Abstract: A method computes page relevance for tabular contents from a document. The method includes receiving a table type and executing a term model to generate a term score of the page. The method further includes executing an embeddings model to generate an embeddings score of the page. The method further includes executing a summary model using a page summary generated with a language model to generate a summary score of the page. The method further includes executing a neighborhood model using a set of pages from the document to generate a neighborhood score of the page. The method further includes executing a combination model using the term score, the embeddings score, the summary score, and the neighborhood score to generate a combined score of the page. The method further includes presenting the combined score to indicate the page includes the table of the table type.

Abstract: A method computes page relevance for tabular contents from a document. The method includes receiving a table type and executing a term model to generate a term score of the page. The method further includes executing an embeddings model to generate an embeddings score of the page. The method further includes executing a summary model using a page summary generated with a language model to generate a summary score of the page. The method further includes executing a neighborhood model using a set of pages from the document to generate a neighborhood score of the page. The method further includes executing a combination model using the term score, the embeddings score, the summary score, and the neighborhood score to generate a combined score of the page. The method further includes presenting the combined score to indicate the page includes the table of the table type.

Abstract: A large language model (LLM) receives a prediction set extracted from a raw data source, and binary feedback labels labeling respective predictions of the prediction set. The LLM generates an explanation set, including explanations of the binary feedback labels of the predictions, based on the raw data source of the prediction set. The LLM further generates a set of attribute extraction instructions, using the explanation set and the prediction set. The set of attribute extraction instructions includes attribute extraction instructions for a particular attribute across a multitude of raw data sources. Training instances are generated from the set of attribute extraction instructions, having attributes, attribute values and extraction instructions. The training instances are annotated with an output probability distributions of the extracted attributes. An attribute extraction model is further trained with the annotated training dataset and deployed for future attribute extractions from raw data sources.

Abstract: A method implements language model powered search on structured records using relationship graphs. A relationship graph representing entity relationships among a set of tables in a database based on multiple schema definitions is constructed. A structured query based on the natural language query and the multiple schema definitions is generated. The structured query is deconstructed to extract a source table of the set of tables, a target table of the set of tables, a query condition, and an aggregation operator. A traversal path is determined across the relationship graph based on the source table and the target table. A set of entity-specific queries are executed using the query condition and the traversal path to retrieve records from the database. A response is generated based on the retrieved records and the aggregation operator. The response includes one or more of a textual summary and a visualization based on the output prompt.

Abstract: An LLM-powered search engine receives a natural language query from a conversational interface. The conversational interface is a first section of a user interface. The LLM-powered search engine generates a first response including a natural language summary, a data payload and an action recommendation. The data payload is validated with respect to the action recommendation to obtain a validation result. If the validation result is an error result, a correction LLM generates a correction prompt based on the error result and the first response. The LLM-powered search engine processes the correction prompt to generate a second response. A second data payload of the second response is validated with respect to a second action recommendation of the second response, to obtain a second validation result.

Abstract: A method including extracting, by a machine learning model executing using an electronic document, data to create extracted data. An error checking controller is executed on the extracted data to identify erroneous data within the extracted data. A label for the erroneous data is generated by a label controller executing on the erroneous data. The label identifies a type of error of the erroneous data and a correction to the type of error. The label is added to the erroneous data to generate labeled erroneous data. A training controller executes iterative steps to train the machine learning model using the electronic document, the labeled erroneous data, and a first instruction to generate new extracted data. The trained machine learning model is returned. The trained machine learning model has a reduced data extraction error rate relative to the machine learning model prior to executing the training controller.

Abstract: In general, a view coordinator receives a response to a user query from an LLM-powered search engine, including a natural language summary and a data payload. The natural language summary includes a reference to a data element of the data payload. The view coordinator displays the natural language summary in a first section of a user interface of a user application. The view coordinator further detects a selection of the reference. Upon selection of the reference, the view coordinator further identifies a viewer type corresponding to the data element, and invokes a corresponding viewing tool to generate a visualization of the data element. The viewing tool renders the visualization in a first viewer section of the user interface. The view coordinator further monitors the first viewer section for user interactions to detect selection of a second reference.

Abstract: A method implements masked text processing for information processing with documents. The method involves receiving a document page as an image including text image data. The method further involves extracting text unit data and text location data from the image corresponding to the text image data using an optical character recognition (OCR) engine. The method further involves generating mask data for the text unit data with color data based on text type data. The method further involves producing a masked image by replacing the text image data with the mask data using the color data with the location data in the image. The method further involves transmitting the masked image to a machine learning model to execute a downstream task.

Abstract: A method involves detecting primary entities in a document, involving determining that a subset of the primary entities are associated with a first primary entity type, and determining a second primary entity type of one of the primary entities. The method further involves processing the primary entity of the second primary entity type to determine a secondary entity type of the primary entity. The secondary entity type is a subcategory of the second primary entity type. The method also involves hierarchically organizing the primary entities into a document layout structure that includes a top level and a child level. The top level is established by the first subset of primary entities based on the first primary entity type identifying the first subset as headings, and the child level is established by the primary entity based on the second primary entity type, the child level identifying the secondary entity type.

Abstract: The present disclosure relates to a method. The method includes receiving resource reference data corresponding to oil and gas resources. The method also includes obtaining, from a first database, a first plurality of resource data associated with a first organization. Further, the method includes obtaining, from a second database different than the first database, a second plurality of resource data associated with a second organization. Further still, the method includes generating an organization semantics model based on the reference data, the first plurality of resource data, and the second plurality of resource data, wherein the organization semantics model is a language-learning model configured to generate a first response based on a received query corresponding to the first organization, and wherein the organization semantics model is configured to generate a second response based on the received query corresponding to the second organization.

Abstract: A method includes identifying entities in a well record database comprising data representing a plurality of objects and attributes of the objects, determining a data gap for at least one attribute of an object of the objects in the well record database, identifying documents in a document database, wherein identifying the documents include determining that the documents are relevant to the object based at least in part on metadata of the documents, extracting values for the data gap from the documents using a machine learning model, determining a data gap filler by aggregating the extracted values, and inserting the data gap filler into the data gap in the well log database.

Abstract: A method includes receiving historical well data comprising trajectories, performance data, and one or more drilling parameters for a plurality of wells, clustering at least a portion of the plurality of wells into a plurality of clusters based on the trajectories, using a machine learning model, receiving trajectory data for a subject well, identifying one of the clusters based on the trajectory data of the subject well, using the machine learning model, selecting one or more of the plurality of wells, or one or more sections thereof, in the cluster that was identified based on the performance data associated with the one or more of the plurality of wells or the portion thereof, and visualizing the selected one or more of the plurality of wells or one or more sections thereof.

Abstract: Systems, computer-readable media, and methods are provided. Relevant documents related to a specific entity are identified based on document metadata. Text and associated spatial coordinates are extracted based on relevant document pages. Significant document entities and associated spatial locations are identified. Page ranking is based on the extracted text and the spatial coordinates, the significant document entities, and image vector representations of the pages. A deep learning language model that utilizes the text and the spatial coordinates, layout information of the document entities, and the image vector representations of the pages is used to extract the user-defined attributes from the relevant document pages. First attribute values associated with the user-defined attributes are aggregated from the pages of one of the relevant documents into a single record. Second attribute values associated with the user-defined attributes are aggregated across the relevant documents.

Abstract: A method includes identifying entities in a well record database comprising data representing a plurality of objects and attributes of the objects, determining a data gap for at least one attribute of an object of the objects in the well record database, identifying documents in a document database, wherein identifying the documents include determining that the documents are relevant to the object based at least in part on metadata of the documents, extracting values for the data gap from the documents using a machine learning model, determining a data gap filler by aggregating the extracted values, and inserting the data gap filler into the data gap in the well log database.

Abstract: A method involves detecting primary entities in a document, involving determining that a subset of the primary entities are associated with a first primary entity type, and determining a second primary entity type of one of the primary entities. The method further involves processing the primary entity of the second primary entity type to determine a secondary entity type of the primary entity. The secondary entity type is a subcategory of the second primary entity type. The method also involves hierarchically organizing the primary entities into a document layout structure that includes a top level and a child level. The top level is established by the first subset of primary entities based on the first primary entity type identifying the first subset as headings, and the child level is established by the primary entity based on the second primary entity type, the child level identifying the secondary entity type.

Abstract: Systems, computer-readable media, and methods are provided. Relevant documents related to a specific entity are identified based on document metadata. Text and associated spatial coordinates are extracted based on relevant document pages. Significant document entities and associated spatial locations are identified. Page ranking is based on the extracted text and the spatial coordinates, the significant document entities, and image vector representations of the pages. A deep learning language model that utilizes the text and the spatial coordinates, layout information of the document entities, and the image vector representations of the pages is used to extract the user-defined attributes from the relevant document pages. First attribute values associated with the user-defined attributes are aggregated from the pages of one of the relevant documents into a single record. Second attribute values associated with the user-defined attributes are aggregated across the relevant documents.

Abstract: A method involves obtaining query pressure transient analysis (PTA) data from a well associated with a reservoir, and obtaining a selected class of physics models from a multitude of classes of physics models using a first machine learning model operating on the query PTA data. A physics model in at least one of the multitude of classes of physics models includes a well model and a reservoir model. The well model and the reservoir model are parameterized with model parameters having model parameter values. The method further involves obtaining a multitude of model parameter value estimates to form a parameterized query physics model of the selected class of physics models, using a second machine learning model operating on the query PTA data; and providing the parameterized query physics model to a user.

Abstract: Systems, computer-readable media, and methods are provided. Relevant documents related to a specific entity are identified based on document metadata. Text and associated spatial coordinates are extracted based on relevant document pages. Significant document entities and associated spatial locations are identified. Page ranking is based on the extracted text and the spatial coordinates, the significant document entities, and image vector representations of the pages. A deep learning language model that utilizes the text and the spatial coordinates, layout information of the document entities, and the image vector representations of the pages is used to extract the user-defined attributes from the relevant document pages. First attribute values associated with the user-defined attributes are aggregated from the pages of one of the relevant documents into a single record. Second attribute values associated with the user-defined attributes are aggregated across the relevant documents.

Abstract: A control system and method utilizing one or more processors that are configured to determine contaminant loading of blades of a turbomachinery compressor based on one or more environmental conditions to which the turbomachinery compressor is exposed and one or more atmospheric air inlet conditions of the turbomachinery compressor. The one or more processors then determine a corrosion contaminant concentration on the blades of the turbomachinery compressor based on the contaminant loading that is determined and determine an upper limit on or a distribution of potential corrosion of the blades of the turbomachinery based on the corrosion contaminant concentration, at least one of the environmental conditions to which the turbomachinery compressor is exposed, and the corrosion contaminant concentration that is determined.

Abstract: A control system and method utilizing one or more processors that are configured to determine contaminant loading of blades of a turbomachinery compressor based on one or more environmental conditions to which the turbomachinery compressor is exposed and one or more atmospheric air inlet conditions of the turbomachinery compressor. The one or more processors then determine a corrosion contaminant concentration on the blades of the turbomachinery compressor based on the contaminant loading that is determined and determine an upper limit on or a distribution of potential corrosion of the blades of the turbomachinery based on the corrosion contaminant concentration, at least one of the environmental conditions to which the turbomachinery compressor is exposed, and the corrosion contaminant concentration that is determined.