Mindbreeze InSpire – The AI-based information hub for your organization

Mindbreeze InSpire leads the industry through its latest AI capabilities. This is done by offering connectors to link over 500 enterprise data sources, providing a centralized and unified view of all business-relevant information and processes, regardless of the location and application of the data. This information is available to the user in a secure and relevant manner via insight services for insight apps, for example, in 360-degree views and by means of generative AI. This information is constantly indexed and kept up to date by our connector technology. In particular, the access authorizations are also mapped and likewise constantly kept up to date. Thus, the change information, as well as the encrypted authorization patterns, also represents an important source for insights, keeping your data up to date and consistently consistent. 

Mindbreeze InSpire combines various retrieval technologies – precise semantic search based on vector retrieval with an expressive relevance model and a highly efficient authorization system

The Mindbreeze InSpire engine provides a platform for a wide range of retrieval technologies, going beyond the capabilities of individual components to provide highly interactive insight applications with integrated artificial intelligence and language understanding.
Mindbreeze InSpire automatically builds vector indices from text areas (sentences, paragraphs, or other regions). In doing so, AI embeddings are generated via so-called transformer models. These models have been specially trained for BI encoding. Vector indexes, often also referred to as a vector database, are not isolated in Mindbreeze InSpire but are seamlessly connected to all the other structures that can be mapped with Mindbreeze InSpire, such as the keyword index, the in-memory graph representation, and the authorization models.

What exactly does “interconnection” mean?

One example of this is that the results of the similarity search can be directly customized using so-called boosting rules based on any search terms or patterns. This makes it possible, for example, to favor answers from more relevant documents. This means that weighting in the embedding models alone is not the only determining factor for the correspondence between the user’s query and the retrieved results since boostings can also be adjusted at the time of the query. The same applies to the authorizations – only the vectors that originate from authorized documents are considered in the vector search – and this is done early on in processing stage to be also used for optimizations.
Another special feature of Mindbreeze InSpire is the continuous synchronization with the contents of the connected data sources. This functionality is available in the same way for the semantic search, so the vector index is always up to date. This is based on the fact that Mindbreeze has decades of experience in processing large amounts of data, and therefore, the vector indices also have a logarithmic compression method.

 

360-degree views – the combination of insight apps and semantic linking

Mindbreeze InSpire offers constant, up-to-date access to all business-relevant information in an enterprise, regardless of where it is stored. The information, which is often stored in heterogeneous locations, is represented centrally and uniformly in the Mindbreeze index. This information can be accessed interactively via Insight Apps.

When Mindbreeze InSpire processes the information, semantic connections between content are recognized. Every type of enterprise information contains references to other information in some form. This can be represented very explicitly by means of links or implicitly, for example, by means of author information, by using product numbers or codes, or by means of references to persons, organizations, places, or other named entities. Mindbreeze recognizes references to entities based on text formatting, the use of codes or abbreviations, or through recognition by AI models, even from unstructured content. The resulting entities can be used to semantically link content. For example, there is an order that has a purchase order number. In addition, there are documents that refer to this purchase order number in an unstructured way. Mindbreeze can interrelate this content and make it navigable and, for example, automatically display relevant content related to a purchase order.

Through this self-learning and automatic linking of relevant information, a navigation system is created through information landscapes. A very powerful way to visualize such a landscape is called 360-degree view Insight App. The great thing about this is that the more information Mindbreeze InSpire processes, the more information it can implicitly link and integrate into the 360-degree views.