Enhancing Online Store Search with Automatic Attribute Recognition and Query Improvement

---

EU-funded project developing an AI search system that understands user intent, improving e-commerce search relevance and customer satisfaction.

Project Description

Online shoppers expect quick and intuitive search functions that understand their intent, even when queries are complex and imprecise. Existing e-commerce search systems often fail to meet these expectations, requiring lengthy manual filtering that leads to frustration and abandoned purchases. We propose the development of an AI-based search system that uses natural language processing and emulates human reasoning by automatically recognizing user intent and key product attributes.

Project Objective

Our goal is to create an enhanced user experience that significantly increases the relevance of search results. This system will improve user satisfaction, boost successful sales rates in e-stores, strengthen customer loyalty, and provide a competitive advantage.

Project Activities and Expected Results

• A1. Developing the solution to TRL-4
• A2. Testing the solution in laboratory conditions
• A3. Managing the innovation cycle
• A4. Preparing documentation for innovation commercialization
• A5. Administrative project management

Concept Feasibility Will Be Demonstrated If Testing Produces Results:

• R1: Recognition of a Predefined Set of Attributes from User Queries The system aims to recognize and identify at least 80% of attributes from a predefined set within user queries. This will confirm the system’s ability to consistently interpret queries and apply relevant search filters.

• R2: Recognition of Store-Specific Attributes from User Queries The target is to achieve recognition of store-specific attributes with over 70% accuracy for most user queries. This will demonstrate the system’s adaptability to different stores and precision in identifying attributes unique to individual product databases.

• R3: User Query Processing and Optimization We will implement a method to improve user queries by adding or replacing key terms. The goal is to increase result relevance by 30% compared to existing search systems that do not integrate this method.

• R4: Sufficiently Fast Search The system will be tested under conditions similar to production to evaluate search speed measured in queries per second (QPS). The goal is to achieve at least 20 QPS under normal load and up to 120 QPS at peak load, confirming system scalability and speed suitable for high-traffic e-commerce applications.

The project aims to reach TRL-4, providing a solid foundation for further development and successful commercialization of this innovative solution in international e-commerce markets.

---

The terms used in the text for masculine persons are neutral and refer to both male and female individuals.
Funded by the European Union – NextGenerationEU.

The views and opinions expressed are those of the author only and do not necessarily reflect the official position of the European Union or the European Commission. Neither the European Union nor the European Commission can be held responsible for them.