15.01.2019 | Blog Cognitive Search
In the media and also on the IntraFind-website there is often talk about "Cognitive Search as the next generation of Enterprise Search". The term "Cognitive Search" comes from the analysts at Forrester and describes Enterprise Search applications that are based on Artificial Intelligence and thus enable completely new application scenarios beyond the classic full-text search.
In this context, terms such as Natural Language Processing (NLP), Cognitive Recommendation, Knowledge Graph or graph databases are often used. But what value does this mean in concrete terms for the user? In order to convince customers of the product, as a sales person I have to deliver hard facts and good arguments. Fortunately, I am a user myself, because I don't just sell iFinder, I also use it. Using the following scenarios, I would like to explain how iFinder as a cognitive search application enriches my daily work:
A search engine that informs and guides me
When I use iFinder to search for a specific presentation or offer, all results are listed, including the data sources. However, I am often only interested in a small part of the document, such as a certain chart or graphic, which I would like to reuse for the next presentation. In order to find this graphic, I don't need to scroll through the hit list and open each file individually. The preview usually tells me at a glance which PowerPoint presentation contains the desired image. I can then open the document directly from the hit list and copy the image into my new presentation.
iFinder is also able to inform me proactively, for example when a colleague has created and saved a new document and it has been re-indexed by the iFinder. The filter suggestions also guide me in my search and show me further terms to which I can limit my search.
What is Natural Language Processing (NLP)?
A search engine that shows me relationships
But iFinder can do much more than just deliver a hit list. As a cognitive search application, it is able to link information from structured and unstructured data sets in a meaningful way and to set them in relation to each other. For example, if I enter the name of a specific customer in the search field, I get all the relevant information at a glance: The offer we have stored, the most important e-mails, all the presentations we have given to the customer and, of course, the address and contact details of my contact person. In addition, I also see the next appointments I have made with this customer. iFinder offers me the perfect tool for preparing for the next customer appointment - as a point of knowledge that puts data from a wide variety of data silos into context and merges them.
The so-called Knowledge Graph makes it possible for me to see all this information. A Knowledge Graph - for example as an info box next to the hit list, as known from web searches - combines this heterogeneous information and makes it easier to use. Another way to display this information is, for example, to combine it in dashboards or even in specific information cockpits.
A search engine that knows what's important to me
In order to be able to make an offer to the customer, I have to know his requirements and compare them with our product. To do this, I need to know which functions the current releases already support. When I'm looking for it in iFinder, I'm less interested in the open tickets from the development team than in release notes, the technical documentation or even data sheets that I can send to the customer. In this case, the search engine knows what is important to me and places the hits that are relevant to me right at the top of the hit list. Search profiles can be used to specify which data sources or file formats (PDF, PowerPoint) should be given greater relevance and which are less important to me (for example, tickets from Jira). E-mails that have been written between me and a developer about this topic can also be booted up in this way.
The iFinder provides me with a tailor-made hit list that is exactly tailored to my needs. The hits can also be adjusted according to how often my other sales colleagues have clicked on a particular document - according to the following scheme: What was important for my colleagues could also be relevant for me. This is done on the basis of machine learning procedures and is called "use-based relevance" or "cognitive recommendation".
A search engine that understands my language
Whether Siri, Alexa and Co. - digital assistants are used to making search queries in question sentences as well as in everyday life. What has proven to be extremely practical in private life is now also finding its way into the office, because iFinder5 elastic also understands natural language. I don't necessarily have to search for a specific keyword, but can also simply ask a question. This is especially advantageous if I know the topic after searching, but I can't narrow it down that much yet. So I can simply ask the iFinder: "Which PowerPoint presentations on the subject of Contract Analyzers were recently changed by Franz Kögl?" or "Which iFinder offers have been available in the past three months? The natural language search is also useful for colleagues from other departments. If they have little previous knowledge of the topic, they can simply search for it intuitively and ask without having to know what the specific search term or product name is.
Conclusion: A search engine that thinks for itself
The pure full text search has become obsolete. Artificial intelligence and machine learning methods extend the enterprise search engine with numerous new application scenarios.
The intelligent search engine understands my language, adds information to the hit list that I initially hadn't thought of when searching, and adjusts the relevance of the results according to my needs. I find what I'm looking for faster and save a lot of time, so that I can devote myself more intensively to my core tasks again.