MÉTODO IMPLEMENTADO POR COMPUTADOR PARA RECONHECIMENTO DE TEXTO E MÍDIA DE ARMAZENAMENTO LEGÍVEL POR COMPUTADOR

Description

1 / 48 COMPUTER-IMPLEMENTED METHOD FOR RECOGNIZING COMPUTER-READABLE TEXT AND STORAGE MEDIA Field of invention

[001] The present invention is situated in the technical field of intelligent systems, more specifically, intelligent systems for text recognition. Thus, the present invention defines a computer-implemented method for recognizing text and computer-readable storage media. Fundamentals of the invention

[002] Reading and interpreting documents that are difficult to read, such as scanned, handwritten, or watermarked documents, represents a significant challenge for traditional Optical Character Recognition (OCR) systems. These systems frequently fail to provide accurate and reliable results, limiting their usefulness in various applications.

[003] Some problems related to optical character recognition of documents that are difficult to read include: low accuracy in converting text in lower quality documents; high error rate in damaged, overlapping or partially visible characters in a document; difficulty in interpreting context and automatically correcting errors, leading to misinterpretations of ambiguous or malformed words; frequent need for manual intervention to correct transcription errors, prone to human error; and inefficiency in processing documents on a large scale.

[004] Traditional OCR systems use pattern-based algorithms to convert images of text into digital format by comparing character shapes in the image with predefined patterns in their database. However, this approach has limitations, which are discussed in detail below.

[005] Low accuracy in complex documents: OCRs Petition 870250006945, dated 01 / 28 / 2025, page 38 / 97 Traditional 2 / 48 scanners often fail to process low-quality documents because they rely on clear and well-defined standards; visual noise, such as smudges, folds, or shadows, interferes with the correct identification of characters.

[006] Difficulty with typographical variations: accuracy decreases significantly with non-standard fonts or unusual typographical styles, and textual marks, special characters or symbols are often misinterpreted due to a lack of exact correspondence with known standards.

[007] Inefficiency in manuscripts: the variability of human handwriting surpasses the ability to recognize based on fixed patterns, resulting in frequent errors. Historical documents or handwritten notes are rarely processed successfully due to the complexity and variability of manuscript forms.

[008] Interference from graphic elements: watermarks, stamps or decorative elements confuse algorithms, as they are interpreted as part of the text to be recognized.

[009] Lack of contextualization: Traditional OCRs operate at the character or word level, without understanding the semantic context. Interpretation errors in ambiguous or misspelled words are common due to a lack of contextual understanding, resulting in meaningless transcriptions.

[010] Extensive need for post-processing: low precision requires extensive manual review and correction, consuming significant time and resources. The manual correction process is prone to human error, potentially introducing new inaccuracies.

[011] Limitations in content analysis: traditional OCR systems focus only on transcription, without the ability to analyze or extract insights from the content. The interpretation and synthesis of complex information remain tasks Petition 870250006945, dated 01 / 28 / 2025, page 39 / 97 3 / 48 exclusively human. Transcription errors can result in the loss or distortion of critical information in important documents, limiting their use in decision-making processes. Furthermore, extensive manual post-processing and the need for human verification create bottlenecks in document-based workflows. The inability to process complex documents restricts access to information for users dependent on scanned versions, and historical or difficult-to-read documents remain inaccessible for automated digital analysis.

[012] In this context, such limitations indicate the need for a more advanced solution, capable of overcoming the challenges presented by traditional OCR systems and offering a more robust and intelligent approach to processing complex documents. State of the art

[013] The document MENG, F.; WANG, C.-A. Artificial Intelligence and Machine Learning Approaches to Text Recognition: A Research Overview. J. Math. Techniques Comput. Math. 3(3), 01-05 (2024) proposes a set of innovative methodologies designed to significantly improve the accuracy of text recognition models. These methodologies encompass strategies for improving data quality and diversity, optimizing processes for large-scale training and inference, offering comprehensive support for a multitude of languages ​​and typographies, addressing variations in text layout and configurations, achieving accurate handwritten text recognition, and improving the interpretability and explainability of models. The document is quite generic and comprehensive.

[014] Furthermore, US patent 11,481,691 discloses a non-transient readable medium that stores instructions to be executed by a processor in order to make the processor receive Petition 870250006945, dated 01 / 28 / 2025, page 40 / 97 4 / 48 A first trained machine learning model generates a transcript based on a document. Furthermore, the instructions cause the processor to execute the first trained machine learning model and a second trained machine learning model to generate a refined transcript based on the transcript. Additionally, the processor executes a quality assurance program to generate a refined transcript score based on the refined transcript and at least one element of the document or transcript.

[015] Additionally, document BR 102022010870-6 teaches a method for extracting and structuring information, in which the method receives an unstructured document as input, extracts its information, reorganizes it, and makes this information available in files that can be consumed by other systems. This document defines a page separator model for the document, a block detection and segmentation model, a table extractor, an image extractor, an image classification model, a text extractor, a computer vision model for improving the image quality of texts, an optical character recognition model, a spelling correction model, models for semantic enrichment of the text, an organizer of output files, and a metadata aggregator for information enrichment. This document teaches a process that allows the digitization and recognition of handwritten texts and answer sheets in Portuguese using artificial intelligence.The process is geared towards correcting exams, but can also be used for medical prescriptions. The document in question uses optical character recognition (OCR) in addition to an OCR algorithm. Brief description of the invention

[016] One objective of the present invention is to address the challenges of the state of the art through a sequential approach to improving OCR results and contextual interaction with the Petition 870250006945, dated 01 / 28 / 2025, page 41 / 97 5 / 48 documents transcribed.

[017] In general, the present invention has the following advantages: - Effective reading and interpretation of documents previously considered difficult to read, with significantly increased accuracy; - Substantial improvement in automatic error correction, based on deep contextual understanding; - Ability to extract implicit information and make inferences based on document content and general LLM knowledge; - Drastic reduction in the need for manual correction, increasing the efficiency of document processing; - Expanding the capabilities for processing complex documents, enabling sophisticated analyses and extracting insights that go beyond mere transcription.

[018] The proposed invention represents a significant technological advance in the field of document processing, offering a comprehensive solution that not only overcomes the limitations of traditional OCR systems, but also introduces advanced document interpretation and analysis capabilities. This results in tangible benefits in terms of efficiency, accuracy, and applicability in various professional and academic contexts, fundamentally transforming the way information is extracted and used from complex documents.

[019] Thus, the present invention describes a computer-implemented method for text recognition, comprising the steps of: - receiving a scanned file, which includes receiving a scanned file on a processing server; - Include at least one metadata entry for the scanned file, where the metadata includes the file size and file format. Petition 870250006945, dated 01 / 28 / 2025, page 42 / 97 6 / 48 and time of file receipt; - Create a unique identifier for a workflow associated with the scanned file, in a workflow management module; - Extract text and structure the scanned file using Optical Character Recognition (OCR); - Refine and enhance the text extracted by OCR, using the Large-Scale Language Model (LLM) module, which includes semantic analysis of the text for contextual understanding and correction of gaps or inaccuracies in the text; - to reconstruct the text; - Temporary storage of the file in primary memory, by the temporary storage module; - The user decides whether to interact with the interactive chat tool; - Implement an interactive chat interface, using the chat interface module, if the user decides to interact with the interactive chat tool; - Process user queries and generate contextualized responses using a Large Scale Language Model (LLM); and - output a response generated by the Large-Scale Language Model (LLM).

[020] The step of receiving a scanned file also includes verifying that the scanned file format matches a supported format, where the supported format includes any of the following formats: PDF, JPEG, PNG, TIFF, DOC, DOCX, XLS or XLSX or any other format that includes a scanned file.

[021] The step of extracting text and structuring the digitized file by Optical Character Recognition (OCR) includes the use of optical character recognition algorithms, including convolutional neural networks (CNNs), networks Petition 870250006945, dated 01 / 28 / 2025, page 43 / 97 7 / 48 recurrent neural networks (RNNs) and Transformers; and in which extracting text and structuring the digitized file also includes processing a plurality of languages ​​and font styles through the use of pre-trained models on a plurality of texts of different languages, styles and formats.

[022] The step of refining and improving the text extracted by OCR also includes identifying inconsistencies and ambiguities in the extracted text based on the explicit guidelines given to the model; spelling and grammar correction based on the document context; reconstruction of lists and tables; adjusting connectors and transitions between paragraphs to improve text flow; harmonizing styles and tones throughout the document; and adjusting the level of formality and terminology based on the document type.

[023] The step of reconstructing the text (5) which includes: - Transform refined text into Markdown format, preserving the document's structural hierarchy (titles, subtitles, paragraphs); - Convert text elements to Markdown syntax, including: formatting ordered and unordered lists, bold and italic emphasis; creating internal and external links, where applicable; formatting quotations and code blocks, if present; - Convert identified tables to Markdown table format; - Align columns and format table headers; - Apply consistent styles to improve readability; - Insert line breaks and appropriate spacing.

[024] The step of temporarily storing the file in primary memory includes associating the stored file with a user session.

[025] Furthermore, the method also includes: - display the file in the user interface, which includes Petition 870250006945, dated 01 / 28 / 2025, page 44 / 97 8 / 48 conversion of Markdown to HTML format for display in the browser and application of CSS styles to improve readability and appearance.

[026] The step of processing user queries and generating contextualized responses using Large Scale Language Model (LLM) includes: - incorporation of relevant conversation history; - inclusion of relevant parts of the processed document; - Structuring the user's question in a format optimized for the LLM to process; - Inclusion of specific instructions based on the type of query; - Generating a response that considers the complete context of the interaction with the document.

[027] Furthermore, according to another preferred embodiment of the present invention, a computer-readable storage medium is defined comprising, stored therein, a set of computer-readable instructions, which when executed by a computer, executes the computer-implemented method for text recognition. Brief description of the figures

[028] In order to complement the present description and to obtain a better understanding of the characteristics of the present invention, and in accordance with a preferred embodiment thereof, figure 1 is presented in the appendix, where its preferred embodiment is represented in an exemplary, though not limiting, manner.

[029] Figure 1 illustrates the flowchart of the computer-implemented method for text recognition.

[030] Figure 2 shows the first document, which is an English version of the United States Declaration of Independence. Petition 870250006945, dated 01 / 28 / 2025, page 45 / 97 9 / 48

[031] Figure 3 shows the second document, which is a document in Portuguese referring to a Property Registry Registration No. 311.

[032] Figure 4 presents a comparison of the different distance metrics implemented to assess the similarity between the generated transcripts and the reference text.

[033] Figure 5 presents a complementary analysis, focusing on similarity metrics that capture different aspects of the relationship between the transcripts and the reference text of the English document.

[034] Figure 6 presents a comprehensive comparison of the different distance metrics implemented to assess the similarity between the generated transcripts and the reference text of the document in Portuguese.

[035] Figure 7 presents a complementary analysis, focusing on similarity metrics that capture different aspects of the relationship between the transcripts and the reference text of the document in Portuguese. Detailed description of the invention

[036] Firstly, as illustrated in Figure 1, it begins with the presentation of a graphical user interface, which serves as an entry point for the process of processing and interacting with scanned files. In this sense, an interface based on user-centered design principles is displayed, such as a clear layout with visual hierarchy, explicit labels for functions, and logical grouping of related elements. In addition, responsive design elements are incorporated to adapt to different devices and screen sizes. In this sense, options for uploading files are provided via a clickable button, and information on support for different file formats is provided to the user, as well as other guidelines such as the maximum document size, and visual feedback on the file upload progress. Petition 870250006945, dated 01 / 28 / 2025, page 46 / 97 10 / 48

[037] Also, in the graphical interface, there is an option to select the desired quality for the transcribed document and an option to ask the user to choose whether to view the transcript after processing or store it only in memory.

[038] The computer-implemented method for text recognition, comprising the steps of: - receive a scanned file (1), which includes receiving a scanned file on a processing server; wherein the step of receiving a scanned file (1) also includes verifying whether the format of the scanned file matches a supported format, wherein the supported format may include any of the following formats: PDF, JPEG, PNG, TIFF, DOC, DOCX, XLS or XLSX or any other format that includes a scanned file; - Include at least one metadata entry for the scanned file, which includes the file size, file format, and time the file was received; - create a unique identifier for a workflow associated with the scanned file (2), in a workflow management module; - extract text and structure the scanned file (3) by Optical Character Recognition (OCR), which includes the use of optical character recognition algorithms, including convolutional neural networks (CNNs), recurrent neural networks (RNNs) and Transformers; extracting text and structuring the digitized file (3) also includes processing a plurality of languages ​​and font styles through the use of pre-trained models in a plurality of texts of different languages, styles and formats; - refine and enhance the text extracted by OCR (4), using the Large Scale Language Model (LLM) module, which includes semantic analysis of the text for contextual understanding and Petition 870250006945, dated 01 / 28 / 2025, page 47 / 97 11 / 48 correction of gaps or inaccuracies in the text; where the step of refining and improving the text extracted by OCR (4) also includes identifying inconsistencies and ambiguities in the extracted text based on the explicit guidelines given to the model; spelling and grammar correction based on the document context; reconstruction of lists and tables; adjusting connectors and transitions between paragraphs to improve text flow; harmonizing styles and tones throughout the document; and adjusting the level of formality and terminology based on the document type; - reconstruct the text (5) which includes: • Transform refined text into Markdown format, preserving the document's structural hierarchy (titles, subtitles, paragraphs); • Convert text elements to Markdown syntax, including: formatting ordered and unordered lists, bold and italic emphasis; creating internal and external links, where applicable; formatting quotations and code blocks, if present; - Convert identified tables to Markdown table format; - Align columns and format table headers; - Apply consistent styles to improve readability; - Insert line breaks and appropriate spacing; - temporary storage of the file in primary memory (6), by the temporary storage module, which includes associating the stored file with a user session; - display the file in the user interface (7), which includes converting Markdown to HTML format for display in the browser and applying CSS styles to improve readability and appearance; - user decides whether to interact (8) with interactive chat tool; Petition 870250006945, dated 01 / 28 / 2025, page 48 / 97 12 / 48 - implement an interactive chat interface (9), through the chat interface module, if the user decides whether to interact with the interactive chat tool; - process user queries and generate contextualized responses using Large Scale Language Model (LLM) (10), which includes: • incorporating relevant conversation history; • Inclusion of relevant parts of the processed document; • Structuring the user's question in a format optimized for the LLM to process; • Inclusion of specific instructions based on the type of query; • Generating a response that considers the complete context of the interaction with the document; - output a response generated by the Large Scale Language Model (LLM) (11).

[039] After the step of outputting a response generated by the Large Scale Language Model (LLM) (11), the method includes an interaction loop that manages the continuous cycle of interaction between the user and the system, including: - assessment of whether the user wishes to continue or end the process; - updating the context of the conversation; - Restart the query processing module if necessary.

[040] Furthermore, the method also includes a finalization step responsible for the orderly termination of the session and the cleanup of the resources used, including the release of memory allocated for the chat file and history, and the termination of active network connections.

[041] In particular, the step of receiving a scanned file (1) includes the transfer of a scanned document from a user device to a server. Petition 870250006945, dated 01 / 28 / 2025, page 49 / 97 13 / 48 processing. Furthermore, regarding the verification of whether the transferred file is in one of the supported formats, any rejection of an unsupported format is carried out with appropriate feedback to the user in the initial graphical interface. In particular, the file transfer or upload is performed by a secure upload mechanism, potentially using encryption during transmission. The upload progress is monitored and displayed to the user. The scanned file is stored in temporary storage, in an area designated for later processing, such as in the primary memory of the processing server or in some cloud storage.

[042] The workflow management module is responsible for coordinating and overseeing all stages of document processing, from upload to final user interaction.The workflow management module defines the sequence of operations to be performed on the document; allocates computational resources to each stage of the process; implements a queuing system to manage multiple documents in processing; executes tasks in parallel to optimize resource usage; manages dependencies between processing stages; monitors the progress of each stage of the process in real time; records detailed logs for later analysis and troubleshooting; implements recovery mechanisms to handle failures in specific stages; notifies the user in case of critical errors that prevent processing; determines the appropriate time to advance to the next processing stage; and ensures that all necessary steps are completed before releasing the final result.

[043] Specifically, the step of extracting text and structuring the scanned file (3) by Optical Character Recognition (OCR) includes the use of Azure Document Intelligence, but the method can be performed by other Petition 870250006945, dated 01 / 28 / 2025, page 50 / 97 14 / 48 commercial tools, such as AWS Textract, Google Cloud Vision AI, OpenAI GPT-4 Vision, or traditional OCR solutions like Tesseract OCR. The choice of specific tool will depend on project requirements, cost considerations, and resource availability.

[044] The Large-Scale Language Model (LLM) module is responsible for constructing a structured prompt for the language model, which comprises: a. Precise instructions: A set of detailed guidelines that instruct the model on how to process and refine the text; b. Document context: Information about the type, format, and origin of the document; c. Raw OCR text: The content extracted by the OCR process; d. OCR Metadata: Information about the reliability of the recognition and the identified structure;

[045] The prompt is based on: a. Iterative optimization: Refinement based on extensive testing with various types of documents; b. Balancing instructions: A balance between specific guidelines and flexibility to adapt to different contexts; c. Parameter control: Fine-tuning of temperature, top_p, and other model parameters to optimize performance;

[046] The prompt includes specific instructions for: a. Maintain fidelity to the original content; b. Correct spelling and grammatical errors; c. Resolve ambiguities based on the document's context; d. Reconstruct complex textual structures when necessary. Petition 870250006945, dated 01 / 28 / 2025, page 51 / 97 15 / 48

[047] While the current implementation uses Azure OpenAI GPT-40, the method of the present invention is flexible and can be adapted to use other large-scale language models, such as Google Gemini 1.5 Pro, Anthropic Claude 3.5, or Meta LLaMA 3.1. The choice of the specific model will depend on factors such as performance, cost, privacy requirements, availability, and operational costs.

[048] The temporary storage module is responsible for the temporary storage of the processed document in primary memory for fast access during the user interaction session. The temporary storage module reserves space in memory to store the processed document; optimizes allocation to balance performance and efficient use of resources; associates the stored document with the user session; implements mechanisms to release memory when the session is terminated; maintains data integrity during read operations and possible updates; synchronizes access in case of multiple threads or processes.

[049] The chat interface module implements an interactive chat interface to allow the user to ask questions about the processed document. The interactive chat interface includes a message display area with automatic scrolling and a text input field for interaction. In addition, chat session management includes maintaining the history of the current conversation and options to clear or save the history.

[050] It is important to emphasize that, although specific tools and models have been mentioned in the current implementation, the proposed method is fundamentally technology-agnostic. The flexibility of the method allows its adaptation to different technological environments, whether they are based on public cloud solutions (such as AWS, Azure, or Google Cloud), on-premises platforms, or a hybrid combination.

[051] Additionally, the present invention relates to a Petition 870250006945, dated 01 / 28 / 2025, page 52 / 97 16 / 48 computer-readable storage media, comprising, stored therein, a set of computer-readable instructions, wherein when the set of computer-readable instructions is executed by one or more processors, the one or more processors implement the method of the present invention, as described above.

[052] In particular, computer-readable storage media can be memory, which can be non-volatile, such as a hard disk drive (HDD) or a solid-state drive (SSD), or volatile, such as random-access memory (RAM). Furthermore, readable storage media can be any other medium or medium that can carry, store, or record the expected program code in the form of an instruction, a data structure, or a set of instructions and can be accessed by one or more computers or one or more processors, but is not limited to them. Alternatively, readable storage media can be a circuit or any other device or medium that can implement a storage, transport, or recording function, such as a signal or carrier.

[053] Specifically, the computer-readable instruction set represents the algorithm or computer program code or data structure that performs the method of the present invention described above.

[054] The processor can be a general-purpose processor, which can be a microprocessor or any conventional processor or similar. Comparisons of the present invention with other state-of-the-art transcription tools.

[055] Optical Character Recognition (OCR) allows the conversion of physical documents into digital formats. Advances Petition 870250006945, dated 01 / 28 / 2025, page 53 / 97 Recent advances in artificial intelligence and machine learning have led to the development of several OCR methods.

[056] The present invention presents a new OCR method and compares its performance with two state-of-the-art technologies: Azure Document Intelligence and GPT-4 Vision. The comparison focuses on transcription accuracy and semantic adequacy, using two historical handwritten documents as test cases: one in English and the other in Portuguese.

[057] The first document is the English document referring to the United States Declaration of Independence, as shown in Figure 2, chosen for its linguistic complexity and historical importance.

[058] The second document, as shown in Figure 3, is the document in Portuguese, referring to a Property Registration - Registration No. 311, with an initial registration date of June 2, 1977, belonging to the collection of the Electronic Property Registration System (SREI), managed by the National Operator of the Electronic Property Registration (ONR). This document represents a significant example of Brazilian notarial documentation, with its peculiarities of formatting and legal terminology. Data relating to the names of owners and other members, CPF, RG and other identifying elements have been hidden by virtue of the General Data Protection Law (LGPD), and replaced by XXXXX (for full names) or each digit replaced by an equivalent X.

[059] Both documents provide challenging benchmarks for evaluating transcript quality across all three methods.

[060] The comparison evaluates character-level accuracy, word recognition rates, and preservation of semantic coherence. By analyzing the performance on these complex documents, we aim to demonstrate the effectiveness of the method of the present invention in handling textual content in different languages ​​and Petition 870250006945, dated 01 / 28 / 2025, page 54 / 97 18 / 48 contexts, potentially advancing the field of document digitization and information extraction.

[061] The specific objectives of the comparison are: - To quantify the literal accuracy of the present method in document transcription, comparing it with Azure Document Intelligence and GPT-4 Vision; - To evaluate the semantic preservation capacity of the present method in comparison with the reference solutions. - To evaluate the technological effectiveness of the present method versus established tools in the transcription of documents of various formats (PDF, images); and - To establish, through quantitative metrics, the performance differences between the present method and existing solutions for the accurate and semantically faithful transcription of documents.

[062] The central hypothesis posits that the present method exhibits superior document transcription capabilities, evidenced by: - Reduced error rate in the literal transcription of words; - Enhanced preservation of the original text's semantics; and - Improved performance across various textual and semantic similarity metrics. Azure Document Intelligence

[063] Azure Document Intelligence, formerly known as Form Recognizer, is an artificial intelligence service from Microsoft Azure that uses machine learning models to extract and analyze content from various documents. The Read model specializes in OCR and can process printed and handwritten text in over 164 languages. The Read model employs deep convolutional neural networks to detect and recognize characters, even under challenging conditions such as varying lighting or image distortions. Key features of Azure Document Intelligence include: Petition 870250006945, dated 01 / 28 / 2025, page 55 / 97 19 / 48 - Ability to extract text, tables, and document structures from scanned documents and photographs; - Support for multiple file formats, including PDF, JPEG, and PNG; - High-precision recognition of handwritten text; - Easy integration with other Azure services and RESTful APIs. GPT-4 view.

[064] GPT-4 Vision is an advanced multimodal model developed by OpenAI, capable of processing and analyzing both text and images. This model represents a significant advancement in AI, combining the powerful language understanding capabilities of GPT-4 with computer vision technology. Key features of GPT-4 Vision include: - Ability to interpret and describe complex visual scenes; - Understanding text in images, including handwritten content; - Ability to process various image formats and PDFs; - Contextual understanding of visual elements and their relationships.

[065] GPT-4 Vision uses a Transformers-based architecture, allowing it to attend to the different parts of an image and correlate them with textual information. This enables the model to perform tasks such as answering visual questions, captioning images, and transcribing documents.

[066] For OCR tasks, GPT-4 Vision leverages its visual processing capabilities to identify regions of text within an image or PDF and then applies its language understanding to transcribe and interpret the content. Potentially, this approach offers advantages in handling complex layouts, varied fonts, and contextual text interpretation. Petition 870250006945, dated 01 / 28 / 2025, page 56 / 97 20 / 48 The method of the present invention

[067] The present method combines image processing, text segmentation and character recognition techniques with large-scale language models. Validation Metrics

[068] The transcription methods were evaluated against a standard reference (ground truth) using a diverse set of metrics. These metrics range from word error rate to vector space representation distances and measures of textual similarity. Both the reference transcription and the outputs of each method are provided in the item named Appendix, below. Metrics Word Error Rate (WER)

[069] The Word Error Rate (WER) is a fundamental metric in the evaluation of OCR and speech recognition systems. The WER quantifies the edit distance between a generated word sequence (hypothesis) and the reference sequence, normalized by the number of words in the reference, given by the equation below: WR = (S + D + I) / N, where: S = number of substitutions; D = number of deletions; I = number of insertions; and N = number of words in the reference.

[070] Lower WER values ​​indicate better performance, with 0 being perfect and 1 (100%) indicating a complete error. Metrics Based on Embeddings

[071] Text embeddings for the following metrics were generated using the OpenAI text-embedding-ada-002 (version 2) template: Normalized Manhattan Distance: Sum of the differences Petition 870250006945, dated 01 / 28 / 2025, page 57 / 97 21 / 48 absolute similarity between embedding elements, normalized by the sum of the absolute values ​​of the reference embedding. The smaller the distance, the greater the similarity. The best possible value is 0; Euclidean distance: The square root of the sum of the squared differences between embedding elements. The smaller the distance, the greater the similarity. The best possible value is 0; - Minkowski distance: Generalization of Manhattan (p=1) and Euclidean (p=2) distances. The smaller the distance, the greater the similarity. The best possible value is 0; - Chebyshev distance: Maximum absolute difference between corresponding embedding elements. The smaller the distance, the greater the similarity. The best possible value is 0; - Bray-Curtis distance: Sum of absolute differences divided by the sum of absolute values. The smaller the distance, the greater the similarity. The best possible value is 0; - Cosine Similarity: Cosine of the angle between embedding vectors, ranging from -1 (opposite) to 1 (identical). In this case, the higher the value, the greater the similarity. The best possible value is 1; - Spearman correlation: Monotonic correlation between the classifications of the embedding elements. The closer to 1, the stronger the positive correlation and therefore the greater the similarity. The best possible value is 1; Kendall's Tau: Measures the ordinal relationship between pairs of embedding elements. The closer to 1, the greater the agreement and therefore the greater the similarity. The best possible value is 1.

[072] This diverse set of metrics provides a Petition 870250006945, dated 01 / 28 / 2025, page 58 / 97 22 / 48 comprehensive evaluation of the performance of each transcription method, capturing various aspects of textual similarity and semantic preservation. Results Document in English - United States Declaration of Independence Word Error Rate (WER)

[073] The Word Error Rate (WER) is a crucial metric for evaluating transcription accuracy, measuring the editing gap between the generated text and the reference. Analysis of the results revealed the following findings, elucidated below.

[074] The method of the present invention: Shows best performance with a WER of 0.12, indicating that approximately 12% of the words would need correction to exactly match the reference text.

[075] Azure Document Intelligence: Ranks second with a WER of 0.28. About 28% of the words in its transcript require some form of editing to achieve a perfect match.

[076] GPT-4 Vision: Registers a WER of 0.38, indicating that approximately 38% of the words need correction. It is important to note that GPT-4 Vision failed to transcribe the entire text due to output token limitations, which significantly impacted its performance on this task. Distance Metrics

[077] Figure 4 presents a comprehensive comparison of the different distance metrics implemented to assess the similarity between the generated transcripts and the reference text. Analyzing the results shown in Figure 4, consistent patterns are observed, which are presented below.

[078] The method of the present invention: It presents Petition 870250006945, dated 01 / 28 / 2025, page 59 / 97 23 / 48 consistently represents the smallest distances across all metrics evaluated, indicating the greatest similarity to the reference text.

[079] Azure Document Intelligence: Occupies an intermediate position in terms of performance, with distance values ​​greater than the present method, but less than GPT-4 Vision.

[080] GPT-4 Vision: Presents the largest distances in all metrics. These high values ​​reflect incomplete transcription due to token limitations, resulting in a greater disparity between the transcript and the reference text. Similarity Metrics

[081] Figure 5 presents a complementary analysis, focusing on similarity metrics that capture different aspects of the relationship between the transcripts and the reference text of the English document. Analyzing the results shown in Figure 5, the following patterns are observed, highlighted below.

[082] The method of the present invention: It presents the highest values ​​in all similarity metrics evaluated, indicating strong structural and semantic correspondence with the reference text.

[083] Azure Document Intelligence: Demonstrates robust performance, with similarity values ​​close to, although slightly lower than, those of the present method.

[084] GPT-4 Vision: Presents the lowest similarity values ​​across all metrics. These low similarity values ​​are largely attributable to incomplete transcription due to output token limitations, significantly impacting its performance in capturing the complete content and structure of the original text.

[085] The superior performance of the present method in these metrics highlights its ability to preserve not only the Petition 870250006945, dated 01 / 28 / 2025, pp. 60 / 97 24 / 48 literal content, but also the structure and semantic relationships present in the original text. The good performance of Azure Document Intelligence suggests its effectiveness in capturing important aspects of the text, although to a lesser degree than the present method. The performance of GPT-4 Vision was significantly hampered by its inability to process the entire document due to token limitations, highlighting a key limitation of this approach for long document transcription tasks. Document in Portuguese - Property Registry - Registration No. 311 Word Error Rate (WER)

[086] The method of the present invention: Shows the best performance, presenting a WER of 0.50.

[087] Azure Document Intelligence: Shows the worst performance, with a WER of 0.60.

[088] GPT-4 Vision: Registers a WER of 0.57, showing performance close to Document Intelligence, although slightly superior. Distance Metrics

[089] Figure 6 presents a comprehensive comparison of the different distance metrics implemented to assess the similarity between the generated transcripts and the reference text of the document in Portuguese. Analyzing the results shown in Figure 6, we also observe consistent patterns, as highlighted below.

[090] The method of the present invention: Consistently presents the smallest distances in all evaluated metrics, indicating the greatest similarity to the reference text.

[091] Azure Document Intelligence and GPT-4 Vision: They exhibit very similar performance, both inferior to the method Petition 870250006945, dated 01 / 28 / 2025, page 61 / 97 25 / 48 was proposed, given that the distances are greater in all metrics analyzed. Similarity Metrics

[092] Figure 7 presents a complementary analysis, focusing on similarity metrics that capture different aspects of the relationship between the transcripts and the reference text of the document in Portuguese.

[093] The method of the present invention: It has the highest values ​​in all similarity metrics evaluated, indicating greater structural and semantic correspondence with the reference text.

[094] Azure Document Intelligence and GPT-4 Vision: Similar to distance metrics, the performance of both models is very similar, being consistently lower than that of the proposed method.

[095] The superior performance of the present method in the metrics analyzed for the Portuguese document highlights its ability to preserve not only the literal content, but also the structure and semantic relationships present in the original text. This is particularly noteworthy considering that the document contains specific legal terminology and formatting characteristic of Brazilian notarial records. Both Azure Document Intelligence and GPT-4 Vision demonstrated inferior performance, with WER above 0.57, and presented very similar results to each other in all metrics evaluated. This consistent pattern of inferior performance can be attributed to the specific complexity of the Portuguese document, including its particular formatting and technical-legal terminology, aspects in which the method of the present invention demonstrated greater robustness.It is interesting to note that, even without the token limitations observed in the English document, GPT-4 Vision failed to significantly outperform. Petition 870250006945, dated 01 / 28 / 2025, page 62 / 97 26 / 48 Azure Document Intelligence, and consequently did not outperform the proposed method.

[096] Thus, the comparisons carried out, using a diverse range of metrics including Word Error Rate (WER), distance measures and similarity indices, demonstrate the superior capability of the present method in manuscript transcription tasks.

[097] The present method exhibited exceptional and consistent performance across all metrics evaluated, as elucidated below.

[098] Literal Accuracy: The present method achieved the lowest WER, indicating high fidelity in the word-for-word transcription of the original text.

[099] Semantic Similarity: In all distance metrics, the present method consistently showed the lowest values, evidencing strong semantic correspondence between its transcripts and the reference text.

[100] Structural Correlation: In similarity metrics, including cosine similarity, Spearman correlation, and Kendall's Tau, the present method achieved the highest scores, demonstrating a superior ability to preserve the structure and contextual relationships of the original text.

[101] This consistently superior performance across multiple evaluation dimensions underscores the effectiveness of the present method in manuscript transcription. The method of the present invention not only accurately captures the literal content of the text, but also robustly preserves the semantic and structural aspects of the original manuscript. In summary, the present method emerges from this study as an exceptionally effective and reliable tool for manuscript transcription. Petition 870250006945, dated 01 / 28 / 2025, page 63 / 97 27 / 48 Appendix Azure Document Intelligence transcript of the English document - United States Declaration of Independence In Congress, July 4, 1776. Che unanimous Declaration of the thirter united States of America, best in the Course of human events, it becomes necessary for one people to difsolve the political bands which have connected them with another, and to~ assume among the flowers of the earth , He feparate and equal ftation to which the Laws of . Nature and of Nature's God ontitle them , a decent respect to the opinions of mankind requires that they Should declar the caufes which impul them totie fiparatien.We hold these truths tobe felf evident, that all men are created equal, that they are endowed by their Creator with certain unalienable Rights , that among these are Life, Liberty and the pursuit of Happiness_ That to secure these lights, Governments are instituted among Men, deriving their just flowers from the consent of the governed, _ That whenever any Form of Government becomes destructive of these ends , it is the Right of the People to alter or to abolish it, and to institute new Government, laying its foundation on such principles and organizing its powers in such form, as to them shall seem most likely to efect their Safety and Happiness. Prudence, indeed, will dictate that Governments long established Should not be changed for light and transient causes ; and accordingly all experience hath fhewn, that mankind are more disposed to puffer, while evils are Sufferable, than to right themselves by abolishing the forms to which they are accustomed.But when along train of abuses and usurpations, pursuing invariably the same Object wines a design to reduce frem under absolute Despotism ; it is their right, itis their duty , to throw of such Government. Petição 870250006945, de 28 / 01 / 2025, pág. 64 / 97 28 / 48 , and to provide nuo Guards for their future fecurity. _ Such has been the patient fuferance of these Colonies ; and fuch is now the necessity which constrains them to alter their former Systems of Government. The history of the present King of Great Britain is a history of repeated injuries and ufufutions, all having in direct object the establishment of an absolute Tyranny over these States . To prove this, let Facts be Submitted to a cand world. He has refused his afsent to Laus , the most wholesome and necessary for the public good. He has forbidden his Governors topafs Laws of immediate andfeeling importance, unlefs fuspended intheir operation till his Afsent should be obtained; and when so Suspended, he has utterly neglected to allend to them _ He has refused to pays other Laws for the accommodation of large districts of people, unless those people would relinquish the right of Representation in the Legislature, a right ineflimable tothem and formidable to tyrants only .. .He has called together legislative bodies at places unusual, uncomfortable, and distant from the depository of this public Records, for the sole purpose of fatiguing them into compliance with his measures . _ He has dissolved. Repristative Houses repeatedly , for opposing with manly firmnefs his invasions on the rights of the people ._ He has refused for along time, after such difsolutions, to cause others to be elected; whereby the Legislative flowers, incapable of Annihilation, have returned to the People at large for their exercise; the State remain ing in the mean time exposed to all the dangers of invasion from without , and convulsions within ?_ He has endeavoured to prevent the population of these States ; for that purpose obstruc. ting the Saus for naturalization of Jorigners ; refusing topafs cthurs to encourage their migrations hither , and raising Petição 870250006945, de 28 / 01 / 2025, pág. 65 / 97 29 / 48 the conditions ofnew appropriations of Sands. He has obstructed the Administration of Justice, by refusing his afsent to Laws for establishing Judiciary flowers. He has made Judges dependent on his Walk alone, for the tenure of their ofics, and the amount and payment of their salaries. _ He has erected a multitude of New Ofices, and sont hither farms of Oficers to kanals our people, and eat out this Substance _ He has kept among us, in times of peace, Standing Armies without the Consent of our legislatures ._ He has affected to under the Military independent of and superior to the Civil power? _ He has combined with others tojubject us to a jurisdiction foreign to our constitution, and unacknowledged by our laws ; giving his sent totheir Acts of pretended Legislation : _ For quartering large bodies of armed troops among us : For protecting them, by a mock Trial,from punishment for any Murders which they should commit on the Inhabitants of these States : ~ For calling of our Trade with allparts of the world : For imposing Faces on us with out our Consent? _ For depriving us in many cases, of the benefits of Trial by Jury :_ For transporting us beyond Seas to be tried forpretended offences : _ For abolishing the file System of English Laws in a neighbouring Province, establishing therein an Arbitrary government, and enlarging it's Bourdais so as to render it at once an example and fit influment for introducing thesame absolute tule into these Colonies : _ For taking away our Charters, abolishing our most valuable Laws, and altering fundamentally the Forms of our Governments: _ For suspending our own Legislatures, and declaring themselves invested with power to legislate for us in all cases whatsoever .~ He has abdicated Government here by declaring us out ofhis Protection and waging War against us . _ He has plundered our seas,ravaged our boats, bunt our cities, and deshoyed, Petition 870250006945, dated 01 / 28 / 2025, p. 66 / 97 30 / 48 the Lives four people. _ He is at this time hansporting large Armies of foreign Mercenaries to compleat the works of death, desolation and tyranny, already begun with circumstances of Cruelty & perfidy fearcely paralleled in the most barbarous ages, and totally unworthy the Head of a civilized nation. _ He has constrained our fellow Citizens taken Captive on the high Seas tobear arms against their bounty , tobecome the executiones of their friends and Buthun, orto fall themselves by their Hands . _ He has excited domestic infursections amongst us , and has endeavoured to bring on the inhabitants of our frontiers, the merciles Indian Savages, whose known rule of warfare, is an undistinguished deshuction of all ages, fees and conditions. In wery frage of these Oppressions le have Petitioned for Redes in the most humble terms . Our repeated Petitions have been answered by repeated injury.A Rince, whose character is thus marked by every act which may define a Tyrant, is unfit tobe the rules of a free people. Not have We been wanting in attentions to our Brittisk brethren . We have warned them from time totime of attempts by their legislature to extend an unwarrant. able jurisdiction over us. We have reminded them of the circumstances of our migration and Settlement here . We have appealed to their native justice and magnanimity, and we have conjured them by the ties of our common kindred to disavow these ufurpations, which, would inevitably interrupt our connections and correspondence They to have been deaf to the voice of justice and of, consanguinity. We must, therefore, acquiesce in the necessity, which denounces our Separation, and hold them, as we hold the rest of mankind, Enemies in War, in Peace Friends . We, therefore, the Representatives of the united States of Mmerica, in General Congress, Afsembled, appealing to the Petição 870250006945, de 28 / 01 / 2025, pág. 67 / 97 31 / 48 Supreme Judge of the world for the rectitude of our in. tentions, do, in the Name,and by authority of the good People of these Colonies, Solemnly publish and declare, That these United Colonies are, and of Right ought tobe Free and Independent States; that they are absolved from all allegiance to the British brown, and that all political connection between them and the State of Great Britain, is and ought to be totally dissolved; and that as Free and Independent States, they have full Power to levy War, conclude Peace, contract alliances, establish Commerce, and to do all other acts and Things which Independent States may of right do . _ And for the fifthort of this Declaration, with afirm reliance on the protection of divine Providence, we mutually pledge to each other our Lives, our Fortunes and our faired Honor John Hancock Mobmorris Josiah Bartlett n° Whipple gym Hooper Joseph Heures, Benjaminhush Button Gewinnst Lymanistalles Geo Walton. Chil Livingstou Sam Adams Bong. Franklin Saan Lewy John Adams John Denn Samuch Chaves Com Paro Lewis Morris Rob Frear Paint Elbridge Gerry Tho: Stone John Norton Edward Rutledge / . GEO. Taylor James Wilson Auch Nochton StepHopkins T.J.STONE SC. WASHY William Ellery 6 Petição 870250006945, de 28 / 01 / 2025, pág. 68 / 97 32 / 48 George Wythe Thomas Lynch Jun? Ins Wichenpourle Richard Henry Lee Roger Sherman Sam Huntington Tra Hopkinson Arthur Middleton Th Jefferson Casar Pormuy John Hart MY Williams Bony Harrisone The Nelson fi Tho M. Jak Abra Clark Oliver Wohnt Matthew Thornton France's Lightfoot Lee Garler Braxton Transcrição do GPT-4 Vision do documento em inglês Declaração de Independência dos Estados Unidos IN CONGRESS, JULY 4, 1776. The unanimous Declaration of the thirteen United States of America, When in the Course of human events, it becomes necessary for one people to dissolve the political bands which have connected them with another, and to assume among the powers of the earth, the separate and equal station to which the Laws of Nature and of Nature's God entitle them, a decent respect to the opinions of mankind requires that they should declare the causes which impel them to the separation. We hold these truths to be self-evident, that all men are created equal, that they are endowed by their Creator with certain unalienable Rights, that among these are Life, Liberty and the pursuit of Happiness. That to secure these rights, Governments are instituted among Men, deriving their just powers from the consent of the governed. That whenever Petition 870250006945, dated 01 / 28 / 2025, p. 69 / 97 33 / 48 any Form of Government becomes destructive of these ends, it is the Right of the People to alter or to abolish it, and to institute new Government, laying its foundation on such principles and organizing its powers in such form, as to them shall seem most likely to effect their Safety and Happiness. Prudence, indeed, will dictate that Governments long established should not be changed for light and transient causes; and accordingly all experience hath shown, that mankind are more disposed to suffer, while evils are sufferable, than to right themselves by abolishing the forms to which they are accustomed. But when a long train of abuses and usurpations, pursuing invariably the same Object evinces a design to reduce them under absolute Despotism; it is their right, it is their duty, to throw off such Government, and to provide new Guards for their future security.Such has been the patient sufferance of these Colonies; and such is now the necessity which constrains them to alter their former Systems of Government. The history of the present King of Great Britain is a history of repeated injuries and usurpations, all having in direct object the establishment of an absolute Tyranny over these States. To prove this, let Facts be submitted to a candid world. He has refused his assent to Laws, the most wholesome and necessary for the public good. He has forbidden his Governors to pass Laws of immediate and pressing importance, unless suspended in their operation till his Assent should be obtained; and when so suspended, he has utterly neglected to attend to them. He has refused to pass other Laws for the accommodation of large districts of people, unless those people would relinquish the right of Representation in the Legislature, a right inestimable to them and formidable to tyrants only.He has called together legislative bodies at places unusual, uncomfortable, and distant from the depository of this public. Petição 870250006945, de 28 / 01 / 2025, pág. 70 / 97 34 / 48 Records, for the sole purpose of fatiguing them into compliance with his measures. He has dissolved Representative Houses repeatedly, for opposing with manly firmness his invasions on the rights of the people. He has refused for a long time, after such dissolutions, to cause others to be elected; whereby the Legislative powers, incapable of Annihilation, have returned to the People at large for their exercise; the State remaining in the meantime exposed to all the dangers of invasion from without, and convulsions within. He has endeavored to prevent the population of these States; for that purpose obstructing the Laws for naturalization of Foreigners; refusing to pass others to encourage their migrations hither, and raising the conditions of new appropriations of Lands. He has obstructed the Administration of Justice, by refusing his assent to Laws for establishing Judiciary powers.He has made Judges dependent on his Will alone, for the tenure of their offices, and the amount and payment of their salaries. He has erected a multitude of New Offices, and sent hither swarms of Officers to harrass our people, and eat out their Substance. He has kept among us, in times of peace, Standing Armies without the Consent of our legislatures. He has affected to render the Military independent of and superior to the Civil power.He has combined with others to subject us to a jurisdiction foreign to our constitution, and unacknowledged by our laws; giving his assent to their Acts of pretended Legislation: For quartering large bodies of armed troops among us: For protecting them, by a mock Trial, from punishment for any Murders which they should commit on the Inhabitants of these States: For cutting off our Trade with all parts of the world: For imposing Taxes on us without our Consent: For depriving us in many cases, of the benefits of Trial by Jury: For transporting us beyond Seas to be tried. Petição 870250006945, de 28 / 01 / 2025, pág. 71 / 97 35 / 48 for pretended offences: For abolishing the free System of English Laws in a neighbouring Province, establishing therein an Arbitrary government, and enlarging its Boundaries so as to render it at once an example and fit instrument for introducing the same absolute rule into these Colonies: For taking away our Charters, abolishing our most valuable Laws, and altering fundamentally the Forms of our Governments: For suspending our own Legislatures, and declaring themselves invested with power to legislate for us in all cases whatsoever. He has abdicated Government here by declaring us out of his Protection and waging War against us. He has plundered our seas, ravaged our coasts, burnt our towns, and destroyed the Lives of our people.He is at this time transporting large Armies of foreign Mercenaries to complete the works of death, desolation and tyranny, already begun with circumstances of Cruelty & perfidy scarcely paralleled in the most barbarous ages, and totally unworthy the Head of a civilized nation. He has constrained ou. Transcrição do método da presente invenção do documento em inglês - Declaração de Independência dos Estados Unidos IN CONGRESS, JULY 4, 1776 The Unanimous Declaration of the Thirteen United States of America When in the course of human events, it becomes necessary for one people to dissolve the political bands which have connected them with another, and to assume among the powers of the earth, the separate and equal station to which the Laws of Nature and of Nature's God entitle them, a decent respect to the opinions of mankind requires that they should declare the causes which impel them to the separation. We hold these truths to be selfevident, that all men are created equal, that they are endowed by their Creator with certain unalienable Rights, that among these are Life, Petição 870250006945, de 28 / 01 / 2025, pág. 72 / 97 36 / 48 Liberty and the pursuit of Happiness. That to secure these rights, Governments are instituted among Men, deriving their just powers from the consent of the governed. That whenever any Form of Government becomes destructive of these ends, it is the Right of the People to alter or to abolish it, and to institute new Government, laying its foundation on such principles and organizing its powers in such form, as to them shall seem most likely to effect their Safety and Happiness. Prudence, indeed, will dictate that Governments long established should not be changed for light and transient causes; and accordingly all experience hath shown, that mankind are more disposed to suffer, while evils are sufferable, than to right themselves by abolishing the forms to which they are accustomed. But when a long train of abuses and usurpations, pursuing invariably the same Object evinces a design to reduce them under absolute Despotism, it is their right, it is their duty, to throw off such Government, and to provide new Guards for their future security. Such has been the patient sufferance of these Colonies; and such is now the necessity which constrains them to alter their former Systems of Government. The history of the present King of Great Britain is a history of repeated injuries and usurpations, all having in direct object the establishment of an absolute Tyranny over these States. To prove this, let Facts be submitted to a candid world. He has refused his assent to Laws, the most wholesome and necessary for the public good. He has forbidden his Governors to pass Laws of immediate and pressing importance, unless suspended in their operation till his Assent should be obtained; and when so suspended, he has utterly neglected to attend to them. He has refused to pass other Laws for the accommodation of Petição 870250006945, de 28 / 01 / 2025, pág. 73 / 97 37 / 48 large districts of people, unless those people would relinquish the right of Representation in the Legislature, a right inestimable to them and formidable to tyrants only. He has called together legislative bodies at places unusual, uncomfortable, and distant from the depository of their public Records, for the sole purpose of fatiguing them into compliance with his measures. He has dissolved Representative Houses repeatedly, for opposing with manly firmness his invasions on the rights of the people. He has refused for a long time, after such dissolutions, to cause others to be elected; whereby the Legislative powers, incapable of Annihilation, have returned to the People at large for their exercise; the State remaining in the meantime exposed to all the dangers of invasion from without, and convulsions within. He has endeavoured to prevent the population of these States; for that purpose obstructing the Laws for naturalization of Foreigners; refusing to pass others to encourage their migrations hither, and raising the conditions of new appropriations of Lands. He has obstructed the Administration of Justice, by refusing his assent to Laws for establishing Judiciary powers. He has made Judges dependent on his Will alone, for the tenure of their offices, and the amount and payment of their salaries. He has erected a multitude of New Offices, and sent hither swarms of Officers to harass our people, and eat out their substance. He has kept among us, in times of peace, Standing Armies without the Consent of our legislatures. He has affected to render the Military independent of and superior to the Civil power. Petição 870250006945, de 28 / 01 / 2025, pág. 74 / 97 38 / 48 He has combined with others to subject us to a jurisdiction foreign to our constitution, and unacknowledged by our laws; giving his Assent to their Acts of pretended Legislation: For quartering large bodies of armed troops among us: For protecting them, by a mock Trial, from punishment for any Murders which they should commit on the Inhabitants of these States: For cutting off our Trade with all parts of the world: For imposing Taxes on us without our Consent: For depriving us in many cases, of the benefits of Trial by Jury: For transporting us beyond Seas to be tried for pretended offences: For abolishing the free System of English Laws in a neighbouring Province, establishing therein an Arbitrary government, and enlarging its Boundaries so as to render it at once an example and fit instrument for introducing the same absolute rule into these Colonies: For taking away our Charters, abolishing our most valuable Laws, and altering fundamentally the Forms of our Governments: For suspending our own Legislatures, and declaring themselves invested with power to legislate for us in all cases whatsoever. He has abdicated Government here by declaring us out of his Protection and waging War against us. He has plundered our seas, ravaged our coasts, burnt our towns, and destroyed the lives of our people. He is at this time transporting large Armies of foreign Mercenaries to complete the works of death, desolation and tyranny, already begun with circumstances of Cruelty & perfidy scarcely paralleled in the most barbarous ages, and totally unworthy the Head of a civilized nation. Petição 870250006945, de 28 / 01 / 2025, pág. 75 / 97 39 / 48 He has constrained our fellow Citizens taken Captive on the high Seas to bear arms against their country, to become the executiones of their friends and Buthun, orto fall themselves by their Hands. He has excited domestic insurrections amongst us, and has endeavoured to bring on the inhabitants of our frontiers, the merciless Indian Savages, whose known rule of warfare is an undistinguished destruction of all ages, sexes, and conditions. In every stage of these Oppressions, we have Petitioned for Redress in the most humble terms. Our repeated Petitions have been answered by repeated injury. A Prince, whose character is thus marked by every act which may define a Tyrant, is unfit to be the ruler of a free people. Nor have We been wanting in attentions to our British brethren. We have warned them from time to time of attempts by their legislature to extend an unwarrantable jurisdiction over us. We have reminded them of the circumstances of our migration and Settlement here. We have appealed to their native justice and magnanimity, and we have conjured them by the ties of our common kindred to disavow these usurpations, which would inevitably interrupt our connections and correspondence. They too have been deaf to the voice of justice and of consanguinity. We must, therefore, acquiesce in the necessity, which denounces our Separation, and hold them, as we hold the rest of mankind, Enemies in War, in Peace Friends. We, therefore, the Representatives of the United States of America, in General Congress, Assembled, appealing to the Supreme Judge of the world for the rectitude of our intentions, do, in the Name, and by authority of the good People of these Colonies, solemnly publish and declare, That Petição 870250006945, de 28 / 01 / 2025, pág. 76 / 97 40 / 48 these United Colonies are, and of Right ought to be Free and Independent States; that they are absolved from all allegiance to the British Crown, and that all political connection between them and the State of Great Britain, is and ought to be totally dissolved; and that as Free and Independent States, they have full Power to levy War, conclude Peace, contract alliances, establish Commerce, and to do all other acts and Things which Independent States may of right do. And for the support of this Declaration, with a firm reliance on the protection of divine Providence, we mutually pledge to each other our Lives, our Fortunes, and our sacred Honor John Hancock Mobmorris Josiah Bartlett N° Whipple Gym Hooper Joseph Heures Benjaminhush Button Gwinnett Lymanistalles Geo Walton Chil Livingston Sam Adams Benj. Franklin Saan Lewy John Adams John Denn Samuch Chaves Com Paro Lewis Morris Rob Frear Paint Elbridge Gerry Petition 870250006945, dated 01 / 28 / 2025, pp. 77 / 97 41 / 48 Tho: Stone John Norton Edward Rutledge GEO. Taylor James Wilson Auch Nochton StepHopkins TJSTONE SC. WASHY William Ellery George Wythe Thomas Lynch Jun Ins Wichenpourle Richard Henry Lee Roger Sherman Sam Huntington Tra Hopkinson Arthur Middleton Th Jefferson Casar Pormuy John Hart MY Williams Bony Harrisone The Nelson Tho M. Jak Abra Clark Oliver Wohnt Matthew Thornton Francis Lightfoot Lee Garler Braxton Azure Document Intelligence transcription of the document in Portuguese - Real Estate Registry - Registration No. 311 . . OnF BOOK No. 2 Petition 870250006945, dated 01 / 28 / 2025, p. 78 / 97 42 / 48 National Operator of the Electronic Real Estate Registry System - General Registry REGISTRATION NUMBER 311 DATE: June 2, 1977. PROPERTY: Endowment No. 104, of the Colégio Glebe, of the Papucaia Colonial Nucleus, 2nd district of this municipality, rural area, as described: from marker 438, 11.6 meters to 439, with 87 / 13 meters at LV 47° 48' NE, from 439 to 440, 10.54 meters, with 76.66 meters at RV 51° 06' NE, bordering the Macacu River; from 440, 10.54 meters to 11, with 588.00 meters at RV 59° 42' SE, bordering lot 105; from 11 to 14 with... 141.00 meters on RV 57° 24 SW, bordering the road named, from 14 / 90 75 with 115.05 meters on RV 70° 5 P'NIN; from 15 to 16 with 85.32 meters on RV 71° 29 NW; from 16 to 17 with 65.17 meters on RV 69° 42'NW; from 17 to 18 with 56.54 meters on RV 73° 26 NW; from 18 to 19 with 31.72 meters on RV 80° 0}, Nn; from 19 to 20 with 46.77 meters on RV 84° 47 NW, bordering. If as correct. go sapucaia; from 20 to 43811.6 meters with 208.35 meters in the RX39900 NW bordering lot no. 106, with the area 10.10 hectares approximately: Owner: XXXXX, farmer, holder of PF No. XXX.XXX.XXX, Previous prescription: biuro 3.D. Ils. 154, No. 2437. 10% 003. 300 2nd du Oliveira, Brazilian, ca. This document is for reference only and does not serve as a certificate. K:01 / 311. Date: 02 / 06 / 77. Telo, the owner, held a Rural Apothecary Certificate of 1st Frau, four hundred and ten euros), due at Banco do Estado do Rio de Janeiro SA, registered above. Cachoeiras de Macaca, 02 of Petition 870250006945, dated 01 / 28 / 2025, p. 79 / 97 43 / 48 registered 410.00 (twenty-day October 1978, cut and un garant Visualization available at www.registradore do 1 Official : Quelles han AV. 02 / 3 / 1. Date, 27. 11. 18. 1 311, by virtue of which and VALUE: R$ 37.89 of the registration support co of the State of Rio de Janeiro SA and that will remain there Cachoeiras de mazaci Vecce 19fz O Official C 1 K: 03 / 311. Date of the un Cídas quenta vil, frizer photo of 1918. Diego 6 of this city, giving Cachoeiras de O Official 9 nio of the above registered property to of 1st fraction, in the value of cr$ 50,338 € nta . to cruzeiros), with due date via against the Banco do Estado do Rio de Janeiro S / A guarantee o property registered above Operadoo Nacional do Sistema de Registro Eletrônico de Imoveis ce 1778 lia ilst. AV: 04 / 311. Date: 27.01.81. On this date, it is cancelled. Petition 870250006945, dated 01 / 28 / 2025, pp. 80 / 97 44 / 48 The above Registration No. 03 / 311, which, by virtue of the receipt provided by the Banco do Estado do Rio de Janeiro B / A, and which is filed in this registry, the court fee was collected, as recorded in the registry book No. 06 / 11. I mean, No. 01, 1 / s. 014, 206 on No. 06 / 81, on today's date. Cachoeiras de Propaganda, January 27, 1981. Official 3 / 3 on KOS / 311, Data: 27.11.81. The property object of this registration was promised for sale to Mr. Xxxxx Xxxxxx, Italian, Merchant, married under the regime of community of property to Mrs. XXXXX, holders of identity card no. X.XXX.XXX issued by BRA, Ligo, SRE / DOPS / SP on 04.05.73, from @ PF no. XXX.XXX.XXX / XX, residing at Rua Tadel Adel, 186, apt. 101, Leblon, Rio de Janeiro, RJ, by virtue of the public deed of promise of purchase and sale of 27.01.81, drawn up in the registry office of Vila de Subaio, 3rd district and municipality, in book no. 21, Hs. 996 / 298, in the amount of Cr$ 2,300,000.00 (two REQUESTED BY: PETRÓLEO PETROBRÁSSP CPF / CNPJ: . 001.670 DATE: 03 / 07 / 2024 15:23:29 VALUE: R$ 37.89 :selected: Transcription of the GPT-4 Vision document in Portuguese: Real Estate Registry - Registration No. 311 .. ONF BOOK No. 2 11 National Operator of the Electronic Real Estate Registry System GENERAL REGISTRY REGISTRATION No. 311 DATE: June 2, 1977. PROPERTY: Endowment No. 104, of the Colégio Glebe, of the Papucaia Colonial Nucleus, 2nd district of this municipality, rural area, described as follows: from marker 438+11.6 meters to 439, with 87 / 13 meters at LV 47° 48' Petition 870250006945, dated 01 / 28 / 2025, p. 81 / 97 45 / 48 NE, from 439 to 440, 10.54 meters, with 76.66 meters on the r.V. 51° 06' NE, bordering the Macacu River; from 440-10.54 meters to 11 with 588.00 meters on the rv 59° 42' SE, bordering lot 105; from 11 to 14 with 141.00 meters on the rv 57° 24' SW, bordering the Road; from 14 / 90 75 with 115.05 meters on the rv 70° 5' P'NIN; from 15 to 16 with 85.32 meters on the rv 71° 29' NW; from 16 to 17 with 65.17 meters on the rv 69° 42' NW; From 17 to 18, measuring 56.54 meters on rV 73° 26' NW; from 18 to 19, measuring 31.72 meters on rV 80° 0' N; from 19 to 20, measuring 46.77 meters on rV 84° 47' NW, bordering Sapucaia; from 20 to 438 + 11.6 meters, measuring 208.35 meters on rV 39° 00' NW, bordering lot no. 106, with an area of ​​approximately 10.10 ha: Owner: XXXXX, farmer, holder of CPF no. XXX.XXX.XXX, Previous prescription: book 3.D. Ils. 154, No. 2437. 10% 003. 300 2nd of Oliveira, Brazilian, ca. FOR SIMPLE CONSULTATION ONLY, NOT VALID AS A CERTIFICATE. Municipality. TransK: 01 / 311. Date: 02 / 06 / 77. Title of ownership: Rural Apothecary Certificate of 1st fraction, four hundred and ten thousand cruzeiros, due at Banco do Estado do Rio de Janeiro SA, already registered. Cachoeiras de Macacu, October 2, 1978, with guarantee. View available at www.registradore of the 1st Official: Quelles han AV. 02 / 3 / 1. Date: 27.11.18. 1 311, by virtue of which VALUE: R$ 37.89 Petition 870250006945, dated 01 / 28 / 2025, pp. 82 / 97 46 / 48 of the supporting registry of the State of Rio de Janeiro SA and which will be filed in Cachoeiras de Macacu. Officer C 1 K: 03 / 311. Date: 11 / 27 / 81. The property subject to this registration was promised for sale to Mr. XXXXX, Italian, merchant, married under the regime of community property to Mrs. XXXXX, holders of identity card no. X.XXX.XXX issued by BRA, Ligo, SRE / DOPS / SP on 05 / 04 / 73, CPF no. XXX.XXX.XXX / XX, residing at Rua Tadeu Adel, 186, apt. 101, Leblon - Rio de Janeiro - RJ, by virtue of the public deed of promise of purchase and sale of 01 / 27 / 81, drawn up at the registry office of Vila de Subaio, 3rd district of this municipality, in book no. 21, pages 996 / 298, for the amount of Cr$ 2,300,000.00 (two million and three hundred thousand cruzeiros). REQUESTED BY: PETRÓLEO PETROBRÁS-SP - CPF / CNPJ: ***. 001.670 -** DATE: 07 / 03 / 2024 15:23:29 - VALUE: R$ 37.89. Transcription of the method of the present invention from the document in Portuguese - Real Estate Registry - Registration No. 311 OnF BOOK No. 2 National Operator of the Electronic Real Estate Registry System GENERAL REGISTRY REGISTRATION NUMBER 311 DATE: June 2, 1977. PROPERTY: Lot No. 104, of the Colégio Glebe, of the Papucaia Colonial Nucleus, 2nd district of this municipality, rural area, thus described: from marker 438, 11.6 meters to 439, with 87 / 13 meters on LV 47° 48' NE, from 439 to 440, 10.54 meters, with 76.66 meters on rV 51° 06' NE, bordering the Macacu River; from 440, 10.54 meters to 11 with 588.00 meters on rV 59° 42' SE, bordering lot 105; from 11 to Petition 870250006945, dated 01 / 28 / 2025, pp. 83 / 97 47 / 48 with 141.00 meters on rV 57° 24' SW, bordering the unnamed road; from 14 to 15 with 115.05 meters on rV 70° 5' N; from 15 to 16 with 85.32 meters on rV 71° 29' NW; from 16 to 17 with 65.17 meters on rV 69° 42' NW; from 17 to 18 with 56.54 meters on rV 73° 26' NW; from 18 to 19 with 31.72 meters on rV 80° 0' N; from 19 to 20 with 46.77 meters on rV 84° 47' NW, bordering the Sapucaia stream; from 20 to 43811.6 meters with 208.35 meters at rV 39° 00' NW, bordering lot no. 106, with an area of ​​approximately 10.10 ha. Owner: XXXXX, farmer, holder of CPF number XXX.XXX.XXX. Previous registration Book 3.D. Folio 154, No. 2437. 10% 003. 300 2nd of Oliveira, Brazilian. FOR SIMPLE CONSULTATION ONLY. NOT VALID AS A CERTIFICATE. Transcription: K:01 / 311. Date: 02 / 06 / 77. Rural Credit Note. The owner holds a first-degree Rural Mortgage Note in the amount of four hundred and ten cruzeiros, due at the Banco do Estado do Rio de Janeiro SA, in favor of the owner registered above. Cachoeiras de Macacu, October 2, 1978. View available at www.registradoredo1official.com AV. 02 / 311. Date: 11 / 27 / 18. In view of the release provided by Banco do Estado do Rio de Janeiro SA, the above Registration No. 03 / 311 is cancelled, as recorded in book No. 06 / 11, page 014, No. 06 / 81, on this date. Cachoeiras de Macacu, January 27, 1981. Promise of Sale Petition 870250006945, dated 01 / 28 / 2025, pp. 84 / 97 48 / 48 The property subject to this registration was promised for sale to Mr. XXXXX, Italian, merchant, married under the regime of community property to Mrs. XXXXX, holders of identity card no. X.XXX.XXX issued by BRA, Ligo, SRE / DOPS / SP on 04 / 05 / 73, CPF no. XXX.XXX.XXX / XX, residing at Rua Tadel Adel, 186, apt. 101, Leblon, Rio de Janeiro, RJ, by virtue of the public deed of promise of purchase and sale of 27 / 01 / 81, drawn up at the registry office of Vila de Subaio, 3rd district of this municipality, in book no. 21, pages 996 / 298, for the amount of Cr$ 2,300,000.00 (two million and three hundred thousand cruzeiros).

[102] Those skilled in the art will appreciate the knowledge presented here and will be able to reproduce the invention in the embodiments presented and in other variants, covered within the scope of the appended claims. Petition 870250006945, dated 01 / 28 / 2025, pp. 85 / 97

Claims

1 / 4 CLAIMS 1. A computer-implemented method for text recognition, characterized by comprising the steps of: - receiving a scanned file (1), which includes receiving a scanned file on a processing server; - including at least one metadata for the scanned file, which includes file size, file format, and file receipt time; - creating a unique identifier for a workflow associated with the scanned file (2), in a workflow management module; - extracting text and structuring the scanned file (3) by Optical Character Recognition (OCR); - refining and enhancing the text extracted by OCR (4), by the Large-Scale Language Model (LLM) module, which includes semantic analysis of the text for contextual understanding and correction of gaps or inaccuracies in the text; - reconstructing the text (5);- Temporary storage of the file in primary memory (6), by the temporary storage module; - User decides whether to interact (8) with the interactive chat tool; - Implement an interactive chat interface (9), through the chat interface module, if the user decides whether to interact with the interactive chat tool; - Process user queries and generate contextualized responses by Large Scale Language Model (LLM) (10); and - Output a response generated by the Large Scale Language Model (LLM) (11).

2. Method, according to claim 1, characterized in that the step of receiving a scanned file (1) further includes verifying whether the format of the scanned file Petition 870250006945, dated 01 / 28 / 2025, page 86 / 97 2 / 4 corresponds to a supported format, wherein the supported format comprises any format among: PDF, JPEG, PNG, TIFF, DOC, DOCX, XLS or XLSX or any other format that includes a scanned file.

3. Method, according to claim 1, characterized in that the step of extracting text and structuring the scanned file (3) by Optical Character Recognition (OCR), includes the use of optical character recognition algorithms, including convolutional neural networks (CNNs), recurrent neural networks (RNNs) and Transformers; and in that extracting text and structuring the scanned file (3) also includes processing a plurality of languages ​​and font styles through the use of pre-trained models on a plurality of texts of different languages, styles and formats.

4. Method, according to claim 1, characterized in that the step of refining and improving the text extracted by OCR (4) also includes identifying inconsistencies and ambiguities in the extracted text based on explicit guidelines given to the model; spelling and grammar correction based on the document context; reconstruction of lists and tables; adjusting connectives and transitions between paragraphs to improve text flow; harmonizing styles and tones throughout the document; and adjusting the level of formality and terminology based on the document type.

5. Method, according to claim 1, characterized in that the text reconstruction step (5) includes: - transforming refined text into Markdown format, preserving the structural hierarchy of the document (titles, subtitles, paragraphs); - converting textual elements into Markdown syntax, including: formatting ordered and unordered lists, bold and italic emphasis demarcation; creation of internal and external links, when applicable; formatting of citations and code blocks, if present; - converting identified tables to Markdown table format; - aligning columns and formatting table headers; - applying consistent styles to improve readability; - inserting appropriate line breaks and spacing.

6. Method according to claim 1, characterized in that the step of temporarily storing the file in primary memory (6) includes associating the stored file with a user session.

7. Method according to claim 1, characterized in that the method further comprises: - displaying the file in the user interface (7), which includes converting Markdown to HTML format for display in the browser and applying CSS styles to improve readability and appearance.

8. Method, according to claim 1, characterized in that the step of processing user queries and generating contextualized responses by Large Scale Language Model (LLM) (10) includes: - incorporation of relevant conversation history; - inclusion of pertinent parts of the processed document; - structuring of the user's question in a format optimized for LLM processing; - inclusion of specific instructions based on the query type; - generation of a response considering the complete context of the interaction with the document.

9. Computer-readable storage media, characterized by comprising, stored therein, a set of computer-readable instructions, which when executed by a computer, executes the method as defined in any one of claims 1 to 8. Petition 870250006945, dated 01 / 28 / 2025, p. 88 / 97 4 / 4. Petition 870250006945, dated 01 / 28 / 2025, p. 89 / 97

Images