Back to results
US 11,379,665 B1Granted
Document analysis architecture
AssigneeAon Risk Services, Inc. of Maryland
InventorsWilliam Michael Edmund, John E. Bradley, III
Life of this patent
Estimating the term…
Abstract
Systems and methods for generation and use of document analysis architectures are disclosed. A model builder component may be utilized to receiving user input data for labeling a set of documents as in class or out of class. That user input data may be utilized to train one or more classification models, which may then be utilized to predict classification of other documents. Trained models may be incorporated into a model taxonomy for searching and use by other users for document analysis purposes.
Claims18
A system, comprising:
- one or more processors;
- and
- non-transitory computer-readable media storing computer-executable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:
- storing, in association with a platform configured to receive documents from one or more databases, first data representing the documents, the documents including at least one of patents or patent applications;
- receiving user input data indicating a first portion of the documents that are out of class, the user input data received via a user interface configured to display components of individual ones of the documents and receive user input indicating whether the individual ones of the documents are in class;
- determining keywords of the first portion of the documents that represent the first portion of the documents;
- generating a classification model configured to classify documents, the classification model trained utilizing the keywords as indicators of subject matter that is out of class;
- and
- determining, utilizing the classification model:
- a second portion of the documents that are out of class;
- and
- a third portion of the documents that are in class, wherein the model:
- determines a first correlation between individual ones of the first portion of the documents and the keywords;
- determines a second correlation between the individual ones of the documents and a set of documents indicated to be in class via user input;
- and
- determines that the first correlation is greater than the second correlation.
13 more elements
A method, comprising:
- storing first data representing documents including at least one of patents or patent applications;
- receiving user input data indicating a first portion of the documents that are out of class;
- determining features of the first portion of the documents that represent the first portion of the documents;
- generating a model configured to classify the documents, the model trained based at least in part on the features;
- and
- determining, utilizing the model, a second portion of the documents that are out of class, wherein the model:
- determines a first correlation between individual ones of the documents and keywords that represent the first portion of the documents that are out of class;
- determines a second correlation between the individual ones of the documents and a set of documents indicated to be in class via user input;
- and
- determines that the first correlation is greater than the second correlation.
7 more elements
A system, comprising:
- one or more processors;
- and
- non-transitory computer-readable media storing instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:
- store first data representing documents including at least one of patents or patent applications;
- receiving user input data indicating a first portion of the documents that are out of class;
- determining keywords of the first portion of the documents that represent the first portion of the documents, wherein determining the keywords includes:
- generating a tokenized version of the first portion of the documents, the tokenized version including lexical tokens representing elements of the first portion of the documents;
- applying a bigram-based language model to the tokenized version of the first portion of the documents, the bigram-based language model configured to determine a bigram frequency of the lexical tokens;
- and
- selecting keywords corresponding to a portion of the lexical tokens having a high bigram frequency with respect to lexical tokens other than the portion of the lexical tokens;
- generating a model configured to classify the documents, the model trained based at least in part on the keywords;
- and
- determining, utilizing the model, a second portion of the documents that are out of class.
10 more elements
Dependencies are inferred from the claim text.