The platform uses an ensemble of AI algorithms and more traditional search models including:
● Natural language based search queries are deconstructed using a natural language processing(NLP) model and plotted against the world’s patent literature.
● Machine learning algorithms are used to index the patent database and compile a thesaurus ofsynonyms that is constantly updated. Words in the query are tokenized and replaced with all or a subset of synonyms based on user input via a slide bar.
● Every query isclassified based on the collaborative patent classification (CPC). Our classifier is better than any published classifier to date.
● Returned results from each of the ensemble of search models are processed to remove duplicates,assign a relevancy score to each reference by comparing the references to search query, and order the references by relevance before they are returned to the user interface.
● An NLP model identifies “Key Phrases” from the text of each document and returns these results to the user interface.
● Advanced Filters uses an NLP based model to weight phrases, sentences, and paragraphs in returned references that include specific “Facet Terms,” and reorder return results based on their use of selected facet terms and, in some cases, synonyms of the selected facet terms.