US 12,608,540 B1 · Breadth scoring using relevant document term frequency

Abstract

Systems and methods for generation and use of document analysis platform architectures are disclosed. The document analysis platform may be configured to determine word frequencies for words included in a given portion of a given document based on a frequency of occurrences of the words in additional documents that are similar to the given document. The document analysis platform may utilize one or more formulas to determine word weights for the words based on the frequencies. The document analysis platform may then determine one or more scores (e.g., word scores, breadth scores, document scores, etc.) by assigning the word weights to the words. The document analysis platform may also determine a likelihood of infringement of a given document in a district court action. The document analysis platform may utilize machine-learned (ML) models and/or generative artificial intelligence (AI) to determine similarities, word weights, and/or likelihoods of infringement.

A method comprising:

receiving a corpus of documents, the corpus of documents comprising at least a first document including first text;
pre-processing the first document to generate one or more first processed document portions from the first document, the one or more first processed document portions including first words from the first text;
identifying second documents that are similar to the first document, individual ones of the second documents including second text;
pre-processing the second documents to generate one or more second processed document portions from individual ones of the second documents, the one or more second processed document portions including second words from the second text;
determining word frequencies for individual ones of the second words included in the one or more second processed document portions, a word frequency of the word frequencies being based at least in part on a first number of the second documents that include a first word of the second words and a second number representing a total number of the second documents that are similar to the first document;
generating a first word weight associated with the first word based at least in part on taking a cube root of a quotient of the first number of the second documents that include the first word divided by the second number representing the total number of the second documents, wherein:
the first word weight is weighted lesser than the second word weight associated with a second word included in the first text based at least in part on determining that the first word frequency is greater than the second word frequency associated with the second word;
or
the first word weight is weighted greater than a third word weight associated with a third word based at least in part on determining that the first word frequency is lesser than a third word frequency associated with the third word:
for individual ones of the one or more first processed document portions, identifying:
one or more common words from among the first words based at least in part on the word frequencies associated with the individual ones of the second words, the common words having a first word frequency;
and
one or more rare words from among the first words based at least in part on the word frequencies associated with the individual ones of the second words, the rare words having a second word frequency that is less than the first word frequency;
generating an element frequency for individual elements included in individual ones of the one or more first processed document portions, the elements comprising one or more of the first words and being determined based at least in part on natural language processing;
generating word scores for the elements based at least in part on assigning word weights to the elements, a first word weight of the word weights being based at least in part on a first element frequency of the element frequencies corresponding to a first element of the elements;
generating breadth scores for the first processed document portions, an individual breadth score corresponding to an individual first processed document portion of the first processed document portions being based at least in part on a sum of the word scores associated with the elements included in the individual first processed document portion;
and
generating a graphical user interface (GUI) configured to display on a computing device, the GUI configured to display the breadth scores for the first processed document portions in proximity to generated document identification numbers associated with the individual first processed document portion.

15 more elements

Breadth scoring using relevant document term frequency

Life of this patent

Breadth scoring using relevant document term frequency

Life of this patent

Abstract

Claims17