Types of Models

Docsumo provides a range of models designed to streamline document processing and automate data extraction. Each model serves a specific purpose and offers unique capabilities to meet different document processing needs. This support document outlines the various types of models that Docsumo offers, along with their functionalities and limitations.

Base models

  1. Baseline Model:
    The baseline model is a foundational/system default model for reliable data extraction. Trained on diverse data, it's applicable to pre-trained document types, eliminating the need for specific training. This adaptability stems from its exposure to varied data during training, enabling it to extract information from unfamiliar documents.
  2. AI Assist:
    AI Assist is an assistant that comes with Generic learning capabilities. Unlike the previous models, users cannot train AI Assist. Instead, it serves as a helpful tool for faster annotation. AI Assist predicts values for requested fields based on its learning from the generic data set. This feature assists users during the annotation process, suggesting potential data extraction points and reducing manual effort.
  3. One-Shot Learning:
    The One-Shot Learning model is based on your' historic actions and interactions with the platform. It leverages the last 5-10 documents processed for that document type to predict the best possible results for similar document types. Similar to AI Assist, you cannot train this model; it comes with predefined capabilities to enhance efficiency in data extraction.

Advanced Models

  1. COA Classification:
    The COA (Chart of Account) Classification model is specialised for predicting the category of chart of accounts for any line item in financial documents. It is trained explicitly for Profit & Loss and Balance Sheets. With this model, users can streamline the process of categorising financial data accurately.
  2. Key Value Model:
    The Key Value Model is designed to extract key value information from documents. It allows you to train the model using historic data, enabling it to learn from past document examples. You can train the model to accurately identify and extract essential information like names, dates, addresses, amounts, and more.
  3. Table Model:
    The Table Model is intended for extracting tables from documents. Like the Key Value Model, you need to train this model using historic data. By annotating tables in the training documents, you can train the model to recognise table structures and extract data effectively. This model is particularly useful for documents with tabular data, such as invoices, receipts, and financial reports.
  4. Key Value and Table Model:
    The Key Value and Table Model integrates the capabilities of both the Key Value Model and the Table Model. The combine model can be trained using historical data with both key-value pairs and table annotations. It is particularly effective for processing complex documents that contain key values such as names, dates, and amounts, along with recognizing and extracting data from tables. this model ensures thorough and accurate document analysis, making it an indispensable tool for comprehensive data extraction.

Docsumo offers a diverse set of models to cater to different document processing requirements. You can choose the most suitable model basis your needs, whether it's extracting key-value information, tables, or making use of AI Assist to speed up annotation. It's important to understand the capabilities and limitations of each model to make the most of Docsumo's document processing features. With the various models at your disposal, you can significantly improve efficiency, accuracy, and automation in your document processing workflows.

Should you have any questions or encounter any issues during the process, feel free reach out to us at [email protected], and we'll be more than happy to help you.