Istella’s data assets are the information foundation that powers the AI ​​platform and builds a cognitive infrastructure capable of enriching organizations’ knowledge base with qualified and selected internal and external data.

From collection to context

Istella’s data assets are external data collected and organized to enrich, alongside with individual organization data, the AI ​​platform’s information base, without affecting the ownership and control of corporate data. This allows the model to work within a richer context and retrieve information useful for providing answers more precisely.

In this way, Istella integrates internal and external data into the platform, selected to enrich the organization’s information base, while respecting the control and protection of such data. The result is a system that provides the client with only more relevant information and concrete support for retrieval, reasoning, and content generation in real-world enterprise contexts.

Qualified, selected, and defined data becomes cognitive infrastructure: the starting point for fueling models, enriching knowledge, and making the AI ​​platform more precise, contextual, and useful.

Istella22

Istella 22 is the dataset with which Istella demonstrates its expertise in ranking and evaluating search models. It was designed to allow fair comparisons between traditional and neural approaches, providing a common benchmark for evaluating the quality of search retrieval and ranking.

This dataset provides a corpus of millions of web documents, 220 handcrafted features, relevance judgments on a 5-level scale, and over 2,000 test text queries, providing a rigorous basis for measuring ranking quality and comparing different approaches on the same data. This makes it both a research tool and a concrete demonstration of the maturity achieved in building infrastructures for ranking and content understanding.

LETOR Dataset

The LETOR dataset represents Istella’s most important public benchmark for Learning to Rank: a dataset designed to train and evaluate models that sort search results based on relevance, testing accuracy, efficiency, and the ability to scale across large volumes of queries and documents.

This dataset provides thousands of queries, 220 features for each query-document pair, and over 10 million examples annotated with relevance ratings from 0 to 4. This dataset allows us to understand how a system sorts results, evaluates relevance, and manages complexity when data becomes large and complex. It is therefore a fundamental tool for realistically measuring ranking effectiveness in information-intensive environments.

Web data

Istella collects online information and transforms it into useful knowledge to enrich the platform’s data. The sources are continuously analyzed and organized, providing a constantly updated database ready to support research, analysis, and AI applications.

Istella collects, analyzes, enriches, and indexes billions of URLs, millions of URLs updated daily, thousands of RSS feeds analyzed regularly, and hundreds of news sites constantly monitored.

The value of this asset lies not only in its scale, but in its ability to provide the platform with a contextual understanding that goes beyond the client’s proprietary data. The web thus becomes a living source of up-to-date information, useful for retrieving what is needed, connecting different sources, and enabling more precise responses.

Multimedia

Istella’s multimedia assets expand the scope of knowledge beyond text and bring visual content to the platform, becoming an integral part of information intelligence. Today, the database includes hundreds of millions of indexed images and videos, with a continuous flow of updates that constantly enrich the platform.

The distinctive value lies in the fact that this content does not remain isolated, but can be linked to documents, web signals, and knowledge graphs, expanding the depth of analysis and the quality of the responses produced by the platform. Multimedia thus becomes a structural component of the data assets, designed to support advanced research and AI applications that must understand the world in ways other than text alone.

Social

Social data also enriches the information base with real-time analysis of digital conversations. The system indexes billions of social interactions and integrates signals that reflect events, opinions, and emerging dynamics.

This foundation serves not only to observe the present, but also to identify emerging trends, recurring themes, and relationships between people, brands, places, and events, fueling analysis and semantic enrichment within the Istella ecosystem. In this sense, social data becomes a strategic source for understanding context, strengthening the knowledge graph, and supporting AI applications that must work with live, rapid, and constantly changing signals.