“Method, Apparatus And Computer Program Product For Graph-Based Encoding Of Natural Language Data Objects” in Patent Application Approval Process (USPTO 20230085697): Patent Application

2023 APR 11 (NewsRx) -- By a News Reporter-Staff News Editor at Insurance Daily News -- A patent application by the inventor BULU, Irfan (Sartell, MN, US), filed on September 21, 2021, was made available online on March 23, 2023, according to news reporting originating from Washington, D.C., by NewsRx correspondents.

This patent application has not been assigned to a company or institution.

The following quote was obtained by the news editors from the background information supplied by the inventors: “Natural language processing and machine learning systems have great potential for providing various technical advancement and technical benefits not only in the field of computer science, but also in other associated technical fields and applications. Applicant has identified many technical challenges, deficiencies and problems associated with natural language processing and machine learning systems and methods.”

In addition to the background information obtained for this patent application, NewsRx journalists also obtained the inventor’s summary information for this patent application: “In general, embodiments of the present disclosure provide methods, apparatuses, systems, computing devices, computing entities, and/or the like.

“In accordance with various embodiments of the present disclosure, an apparatus is provided. The apparatus may comprise at least one processor and at least one non-transitory memory comprising a computer program code. The at least one non-transitory memory and the computer program code may be configured to, with the at least one processor, cause the apparatus to retrieve a plurality of natural language data objects from a database; determine, based at least in part on the plurality of natural language data objects and by utilizing an entity extraction machine learning model, a plurality of entity identifiers for the plurality of natural language data objects, wherein: (i) the entity extraction machine learning model comprises an encoder sub-model and an entity classification sub-model, (ii) the encoder sub-model is configured to generate a plurality of text embeddings based at least in part on the plurality of natural language data objects, (iii) the entity classification sub-model is configured to determine an entity classification for each text embedding, and (iv) the plurality of entity identifiers are determined based at least in part on each entity classification; determine, based at least in part on the plurality of entity identifiers and by utilizing the entity extraction machine learning model, one or more entity relationship identifiers for the plurality of natural language data objects, wherein: (i) the entity extraction machine learning model comprises an entity relationship classification sub-model, (ii) the entity relationship classification sub-model is configured to determine an entity relationship classification for each entity pair from the plurality of entity identifiers based at least in part on a subset of the plurality of text embeddings that corresponds to the entity pair, and (iii) the one or more entity relationship identifiers are determined based at least in part on each entity relationship classification; generate, based at least in part on the plurality of entity identifiers and the one or more entity relationship identifiers, a graph-based data object for the plurality of natural language data objects; and perform one or more prediction-based actions based at least in part on the graph-based data object.

“In some embodiments, the encoder sub-model is associated with a multi-headed attention mechanism.

“In some embodiments, the encoder sub-model comprises a Bidirectional Encoder Representations from Transformers (BERT) model.

“In some embodiments, when generating the graph-based data object, the at least one non-transitory memory and the computer program code are configured to, with the at least one processor, cause the apparatus to: generate a plurality of nodes of the graph-based data object based at least in part on the plurality of entity identifiers; and generate a plurality of edges of the graph-based data object based at least in part on the one or more entity relationship identifiers.

“In some embodiments, the plurality of natural language data objects comprises at least one textual contract data object and at least one medical record data object.

“In some embodiments, when generating the graph-based data object, the at least one non-transitory memory and the computer program code are configured to, with the at least one processor, cause the apparatus to: in response to determining that the at least one textual contract data object and the at least one medical record data object are associated with a first patient entity identifier of the plurality of entity identifiers, generate a patient entity node; generate, based at least in part on the at least one medical record data object, at least one symptom node and a first edge connecting the at least one symptom node to the patient entity node; and generate, based at least in part on the at least one textual contract data object, at least one procedure node and a second edge connecting the at least one procedure node to the patent entity node.

“In some embodiments, the at least one procedure node is associated with at least one International Classification of Diseases (ICD) code.

“In some embodiments, the graph-based data object comprises a plurality of nodes and a plurality of edges connecting the plurality of nodes. In some embodiments, each of the plurality of nodes corresponds to an entity associated with the plurality of natural language data objects. In some embodiments, each of the plurality of edges corresponds to a relationship between entities associated with the plurality of natural language data objects.

“In some embodiments, the plurality of nodes is associated with a plurality of node types. In some embodiments, the plurality of edges is associated with a plurality of edge types that is determined based at least in part on the plurality anode types.

“In some embodiments, when performing the one or more prediction-based actions, the at least one non-transitory memory and the computer program code are configured to, with the at least one processor, cause the apparatus to: receive, from a client computing device, a data prediction request associated with at least one entity identifier of the plurality of entity identifiers; in response to receiving the data prediction request, identify, based at least in part on the at least one entity identifier, a related sub-graph of the graph-based data object that corresponds to the at least one entity identifier; generate, based at least in part on the related sub-graph, at least one prediction data object using a data prediction machine learning model; and transmit the at least one prediction data object to the client computing device.

“In some embodiments, when identifying the related sub-graph of the graph-based data object that corresponds to the at least one entity identifier, the at least one non-transitory memory and the computer program code are configured to, with the at least one processor, cause the apparatus to: determine at least a first node from a plurality of nodes of the graph-based data object that is associated with the at least one entity identifier; and determine at least a first edge from a plurality of edges of the graph-based data object that connects the first node to at least a second node.

“In some embodiments, when performing the one or more prediction-based actions, the at least one non-transitory memory and the computer program code are configured to, with the at least one processor, cause the apparatus to: generate the at least one prediction data object based at least in part on the first node, the first edge, and the second node using the data prediction machine learning model.

“In some embodiments, the at least one non-transitory memory and the computer program code are configured to, with the at least one processor, cause the apparatus to: train the data prediction machine learning model using a training data set, wherein the training data set comprises a plurality of historical data prediction requests that corresponds to a plurality of historical response data objects; and subsequent to training the data prediction machine learning model, generate the at least one prediction data object based at least in part on the data prediction request and the graph-based data object.

“In some embodiments, the data prediction request is associated with a preauthorization request and comprises a procedure identifier, a patient entity identifier, a healthcare provider entity identifier, and a health insurance provider entity identifier.

“In some embodiments, when generating the at least one prediction data object, the at least one non-transitory memory and the computer program code are configured to, with the at least one processor, cause the apparatus to: identify, from a plurality of nodes of the graph-based data object, a patient entity node associated with the patient entity identifier; identify, from the plurality of nodes of the graph-based data object, a healthcare provider entity node associated with the healthcare, provider entity identifier; identify, from the plurality of nodes of the graph-based data object, a procedure node associated with the procedure identifier; and calculate, based at least in part on the data prediction machine learning model, (i) a prediction data object indicating a predicted probability of at least one edge connecting the procedure node to the patient entity node and to the healthcare provider entity node and (ii) a prediction confidence score associated with the prediction data object. In some embodiments, the data prediction machine learning model is an unsupervised machine learning model.

“In some embodiments, the at least one non-transitory memory and the computer program code are configured to, with the at least one processor, cause the apparatus to: determine whether the prediction confidence score satisfies a data prediction threshold.

“In some embodiments, the at least one non-transitory memory and the computer program code are configured to, with the at least one processor, cause the apparatus to: in response to determining that the prediction confidence score satisfies the data prediction threshold, generate at least one recommendation data object based at least in part on the at least one prediction data object.

“In some embodiments, the at least one non-transitory memory and the computer program code are configured to, with the at least one processor, cause the apparatus to: in response to determining that the prediction confidence score does not satisfy the data prediction threshold, transmit a data prediction review request to the client computing device.”

There is additional summary information. Please visit full patent to read further.”

The claims supplied by the inventors are:

“1. An apparatus comprising at least one processor and at least one non-transitory memory comprising a computer program code, the at least one non-transitory memory and the computer program code configured to, with the at least one processor, cause the apparatus to: retrieve a plurality of natural language data objects from a database; determine, based at least in part on the plurality of natural language data objects and by utilizing an entity extraction machine learning model, a plurality of entity identifiers for the plurality of natural language data objects, wherein: (i) the entity extraction machine learning model comprises an encoder sub-model and an entity classification sub-model, (ii) the encoder sub-model is configured to generate a plurality of text embeddings based at least in part on the plurality natural language data objects, (iii) the entity classification sub-model is configured to determine an entity classification for each text embedding, and (iv) the plurality of entity identifiers are determined based at least in part on each entity classification; determine, based at least in part on the plurality of entity identifiers and by utilizing the entity extraction machine learning model, one or more entity relationship identifiers for the plurality of natural language data objects, wherein: (i) the entity extraction machine learning model comprises an entity relationship classification sub-model, (ii) the entity relationship classification sub-model is configured to determine an entity relationship classification for each entity pair from the plurality of entity identifiers based at least in part on a subset of the plurality of text embeddings that corresponds to the entity pair, and (iii) the one or more entity relationship identifiers are determined based at least in part on each entity relationship classification; generate, based at least in part on the plurality of entity identifiers and the one or more entity relationship identifiers, a graph-based data object for the plurality of natural language data objects; and perform one or more prediction-based actions based at least in part on the graph-based data object.

“2. The apparatus of claim 1, wherein the encoder sub-model is associated with a multi-headed attention mechanism.

“3. The apparatus of claim 2, wherein the encoder sub-model comprises a Bidirectional Encoder Representations from Transformers (BERT) model.

“4. The apparatus of claim 2, wherein, when generating the graph-based data object, the at least one non-transitory memory and the computer program code are configured to, with the at least one processor, cause the apparatus to: generate a plurality of nodes of the graph-based data object based at least in part on the plurality of entity identifiers; and generate a plurality of wedges of the graph-based data object based at least in part on the one or more entity relationship identifiers.

“5. The apparatus of claim 4, wherein the plurality of natural language data objects comprises at least one textual contract data object and at least one medical record data object.

“6. The apparatus of claim 5, wherein, when generating the graph-based data object, the at least one non-transitory memory and the computer program code are configured to, with the at least one processor, cause the apparatus to: in response to determining that the at least one textual contract data object and the at least one medical record data object are associated with a first patient entity identifier of the plurality of entity identifiers, generate a patient entity node; generate, based at least in part on the at least one medical record data object, at least one symptom node and a first edge connecting the at least one symptom node to the patient entity node; and generate, based at least in part on the at least one textual contract data object, at least one procedure node and a second edge connecting the at least one procedure node to the patient entity node.

“7. The apparatus of claim 6, wherein the at least one procedure node is associated with at least one International Classification of Diseases (ICD) code.

“8. The apparatus of claim 1, wherein the graph-based data object comprises a plurality of nodes and a plurality of edges connecting the plurality of nodes, wherein each of the plurality of nodes corresponds to an entity associated with the plurality of natural language data objects, and wherein each of the plurality of edges corresponds to a relationship between entities associated with the plurality of natural language data objects.

“9. The apparatus of claim 8, wherein the plurality of nodes is associated with a plurality of node types, and wherein the plurality of edges is associated with a plurality of edge types that is determined based at least in part on the plurality of node types.

“10. The apparatus of claim 1, wherein, when performing the one or more prediction-based actions, the at least one non-transitory memory and the computer program code are configured to, with the at least one processor, cause the apparatus to: receive, from a client computing device, a data prediction request associated with at least one entity identifier of the plurality of entity identifiers; in response to receiving the data prediction request, identify, based at least in part on the at least one entity identifier, a related sub-graph of the graph-based data object that corresponds to the at least one entity identifier; generate, based at least in part on the related sub-graph, at least one prediction data object using a data prediction machine learning model; and transmit the at least one prediction data object to the client computing device.

“11. The apparatus of claim 10, wherein, when identifying the related sub-graph of the graph-based data object that corresponds to the at least one entity identifier, the at least one non-transitory memory and the computer program code are configured to, with the at least one processor, cause the apparatus to: determine at least a first node from a plurality of nodes of the graph-based data object that is associated with the at least one entity identifier; and determine at least a first edge from a plurality of edges of the graph-based data object that connects the first node to at least a second node.

“12. The apparatus of claim 11, wherein, when performing the one or more prediction-based actions, the at least one non-transitory memory and the computer program code are configured to, with the at least one processor, cause the apparatus to: generate the at least one prediction data object based at least in part on the first node, the first edge, and the second node using the data prediction machine learning model.

“13. The apparatus of claim 10, wherein the at least one non-transitory memory and the computer program code are configured to, with the at least one processor, cause the apparatus to: train the data prediction machine learning model using a training data set, wherein the training data set comprises a plurality of historical data prediction requests that corresponds to a plurality of historical response data objects; and subsequent to training the data prediction machine learning model, generate the at least one prediction data object based at least in part on the data prediction request and the graph-based data object.

“14. The apparatus of claim 10, wherein the data prediction request is associated with a preauthorization request and comprises a procedure identifier, a patient entity identifier, and a healthcare provider entity identifier.

“15. The apparatus of claim 14, wherein, when generating the at least one prediction data object, the at least one non-transitory memory and the computer program code are configured to, with the at least one processor, cause the apparatus to: identify, from a plurality of nodes of the graph-based data object, a patient entity node associated with the patient entity identifier; identify, from the plurality of nodes of the graph-based data object, a healthcare provider entity node associated with the healthcare provider entity identifier; identify, from the plurality of nodes of the graph-based data object, a procedure node associated with the procedure identifier; and calculate, based at least in part on the data prediction machine learning model, (i) a prediction data object indicating a predicted probability of at least one edge connecting the procedure node to the patient entity node and to the healthcare provider entity node and (ii) a prediction confidence score associated with the prediction data object, wherein the data prediction machine learning model is an unsupervised machine learning model.

“16. The apparatus of claim 15, wherein the at least one non-transitory memory and the computer program code are configured to, with the at least one processor, cause the apparatus to: determine whether the prediction confidence score satisfies a data prediction threshold.

“17. The apparatus of claim 16, wherein the at least one non-transitory memory and the computer program code are configured to, with the at least one processor, cause the apparatus to: in response to determining that the prediction confidence score satisfies the data prediction threshold, generate at least one recommendation data object based at least in part on the at least one prediction data object.

“18. The apparatus of claim 16, wherein the at least one non-transitory memory and the computer program code are configured to, with the at least one processor, cause the apparatus to: in response to determining that the prediction confidence score does not satisfy the data prediction threshold, transmit a data prediction review request to the client computing device.”

There are additional claims. Please visit full patent to read further.

URL and more information on this patent application, see: BULU, Irfan. Method, Apparatus And Computer Program Product For Graph-Based Encoding Of Natural Language Data Objects. U.S. Patent Application Number 20230085697, filed September 21, 2021 and posted March 23, 2023. Patent URL (for desktop use only): https://ppubs.uspto.gov/pubwebapp/external.html?q=(20230085697)&db=US-PGPUB&type=ids

(Our reports deliver fact-based news of research and discoveries from around the world.)