Abstract: An auto-encoder model processes a datasets describing a physical part from a part catalogue in the form of a property co-occurrence graph is provided, and performs entity resolution and auto-completion on the co-occurrence graph in order to compute a corrected and/or completed dataset. The encoder includes a recurrent neural network and a graph attention network. The decoder contains a linear decoder for numeric values and a recurrent neural network decoder for strings. The auto-encoder model provides an automated end-to-end solution that can auto-complete missing information as well as correct data errors such as misspellings or wrong values. The auto-encoder model is capable of auto-completion for highly unaligned part specification data with missing values.

Abstract: For ticket classification in a building automation system, a large language model (LLM) is used to classify. In one approach, a prompt is generated for zero-shot classification, and a prompt is generated for few-shot classification. In another approach, a hybrid annotation provides corrections (review) by an expert to correct LLM classification for sample tickets to be used as examples in the few-shot classification. The LLM may operate on a diverse and complex range of tickets in an efficient and scalable manner.

Abstract: A method for integrating data from different data sources into a knowledge graph storage unit including: transferring data from different data sources into a receiving and extraction module of an extraction-transformation-loading, ETL, data pipeline framework; extracting the loaded data using an extraction data frame and transferring the extracted data to a transformation module; transforming the extracted data using transformation functionalities and transferring the transformed data to a mapping module; mapping the transformed data using mapping rule functionalities and transferring the mapped data to a post-processing module; post-processing the mapped data using a library and transferring the post-processed data to a loading module; and loading the processed data by the loading module to the knowledge graph storage unit; wherein the mapping module uses a JSON-based mapping notation to convert the transformed data into a serialized form, is provided.

Abstract: In an iterative loop a machine learning model is trained with labeled data, thereby forming a trained machine learning model, and then deployed. The trained machine learning model then receives as input at least pairs of equipment identifiers contained in unlabeled data and calculates predictions, wherein the predictions contain at least one match prediction for a pair of equipment identifiers indicating that the equipment identifiers refer to a same industrial equipment, and wherein the predictions contain in particular at least one different prediction for a pair of equipment identifiers indicating that the equipment identifiers refer to different industrial equipment. An inaccuracy detector using an inaccuracy heuristic containing in particular probability thresholds for correct predictions, detects accurate and inaccurate predictions among the predictions, and collects the accurate predictions as automatically labeled data.

Abstract: To restore consistency of a digital twin database, identifiers with metadata imported from various data sources are processed by an encoder, which computes latent representations of the identifiers that are compared by an efficient similarity metric. If the respective similarity score exceeds a threshold, a match is detected between the identifiers. In that case, the digital twin database is updated by aligning the first identifier and the second identifier. This matching algorithm for equipment identifiers updates the digital twin data automatically and continuously by aligning identifiers which refer to the same piece of equipment. The updates flow directly into the digital twin database, thereby removing the manual effort. Using approximate nearest neighbor methods is highly efficient, especially for large plants. The encoder is implemented as an autoencoder which relies only on unlabeled training data.

Abstract: A computer-implemented method for determining a denoised named entity recognition (NER)-model and denoised relation extraction (RE)-model. A computer-implemented method for propagating an input dataset into a graph database representation, a computing unit and a computer program product, is also provided.

Abstract: Provided is a computer-implemented method for receiving the at least two log files; wherein each log file of the at least two log files includes at least one log entry with at least one time stamp and at least one message; wherein the at least two log files differ from one another with respect to at least one distinctive criteria; extracting at least one additional information of each log file of the at least two log files; and combining each log file of the at least two log files with the extracted additional information into at least two processed log files; wherein the at least two processed log files comply with a coherent representation. A corresponding computer program product and generating unit is also provided.

Abstract: An auto-encoder model processes a datasets describing a physical part from a part catalogue in the form of a property co-occurrence graph is provided, and performs entity resolution and auto-completion on the co-occurrence graph in order to compute a corrected and/or completed dataset. The encoder includes a recurrent neural network and a graph attention network. The decoder contains a linear decoder for numeric values and a recurrent neural network decoder for strings. The auto-encoder model provides an automated end-to-end solution that can auto-complete missing information as well as correct data errors such as misspellings or wrong values. The auto-encoder model is capable of auto-completion for highly unaligned part specification data with missing values.

Abstract: Provided is a computer-implemented method for generating automatically annotations for tabular cell data of a table having column and rows, wherein the method includes: supplying raw cell data of cells of a row of the table as input to an embedding layer of a semantic type annotation neural network which transforms the received raw cell data of the cells of the supplied row into cell embedding vectors; processing the cell embedding vectors to calculate attentions among the cells of the respective row of the table encoding a context within the row output as cell context vectors; and processing the cell context vectors generated by the self-attention layer by a classification layer of the semantic type annotation neural network to predict semantic column type annotations and/or to predict relations between semantic column type annotations for the columns of the table.

Abstract: To restore consistency of a digital twin database, identifiers with metadata imported from various data sources are processed by an encoder, which computes latent representations of the identifiers that are compared by an efficient similarity metric. If the respective similarity score exceeds a threshold, a match is detected between the identifiers. In that case, the digital twin database is updated by aligning the first identifier and the second identifier. This matching algorithm for equipment identifiers updates the digital twin data automatically and continuously by aligning identifiers which refer to the same piece of equipment. The updates flow directly into the digital twin database, thereby removing the manual effort. Using approximate nearest neighbor methods is highly efficient, especially for large plants. The encoder is implemented as an autoencoder which relies only on unlabeled training data.

Abstract: Provided is a computer-implemented method for receiving the at least two log files; wherein each log file of the at least two log files includes at least one log entry with at least one time stamp and at least one message; wherein the at least two log files differ from one another with respect to at least one distinctive criteria; extracting at least one additional information of each log file of the at least two log files; and combining each log file of the at least two log files with the extracted additional information into at least two processed log files; wherein the at least two processed log files comply with a coherent representation. A corresponding computer program product and generating unit is also provided.

Abstract: A computer-implemented method for determining a denoised named entity recognition (NER)-model and denoised relation extraction (RE)-model. A computer-implemented method for propagating an input dataset into a graph database representation, a computing unit and a computer program product, is also provided.

Abstract: A method for integrating data from different data sources into a knowledge graph storage unit including: transferring data from different data sources into a receiving and extraction module of an extraction-transformation-loading, ETL, data pipeline framework; (b) extracting the loaded data using an extraction data frame and transferring the extracted data to a transformation module; (c) transforming the extracted data using transformation functionalities and transferring the transformed data to a mapping module; (d) mapping the transformed data using mapping rule functionalities and transferring the mapped data to a post-processing module; (e) post-processing the mapped data using a library and transferring the post-processed data to a loading module; and (f) loading the processed data by the loading module to the knowledge graph storage unit; wherein the mapping module uses a JSON-based mapping notation to convert the transformed data into a serialized form, is provided.

Abstract: Provided is a computer-implemented method for generating automatically annotations for tabular cell data of a table having column and rows, wherein the method includes: supplying raw cell data of cells of a row of the table as input to an embedding layer of a semantic type annotation neural network which transforms the received raw cell data of the cells of the supplied row into cell embedding vectors; processing the cell embedding vectors to calculate attentions among the cells of the respective row of the table encoding a context within the row output as cell context vectors; and processing the cell context vectors generated by the self-attention layer by a classification layer of the semantic type annotation neural network to predict semantic column type annotations and/or to predict relations between semantic column type annotations for the columns of the table.

Abstract: Provided is a Computer-implemented method for Receiving the at least two log files; wherein each log file of the at least two log files includes at least one log entry with at least one time stamp and at least one message; wherein the at least two log files differ from one another with respect to at least one distinctive criteria; Extracting at least one additional information of each log file of the at least two log files; and Combining each log file of the at least two log files with the extracted additional information into at least two processed log tiles; wherein the at least two processed log files comply with a coherent representation. Further, the invention relates to a corresponding computer program product and generating unit.