The Technology Behind the Understanding
Our technology produces the most advanced and accurate natural language understanding in e-commerce through a combination of state-of-the-art NLP techniques, machine learning, and knowledge engineering.
How It Works
Twiggle uses ontological modeling and advanced linguistic algorithms to drive semantic understanding of both search queries and product listings. Our AI engine’s core components include our Natural Language Analyzer (NLA) and proprietary product ontology, which work together to extract attributes, product types, and categories from raw text. The data is enriched with structured, semantic representations that serve as the building blocks for various applications, including search enhancement, analytics, filtering, and catalog enrichment.
NATURAL LANGUAGE ANALYZER
Turning Words into Meaning
Our patented Natural Language Analyzer applies advanced NLP algorithms to extract ontological categories, product types, and attributes from raw queries and product specifications. It supports more than 30 linguistic phenomena, including ambiguities, typos, inferences, conjunctions, disjunctions, negations, dimensions, and more.
UNIVERSAL PRODUCT ONTOLOGY
The Knowledge Standard for E-Commerce
Our world-class team of AI experts, linguists, algorithm developers, and knowledge engineers have built the first universal product ontology designed specifically for retail. Comprised of thousands of product types, categories, and attributes, our ontology is a rich, highly granular representation of the hierarchical relationships between these entities. Its infrastructure supports the unique phenomena encountered specifically in retail, incorporating hierarchical values (t-shirt is a descendent of tops), attribute types (strings, numbers, booleans, etc.), dynamic values (such as brands), constraints (e.g. booties cannot have a high shaft), and contextual synonym expansion. This enables our ontology to reach the highest level of granularity to understand semantic complexity at scale.
queries processed per second
Understand Product Data
For retailers who manage hundreds of thousands of SKUs, ensuring that your product data is complete, structured, and standardized can make the difference between losing your customer or making a sale. Combining NLP, advanced machine learning algorithms, statistical analysis, and linguistic expertise, our technology enriches, standardizes, and fixes product information from any catalog, achieving best-in-industry precision and recall.
Understand Customer Intent
Behind every search query is a shopper with an intent to purchase a specific product—but whose words are not always understood. Our AI engine is designed to understand the precise intent of even the most ambiguous or complex search query. Whether the shopper is simply looking for a “dress” or for a “long-sleeved red dress for a wedding under $100,” our engine extracts the ontological entities from raw search query data by conducting an advanced semantic and syntactic analysis. Every query is processed through our retail-specific spell corrector, as well as our scope classifier to ensure that queries are interpreted within the scope of the catalog (e.g., a “keyboard” is understood to be a computer accessory, not a musical instrument, in an electronics store.)