Innovation Lab
19 NOV 2026

The Innovation Lab is a dedicated, cross-functional environment designed to explore new technologies, unconventional ideas, and potential future products It allows small teams to experiment quickly, fail fast safely, and develop viable prototypes for future product lines. Right now, a new team of Anixians is working on a new project.

The goal of this project is to evaluate and validate modern approaches to hotel deduplication, with a particular focus on challenging supplier data, especially from bed banks. The project aims to identify technologies and methodologies that can improve matching accuracy, provide measurable performance metrics, and establish a reusable framework for addressing future deduplication challenges.

Hotel data received from multiple suppliers must be accurately grouped and mapped to master hotel entities. Existing solutions are only partially effective, primarily due to their strong reliance on heuristic, rule-based algorithms and fixed-threshold values.

While such approaches perform well for straightforward cases, their effectiveness decreases significantly when dealing with edge cases, incomplete information, conflicting supplier records, or other complex matching scenarios. As a result, achieving consistent and highly accurate hotel deduplication remains a challenging problem.

As part of this initiative, we would like to explore a range of approaches that could help us solve this problem more effectively. Our goal is to evaluate different technologies, compare their performance, and ultimately propose practical solutions, or even demonstrate working prototypes, that could significantly improve the accuracy of hotel deduplication.

Our research areas include:

  • Heuristic algorithms based on existing domain knowledge and focused on detailed analysis of hotel-specific attributes and metadata.
  • Neural network-based solutions, including experiments with the ML.NET framework.
  • Semantic similarity and transformer-based models, such as BERT, to better understand relationships between hotel descriptions, names, and associated metadata.
  • Evaluation of publicly available AI models and platforms, including Jina AI and Llama, to assess their potential applicability to hotel deduplication challenges.

We will post results and our conclusions when we get them! Stay tuned!

Tags:ANIXEInnovation Labhotel deduplicationsupplier databed banks