Automate tax computation with machine learning

Overview

In the traditional approach to tax filing, multiple teams of tax specialists source relevant data, and undertake analysis and reconciliation for tax computation. This people-intensive process requires a significant investment of time and effort for rigorous analysis and high quality review before tax filing on an industrial scale.

Infosys blends automation, digitization, and process standardization to transform tax computation for professional services organizations. Our solution leverages automation to extract relevant taxation data from structured, semi-structured, and unstructured documents across multiple sources. Further, we use bots and smart workflows to ensure accurate application of tax rates for diverse financial transactions.

The Infosys Automated Text Extraction (ATE) solution for tax computation capitalizes on Tesseract, an open-source optical character recognition (OCR) engine. It defines elements of tax data relevant for extraction from source documents. This versatile engine uses pre-trained long short-term memory (LSTM) to extract text from images. Significantly, it standardizes the input format of taxation data and eliminates encoding issues.

---

Data discovery and acquisition

• Identify documents from multiple sources
• Extract document metadata
• Create document catalogue

---

Data validation and analysis

• Identify and analyze initial attributes
• Use pattern recognition techniques (NLP) to extract attributes
• Adopt custom document classifier for attributes that do not follow a pattern

---

Solution development and configuration

• Create a repository of normalized attributes referenced in documents
• Create templates with
• custom attributes to ensure configurability

---

Training, review and implementation

• Train machine learning model using historical data
• Undertake manual review of initial set of extracted attributes