Talent Data Exchange (TDX) has proprietary IP and workflow that extends beyond data extraction and concentrates on informing business decisions. We have built proprietary NLP models to help TDX extract, train, and continuously learn. TDX leverages optical character recognition (OCR) and machine learning technologies to read your SOWs, Change Orders, SLAs, and MSAs, turning your unstructured agreements into structured data. We follow the following 4 step augmented analytics framework to enable us to extract metadata and convert it into actionable insights. We capture over 100+ attributes (depending on the contract type) using our machine learning models.
1) Metadata Extraction - As part of this step, TDX proprietary AI algorithms will read the document and extract metadata from the record. The AI algorithms are not based on the rules-based model and can self-learn the metadata fields of interest-based on training. TDX AI models do not depend on formats and the place of the metadata in the document. They are intelligent to understand the field of interest and extract the metadata from any document (post-training).
2) Metadata Classification - Once the metadata is extracted, TDX can classify the metadata into appropriate fields (Text, numerals, dates, etc.) to be presented in the product interface. The classification models can understand the information hierarchy and offer the information for interpretation and analysis.
3) Data interpretation - Detailed metadata is further analyzed, and interpretations are presented for various use cases, E.g., Charts for enabling visibility, understand the difference between the contractual clauses in numerous documents, comparing the contractual language and versions within a given document, etc.
4) Product Insights - Insights are presented in a seamless workflow-based application that will help navigate the user to appropriate analysis. The goal of the product insights is to enable decision making through the process of review, comparisons, and judgment.
Talent Data Exchange (TDX) leverages several machine learning techniques to extract data and create insights. An illustrative list of our key models are Whitespace, Similarity, Topic, and Classification. In addition, Brightfield Domain Experts (Human Curation) provide extensibility and learning continuously.