Machine learning and cognitive computing
In times of increasing information density and communication culture, the efficient categorization and channelling of data is becoming more and more important. Too many different topics require our attention and the mountains of documents, data or information are growing at dizzying speed. This trend is taking place both at home and at work and the call for modern solutions to stop this chaos is getting louder and louder. The Cognitive Search is a solution approach of a new search, which manages the balancing act between exact and conceptual search.
Artificial Intelligence, Machine Learning and Cognitive Computing are concepts and attempts to enable computers through algorithms to grasp complex facts, process them and make autonomous decisions. Tasks that previously required human examination or complex background knowledge can now be processed much more easily and quickly with the help of machines. It is expected that computers will work independently and after a short training phase already understand how to classify certain information and how, at best, to react to it. Understanding” is transferred to the machine and a form of intelligence is programmed.
From the exact to the conceptual search
Especially the finding of information in sometimes scattered files and systems requires a high search competence, which is based on a broad experience and overtaxes especially newcomers. Search engines for the Internet, such as Google, Bing or Yahoo, have meanwhile established search as a core concept and information in the vastness of the World Wide Web is mainly found via this way. Only rarely do we use edited lists or well-maintained catalogues for our research. Instead, we rely on the technology of large IT groups and their processing of data and information.
What is now the norm in the Internet – namely the comprehensive search in distributed and rarely interconnected data sources – is only gradually becoming established in the working world. intergator offers a comprehensive and modern search solution that offers a multitude of possibilities to quickly and efficiently find information and thus transform it into knowledge. In doing so, we rely on an exact search that returns hits based on terms and sentence fragments that contain them. Filters and facets can be used to limit these and lead quickly and elegantly to relevant information. This approach becomes problematic if data and documents match in content, but the search term used does not appear in them. Here the search reaches its limits and delivers hardly useful to no results.
A new approach to solving this dilemma is the use of machine learning methods. Here, a trained computer independently classifies and categorizes content and data and locates them within a multidimensional vector space. In this way, each data set receives its own position, which is defined by various parameters. Contents are evaluated by a trained machine until the hit rate has reached a defined threshold value. From this point on, the machine works autonomously and classifies information unerringly. If you now search for a term, the search returns not only the exact hits, but also the hits relevant to the content – regardless of whether the term appears in it or not. The exact search is replaced by a conceptual search. In this way, new and sometimes unexpected insights and insights are gained.
Unlike conventional searches with exact terms and search hits, the concept search approach provides a kind of cognitive hit map that locates the data and documents in terms of content. The graphical plot places all hits on a single level and thus visually shows connections and topics. Polyseme or homonymous terms (i.e. words with double or multiple meanings) are interesting. People often recognize their meaning out of context, machines find this more difficult at first. Documents with homonyms or polysemes are easier to locate correctly due to the holistic information approach and are displayed on the cognitive hit map in connection with similar results.
The intergator Cognitive Search is a new approach to combine the classic and exact company search with the conceptual search. Depending on the type of search, intergator offers the possibility to find data from a variety of sources. Content is stored in an index as well as autonomously assigned to a trained machine learning model. The user has the possibility to search in both “worlds” and to get the optimal results quickly.
With intergator 6, this search concept is completely redefined and the Enterprise Search becomes a Cognitive Search that autonomously categorizes and catalogs information. This results in new insights into complex relationships and moves the focus even further towards content-based searches.