Abstract: A method of extracting meta-data from a document includes capturing style attributes from the document, identifying cell-wise location coordinates for text characters using page segmentation and border table extraction, and finding relationship between nearby cells using surrounding embedding by determining shortest distant text cell in top, left, right, and bottom direction. The method further includes applying Graph Convolution Network with Informative Attention (GCN-IA) for providing more attention to informative nodes for generating better representation of surrounding embedding and capturing a deep contextual meaning from text cells. A domain specific language model is utilized and improved by a domain aware tokenizer.

Abstract: A system and method of extracting tables and figures from a drawing document is disclosed. The method may include processing coloured image to segmented binary image and extracting a plurality of horizontal lines and a plurality of vertical lines from a foreground of the image. The method may further include detecting a set of candidate table region from the plurality of horizontal lines and the plurality of vertical lines in the image. Further, the method may include calculating textual region density corresponding to each of the set of candidate table regions in the image. The method may further include identifying at least one relevant table region from the set of candidate table regions in the image and a text free region from the at least one additional region in the image. The method may further include identifying at least one figure region from the dilated text free region.

Abstract: A method of extracting dimension data from a document is disclosed. The method includes receiving the document comprising at least one two-dimensional figure and a plurality of dimension sets associated with the at least one two-dimensional figure. The method may include detecting the at least one two-dimensional figure in the document. The method may further include detecting the plurality of dimension sets distinctly from the at least one two-dimensional figure in the document. Further, the method may identify a plurality of arrowheads associated with the plurality of dimension sets. The method may include clustering the plurality of arrowheads to obtain a plurality of set of arrowheads. The method may further include mapping each of the plurality of set of arrowheads with the dimension value and extracting dimension data corresponding to each of the plurality of set of arrowheads based on the mapping.

Abstract: A method and system of extracting borderless structure using image processing is disclosed. The method may include converting a received document into a binary image comprising a plurality of text characters. A first image is created comprising a plurality of text blobs by connecting text characters and merging the plurality of text blobs to create one or more text line blobs to generate a second image. Further the first image and the second image are compared to generate a third image comprising a plurality of gap blobs. The gap blobs are clustered into one or more groups to determine a localized region of interest (ROI). Further lines are identified within the ROI using pixel density and separated into rows and columns. The final output contains list of cell coordinates.

Abstract: A method and system of extracting data from a set of documents is disclosed. A processor determines a plurality of spatial features for each of the set of documents based on a set of keywords and a set of entities extracted from the documents. A variance is determined between at least one of the plurality of spatial features determined for each of the documents. Layout is determined based on the spatial features and variance. Each set of documents is clustered in at least one cluster based on a similarity between the layouts of the documents using a first machine learning model. Spatial features are selected using a second machine learning model. Data of the set of entities is extracted corresponding to the keywords in the documents based on the selection of the features and the similarity between the layouts using a third machine learning model.

Abstract: A method and system for extracting text from a video stream is disclosed. The method may include determining at least one visual text region in each image of a plurality of images. The at least one visual text region may include a plurality of text characters and determining the at least one visual text region is based on analysis of one of lines and curves associated with each of the plurality of text characters, blob associated with the plurality of text characters along an axis, distribution of the plurality of text characters along a major axis, and a common attribute associated with the plurality of text characters. The method may further include segmenting the plurality of text characters from a background, and inpainting each of the plurality of text characters with a predefined color.

Abstract: A method and system of optimizing query input for generating a set of code using a large language model (LLM), is disclosed. A processor receives a user query input by a user for querying the LLM to generate the set of code corresponding to a user-defined coding language. A set of primitive queries are determined from the user query based on the predefined keyword reference table and using an NLP model. Metadata is determined from the set of primitive queries. A query type is determined from a set of predefined query types of the user query using a machine learning model. An optimized query is determined based on the query type and the metadata using the NLP model and a historical database. The set of code is determined corresponding to the user-defined coding language based on the query type and the optimized query by querying the LLM.

Abstract: A method and system of classifying documents is disclosed. The method includes determining line-text data for each of a plurality of lines of the document image using a text extraction technique. A set of unique keywords in the document image is determined based on a predefined list of keywords in the line-text data for each of the plurality of lines. For each keyword, two forward nodes are determined as next two subsequent keywords in the set of unique keywords based on determination of a shortest distance between the corresponding keyword and the next two subsequent keywords and based on the pre-defined reading sequence. Weights of each of the two forward nodes are determined based on the shortest distance and an angle of each of the two forward nodes with the corresponding keyword. A cluster is determined based on the feature matrix using a machine learning clustering model.

Abstract: The present invention generally relates to the field of electro-chemical and energy storage technology. In particular, the present invention relates to the laminar growth mechanism of the Li3VO4 (LVO) anode material and its ultra-long cycling under high C-rate for its application in metal ion batteries such as lithium-ion batteries, sodium ion batteries, or Zinc ion batteries.

Abstract: A method and a device for creating and training machine learning models is disclosed. In an embodiment, a method for training a machine learning model for identifying entities from data includes creating a first plurality of clusters from a first plurality of data samples in a first dataset and a second plurality of clusters from a second plurality of data samples in a second dataset. The method further includes determining a rank for each of the first plurality of clusters and a rank for each of the second plurality of clusters. The method includes retraining the machine learning model using at least one of the first plurality of clusters weighted based on the rank determined for each of the first plurality of clusters and at least one of the second plurality of clusters weighted based on the rank determined for each of the second plurality of clusters.

Abstract: A method and system for performing optical character recognition (OCR) error correction is disclosed. The method includes receiving at least one document image. One or more text entities are determined in the data using an OCR technique. A character embedding, a layout embedding, and a style embedding is determined of each of the one or more text entities in the at least one document image. A concatenated embedding is determined of each of the one or more text entities based on the corresponding character embedding, style embedding, and layout embedding of each of the one or more text entities. A corrected character embedding of each character recognized in the corresponding text entity based on the corresponding concatenated embedding using an encoder-decoder model.

Abstract: A method and system for generating binary image of a document image is disclosed. The method includes determining a negative map image, an inverse negative map image, an HSV image and a grayscale image of the document image. One or more cells corresponding to at least one table are detected based on detection of lines in the negative map image. For each of the cells, a foreground mean value, a background mean value, and a background mean HSV value is determined. Each of the cells are categorized as a dark cell or a light cell based on the foreground mean value and the background mean value. Contrast value of each cell is determined based on the foreground mean value and the background mean value. The binary image is determined based on the contrast value, a pre-defined threshold value, the background mean HSV value and the categorization of the corresponding cell.

Abstract: A method and system of classifying documents is disclosed. The method includes determining line-text data for each of a plurality of lines of the document image using a text extraction technique. A set of unique keywords in the document image is determined based on a predefined list of keywords in the line-text data for each of the plurality of lines. For each keyword, two forward nodes are determined as next two subsequent keywords in the set of unique keywords based on determination of a shortest distance between the corresponding keyword and the next two subsequent keywords and based on the pre-defined reading sequence. Weights of each of the two forward nodes are determined based on the shortest distance and an angle of each of the two forward nodes with the corresponding keyword. A cluster is determined based on the feature matrix using a machine learning clustering model.

Abstract: A method and system of grouping a plurality of documents is disclosed. The method includes determining a plurality of text features of each of the plurality of documents. A plurality of image features of each of the plurality of documents are determined. A layout features set of each of the plurality of documents are determined based on the plurality of text features and the plurality of image features. Each of the plurality of documents in one group from a set of groups is grouped based on determination of a document layout of each of the plurality of documents.

Abstract: A method and system of extracting one or more non-semantic entities in a document image including data entities is disclosed. The methodology includes extraction, by a processor, of row entities and corresponding row location based on a text extraction technique from the document image. The row entities are split into split-row entities based on a splitting rule. Semantic entities are determined from alphabetic entities using semantic recognition technique. The non-semantic entities are determined as split-row entities other than semantic entities. Feature values of each feature type for each of the non-semantic entities is determined. The processor further determines a first probability output for non-semantic entities and a second probability output for semantic entities surrounding the non-semantic entities. The system further labels each of the non-semantic entities based on determination of a highest probability value from a sum of the first probability output and the second probability output.

Abstract: A method and system for relevant data extraction from a document is disclosed. The method includes determining first positional information corresponding to a key from a plurality of predefined keys in the document image based on a deep learning model. Further, second positional information corresponding to the key is determined based on OCR of the document image and an NLP model. Final positional information is determined based on the first positional information and the second positional information, in case a difference between the first positional information and the second positional information is minimal. Relevant data is extracted for the key in the OCR document image based on the final positional information.

Abstract: A method and system for classifying text data in a document based on hierarchy classification is disclosed. A plurality of line regions comprising text data in the document are determined. Positional information and text-characteristic information for each of the plurality of the line regions is determined. For each of the plurality of line regions, a first hierarchy classification from the plurality of hierarchy classifications based on a plurality of predefined rules is determined and a second hierarchy classification from the plurality of hierarchy classifications and a respective probability value based on a machine learning technique is determined. Each of the plurality of line regions are classified based on the first hierarchy classification or the second hierarchy classification. The second hierarchy classification is selected in case the respective probability value of the second hierarchy classification is greater than equal to a predefined threshold.

Abstract: Disclosed herein is system and method for examining relevancy of documents. The system, based on request from the user extracts documents from one or more data sources. The system then obtains from the user, user intention information and user queries. The system then analyses each document with respect to user intention information in order to determine a relevancy level of each document. The relevancy level is indicated in the form of a ranking score. The system ranks and displays the documents to the user in the order of their scores. The system also highlights important excerpts from the documents and provides one or more responses to the one or more user queries submitted by the user for each and every document. Based on the received responses, user provides feedback for further training the system, thereby achieving better accuracy.

Abstract: A method for detecting point anomaly in a text data is disclosed. The method may include tokenizing text input comprising a plurality of text entities into a plurality of tokens, annotating, by the point anomaly detection device, the plurality of tokens, based on one or more annotation parameters. One or more annotation parameters comprise a part-of-speech, a sentiment polarity, a negation statement, and domain rules. The method further includes identifying from one or more annotated tokens, at least one of: one or more anomaly subject tokens, one or more anomaly type tokens, and one or more action type tokens from plurality of tokens. The method further includes generating inferences based on the identified one or more anomaly type tokens, the one or more action type tokens from the plurality of tokens, and the one or more annotation parameters.

Abstract: A method and system of determining shape of a table in a document is disclosed. A region of interest (ROI) from a binarized image of the document is determined is detected corresponding to the table based on detection of a plurality of lines. The ROI is extracted based on a minimum height threshold and a minimum width threshold of the document image. A cluster of points corresponding to each corner of the ROI are determined based on a height of the ROI and contour detection. A corner type of each corner is determined to be one of a 10 pointed corner or a curved corner and in case the corner type of least two corners is determined as the curved corner the shape of the table is determined as a rounded corner structure.