How does it work?
The artificial intelligence (AI) reads in all patent documents from the database and uses word and document embeddings to calculate a numerical representation of each document. In the same manner the input text is also transformed into a numerical representation. By comparing those numerical representations with each other a similarity can be calculated.
Do I need keywords?
Due to the use of the artificial intelligence (AI) no special keywords are necessary compare or find documents. The similarity of documents is calculated using the structure of the text in such a matter that words are interchangeable. Therefore keywords have a very minor impact in finding similar documents.
What does the main graphics show?
The main graphics shows all documents from the database in principle. Because of the vast amount of documents each dot represents a few hundred documents in the vincinity. Clicking on one of those representation points brings up a detailed view with all documents. The graphics is centered around the input text. The more similar a document is the closer it is to the center. This means that a dot close to the center represents a document that is very similar to the input text. A very not similar text will be shown close or at the boundary of the graphics. The angle at which each document is shown represents the main topics the document has. Those topics are chosen by the AI and are not important for the similarity analysis. The topics of documents that can be used for further text analysis will be calculated separately (not finished yet).
What is a knowledge-graph?
A knowledge-graph is a visualisation of recognized entities in the input description text and their relationship to each other. It helps to find out what are the important bits in the description and how are they related. A knowledge graph can help profoundly to recognize what information is important in the description and if it might be helpful or hurts to fulfil the patent requirements.
What does the pdf-report include?
The pdf-report includes the main graphic to show the patent landscape around the input text and can be used as a first impression of the freedom-to-operate (fto). A list of the five most similar patents to the selected position in the main graphic is shown. If the selected position is the center of the graph, then the 5 most similar documents to the input text are listed. For each patent in the list the summary, a knowledge-graph with recognized entities and their relations and the description of the entities is shown. Additionally the several sections of the input text are analyzed with regard of the inpact on the similarity to the compared patent. Sections that make the overall text more or less similar to the compared patent can be easily identified. This simplifies the task of changing the input (description) text efficiently to increase or decrease the similarity. It helps to describe your idea as new.
In-depth comparison of the input description and chosen patents.
The in-depth comparison of the input description and chosen patents is done by a side-by-side presentation of the text and the findings of several AI tools.