UiPath Document Understanding is a powerful framework that allows organizations to automate document processing tasks by extracting relevant data from various types of documents using artificial intelligence (AI) and machine learning (ML) technologies. It provides a set of activities through two main packages: UiPath.IntelligentOCR.Activities and UiPath.DocumentUnderstandingML.Activities. Let’s take a closer look at each of these packages and their functionalities:
UiPath Document Understanding (UiPath.IntelligentOCR.Activities) follows a 6-stage framework for document processing, which includes:
- Load Taxonomy: In this stage, a predefined taxonomy is loaded using the “Load Taxonomy” activity. The taxonomy is a hierarchical structure that defines the categories or types of documents to be processed. It serves as a reference for document classification in the subsequent stages.
- Digitize: In this stage, documents are converted into machine-readable text using the “Digitize” activity. Optical Character Recognition (OCR) techniques are used to extract text and structural information from documents such as scanned images or PDFs. The digitized documents are used for further processing in subsequent stages.
- Classify: In this stage, documents are automatically categorized into different categories based on predefined rules or machine learning models using the “Classify” activity. The documents are classified into relevant groups for further processing based on their content or characteristics. Machine learning models can be trained using the “UiPath.DocumentUnderstanding.ML.Activities” package for more advanced document classification.
- Extraction: In this stage, relevant data is extracted from the documents using the “Extraction” activity. Predefined extraction rules or templates are used to extract data such as names, dates, addresses, and other key information from the documents. Regular expressions or pattern matching techniques can be used for data extraction.
- Validation: In this stage, the extracted data is validated against predefined validation rules or business logic using the “Validation” activity. Data format, data consistency, and data completeness are checked to ensure the accuracy and integrity of the extracted data.
- Export: In this stage, the extracted and validated data is exported into various formats such as Excel, CSV, or JSON using the “Export” activity. This allows the data to be further processed or integrated with other systems as needed.
The “UiPath.DocumentUnderstandingML.Activities” package contains several activities that are used for training and fine-tuning machine learning models for document understanding in UiPath Studio. These activities include:
- Train Classifiers: This activity is used to train machine learning models for document classification. It allows you to provide labeled training data and configure various training parameters such as the model type, feature extraction method, and training algorithm.
- Train Extractors: This activity is used to train machine learning models for data extraction from documents. It allows you to provide labeled training data and configure various training parameters such as the extractor type, feature extraction method, and training algorithm.
- Fine-Tune Classifier: This activity is used to fine-tune a pre-trained machine learning model for document classification. It allows you to provide additional training data and configure various fine-tuning parameters such as the learning rate, batch size, and number of epochs.
- Fine-Tune Extractor: This activity is used to fine-tune a pre-trained machine learning model for data extraction from documents. It allows you to provide additional training data and configure various fine-tuning parameters such as the learning rate, batch size, and number of epochs.
- Train Validator: This activity is used to train machine learning models for data validation. It allows you to provide labeled training data and configure various training parameters such as the validation type, feature extraction method, and training algorithm.
- Fine-Tune Validator: This activity is used to fine-tune a pre-trained machine learning model for data validation. It allows you to provide additional training data and configure various fine-tuning parameters such as the learning rate, batch size, and number of epochs.
- Evaluate Model: This activity is used to evaluate the performance of a trained machine learning model using test data. It provides various evaluation metrics such as accuracy, precision, recall, and F1 score.
UiPath Document Understanding provides a comprehensive framework for document processing that involves loading taxonomy, digitizing documents, classifying documents, extracting relevant data, validating extracted data, and exporting the data for further processing or integration with other systems. The activities provided by the “UiPath.IntelligentOCR.Activities” and “UiPath.DocumentUnderstandingML.Activities” packages offer a wide range of capabilities for automating document understanding workflows in UiPath Studio.
