions through its no-code AI capability, saving time and money, crucial to the growth and success of a company.
In terms of the tool’s functionality, it follows a three-step process for data categorization. The classification process begins with model creation, is followed by model analysis, and ends with data classification. The model creation process requires some training documents shared by the user and comprises patent datasets with their assigned categories. It is performed in one of two ways. Either an excel file with patent numbers and categories can be imported, or the tool can be searched for documents posing as training sets. After the model is created, it can be further analyzed to gauge its accuracy, before it is used for the final stage of the document classification.
The analysis process commences with a selection of the model type, followed by the algorithm, where the user gets to choose from among several types. It is important to mention here that a key attribute of the Classifier tool is that it uses self-learning algorithms to look for patterns in the data sets. The patterns are then used to classify the documents with the help of some features. Interestingly, the user also has the option of selecting fields that form a part of the patents and allocating a specific percentage to them (weights), depending on how significant each one is to the user.
When the tool completes its analysis, it throws up relevant performance measures including precision and recall. High recall and precision are key to a successful classification, so they’re integrated into a mathematical formula or the F1-score. Finally, since most of us assimilate information visually, the confusion matrix provides the required visual representation of the categorization process. Should the user find the data from the model analysis ideal, they can upload sizable data volumes for classification.
While the tool is fairly new, it is remarkable to note that there have already been several user-configurable features added to further enhance its efficiency. This includes selections and weights on raw text fields, semantic concepts, and metadata on technologies and sub-technologies generated by the Relecura platform. The model uses the selections and associated weights to create optimal classification models for the specific use case.
Since we’ve shared so much information on the tool, it makes sense to share the output of the tool through some significant images for your benefit. Please view the ones included below.
In conclusion, you can always check out the ‘products’ section of our portal, https://relecura.com/products/, or request a demo to get a first-hand look and feel of our offerings.