How does Vantage apply machine learning to the training of Skills?

Vantage allows users to train a new skill at design time and improve existing skills further at runtime using Online Machine Learning. The online learning will accommodate feedback from manual review and automatically train skill. To train a new Document Skill, user uploads and labels the fields on a document in Vantage using the Skill Designer. In the case of a Classification Skill, user uploads and assigns documents to different classes.

ABBYY’s machine learning for classification utilizes a multimodal learning approach considering both text and image features. It is based on Convolutional Neural Network to recognize image patterns, bag of words to take frequency of words and phrases, Recurrent Neural Networks and NLP tokens to get sense from the text, and SVM (Support Vector Machine) makes a final decision on document class using image and text features.

ABBYY’s fast machine learning for data extraction groups similar semi-structured documents together by layouts and trains a specific model for each group. The machine learning technology relies on interest points, like logos, separators, stamps and frequently used document words, to build a heatmap of fields locations. ABBYY’s machine learning for extraction also uses text patterns to understand labels and learn typical fields content.

Was this article helpful?

0 out of 0 found this helpful



Please sign in to leave a comment.