With Insight Engine, all of a company's data is stored centrally instead of ...

© Mindbreeze

Because information is stored both in various applications within the company and in the cloud, all company data sources serve as a database. Otherwise, isolated solutions are created, which in turn mean additional work for the user.

... as before in the respective disciplines.

© Mindbreeze

With the help of connectors for data sources and filters for different file formats, numerous different sources can be connected with minimal effort and integrated into the central knowledge management system. This creates a knowledge database that can be queried context-specifically and efficiently for the respective departments, such as maintenance.

In addition to structured data, companies must also be able to extract and link unstructured data - such as emails with attachments, texts, audio and video files.

Extraction of information

Different technologies are used for this purpose:

1. Rule-based methods: Based on certain, clearly recognizable structures, so-called entity recognition can be used to determine certain patterns in the text. Predefined rules are used to assign a document type.
2. Statistical methods: Methods such as 'latent semantic analysis' are usually used to automatically recognize existing similarities and thereby determine any content-related or semantic correlations. This analysis is often also weighted in combination with probabilities.
3. Linguistic methods: This involves concrete language-specific characteristics such as the generation of word variants, the decomposition of compounds, the creation of synonyms and word derivations, the recognition of basic forms and the understanding of individual parts and sentences of a language.

Machine learning and artificial intelligence

The Insight Engine classifies automatically and learns from training documents and corrected documents.

© Mindbreeze

With the help of machine learning, intelligent systems are able to learn from accumulated knowledge from the past. They can therefore classify and assign documents completely automatically. The self-learning algorithm optimizes the unequal treatment of categories based on sample data and thus understands similar classes of information and their category properties.

The Insight Engine scans the incoming documents, extracts and analyzes the information. It then compares this with existing and already processed documents. The intelligent solution recognizes patterns and text combinations and forwards the document to the relevant department. Manual assignment is therefore no longer necessary. As a self-learning system, the program continues to develop with every action it performs. During automatic classification, for example, the system learns not only from the training documents, but also from corrected documents. This capability is particularly useful for companies with a high volume of incoming mail. As a result, processes can be optimized and employees supported efficiently.

Interpret requests

With a 360° view of a component, all information can be viewed.

© Mindbreeze

In order to make optimal use of both the extracted data and its correlations, the user's queries must be correctly understood and interpreted. The following options are available for this:

1. Natural Language Processing (NLP):
   Human language is not always precise, as it is heavily dependent on variable factors such as technical language and dialect. Based on machine learning, NLP is concerned with understanding human language with the help of patterns and word combinations and, for example, translating, interpreting, understanding text content and ideally summarizing it in a dialog.
2. Natural Language Question Answering (NLQA):
   NLQA enables dialog between users and the machine. Search queries no longer consist of mere keywords, but can be interpreted and understood linguistically with the aim of identifying the user's needs in the right context and tailoring search results accordingly.

The visualization of the data is completely customizable. Search Apps' can be used to personalize the display based on user roles, departments or even behaviour and expertise in a 360° view of the information.

Context-specific display

It is essential that these apps are designed to be extremely user-friendly. This means that users who know the information requirements but have no programming knowledge can easily create individual display models themselves.

Nevertheless, it is always ensured that only people with the appropriate rights receive the desired information.

Goal: Pooling and providing knowledge

Artificial intelligence is already incorporated in many applications in various forms today. Intelligent technologies help companies to operate much more flexibly, faster and at a higher quality level. They bundle the existing knowledge in the company into a 360° view, extract the relevant facts and make them available in their entirety.

By comprehensively linking internal and external company information, intelligent systems are therefore also able to provide information on the expected service life of components or systems.

Author:
Gerald Martinetz is responsible for sales in the Insight Engine division at Mindbreeze.