What is HIBU?

The HIBU platform provides a tried and tested toolset for developing customized solutions in which the extraction of content from structured and unstructured data collections plays a central role.

The basis is a state-of-the-art search and analysis platform, where the core components are constantly being refined. These components apply cutting-edge AI technologies such as statistical models and neural networks, but also classic rule-based approaches. This mix enables the HIBU platform to support a wide range of business cases.

Pre-packaged functions speed up your project

What users expect from a search solution

Modern search solutions for documents and other textual data offer at least a full-text search, which finds “risks” if “risk” is entered, and “swam” if “swim” is entered, for example. In addition, you can expect automatic suggestions during search entry, spelling correction and normally also a result preview.

HIBU offers all these functions – for many languages.

Semantic full-text search with spelling correction
Quick findings with search filters

Quick findings with search filters

With its diverse and flexible search filters, HIBU allows to find relevant information quickly: users can easily start with a simple search query and then limit the number of hits to the documents of real interest.

Search functions for special use cases

A synonym search finds documents that contain terms with the same meaning. HIBU’s multilingual search function also identifies, for instance, French documents for a German search query. The expert search implements the search for specific metadata (e.g. author or date).

Using HIBU’s personal search history, users can easily re-run recent searches. Personal favourites let users quickly return to results of previous searches. The comparison functionality implements a comparison of search results (documents, products, etc.).

Search history
Text analysis functionality of HIBU

Information extraction from texts

HIBU’s extractors identify relevant information in a text (e.g. all occurrences of persons, organizations, locations, invoice data, etc.). The classifiers assign appropriate categories to a text, i.e., a document’s content type (contract, account statement, invoice, user manual, etc.). HIBU’s topic recognition then summarizes a text’s essential character, for instance, “Annual electricity bill 2019”.

Sentiment analysis automatically recognizes opinions in texts, particularly positive / negative expressions relating to a certain topic or product. In addition, HIBU calculates content similarity between two documents and can determine for each document the N most similar documents or detect duplicates.

Extracting information from tables

Standard extraction methods are optimised for continuous text and generally deliver inadequate results for tables, e.g. because the content of a table cell is multi-line or because it can only be interpreted with the help of the associated column heading. HIBU recognises tabular document sections and captures the respective table structure, thus enabling the targeted extraction of table data without loss of information.

Using different approaches, HIBU does this for scanned documents as well as for digitally generated documents and can be configured for a wide range of table layouts (e.g. with vs. without vertical lines).

Information extraction from table
Examples of recommendation functionality

Personalized recommendations

Many users are familiar with the feature of web shops and news sites in which the system automatically suggests other products or content (documents, news articles, etc.) that may also be of interest.

HIBU offers solutions from the individual determination of recommendations for relevant products or content down to the automatic compilation of personalized websites. 

Do you want to see HIBU in action? Request a demo!

write email