How Zia works?
Zia uses rule-based, syntax-based, and statistics-based methods, as well as neural network models, to detect grammatical and spelling errors, and help you write concisely and clearly.
Deep learning-based grammar check
Zia uses a deep learning-based grammar checker to spot and correct 100+ types of common errors, including subject-verb disagreement, incorrect preposition/article pairings, homophone confusion, and misuse of punctuation. Zia’s skills are constantly updated and fine- tuned in response to new use cases and user feedback.
Natural language processing- based preprocessing and tagging
Zia uses natural language processing (NLP) operations, such as:
- Sentence segmentation, word tokenization, part of speech (POS) tagging, chunking/shallow parsing, and text cleaning, as well as language detection and preprocessing steps, like stemming and stop word removal
- Detecting sentence fragments via dependency parsing
- Identifying important named entities, like people, places, and organizations, to suggest proper capitalization
Machine learning-based corrections
A corpus-based machine learning approach helps Zia co- reference resolutions and detect unclear pronouns. Zia employs trained, separate, multi-class classifiers to categorize error types as subject-verb agreement (SVA), and incorrect verbs, prepositions, articles, homophones, etc.
Rule-based suggestions
Zia includes 200+ predefined rules for spotting gender bias, wordiness, vagueness and hedging, poor word choices, and informal sentences—and offers suggestions to fix them.