Document Recommendation – Cognitive-guided Knowledge Retrieval

Document Recommendation

The task of document recommendation to knowledge workers differs from the task of recommending products to consumers.

Collaborative approaches, as applied to books or videos, attempt to communicate patterns of shared interest to augment conventional search results. However, it turns out that subtle variations in search context can undermine the effectiveness of collaborative filtering. There are well-known problems with these approaches.

For information seeking, what seems to be required is a recommendation system that takes into account both the user’s query and certain cognitive features from the context. Being able to leverage existing taxonomies and inter-document and inter-library relationships helps to recommend related and similar documents.

The Noggle recommendation engine is optimized and can detect all related documents for a given document. If a document is selected from the search results, the engine pulls up all related or similar documents. Regardless of the filename or file type. The recommendation intelligence is based on full-text/content-similarity deep-search algorithms. It can even pull up new versions of existing documents that have been edited by your colleagues and saved in completely different locations. You can’t locate these documents with simple search queries on your own. Imagine that you find an old PowerPoint document and you want to see the latest version of the document and its Excel calculation sheet. They might be anywhere on the network, but our recommendation engine detects them instantly.

Please review the following example:


How to use the managed TEDTalks library?

This tutorial shows how to use the managed TEDTalks library. The integrated managed TEDTalks library can scan the TEDTalks via the noggle client. Furthermore, you can use all integrated cognitive recommendation features to link public TEDTalks with your individual, personal documents.

Step 1: Select the managed TEDTalks Library

Open the library panel, switch to managed libraries and select the TEDTalks library:

Managed TEDTalks Library

Managed TEDTalks Library




Step 2: Search for specific TED Talks

Now your are able to specifiy search querries to browse and search the TED Talks library from within your client. With a click on the ” Intelligent Open ” button, you will directly forwarded to the page with the respective video talk.



Step 3: Cluster the search results and build cognitive curated playlists via our KnowledgeMap

Noggle can create and cluster all found TED Talks for the subject with the integrated KnowledgeMap feature. By using the integrated cognitive AI processing engine, you can use and browse automatically generated knowledge maps to research and browse the search results.




Step 5: Use the cognitive “Recommender Button” to show related content

If you have found an interesting TED Talk, you can use the “Recommend” button and Noggle will instantly search and present a list with all TED Talks that are related to the selected one. This cognitive retrieval feature also works across different libraries. You can, for example, select different libraries with the Library Manager panel. When you press the “Recommend” button, noggle will also search for related documents from all current selected libraries. This way, you can use public TED talks to retrieve personal documents that are related with a presentation. Or the other way: You can search documents, activate the TED library within the Library Manager panel, and press the “Recommend” button on your document. This way, Noggle will pull up TED talks that are related to your personal document.


What would a robot see in TED Talks? …beautiful TED Talk maps

What would happen if we fed a robot with all the TED talks from the most inspiring leaders in the world over the last 10 years and asked key questions of the robot afterwards? Review beautiful TED Talk maps generated by the Noggle knowledge assistant.

Would the robot answer similarly to how we as humans would? Is current machine-learning and cognitive artificial intelligence able to learn and teach a “dump” robot about our world? No theoretical talk about the future of AI—let’s see what we can get out of this technology today.

So I took the publicly available TED talks from their website (, 2,224 talks as of today, 2006-2016) and used the summaries and transcripts to feed our robot; a state-of-the art machine-learning AI algorithm.

After the robot gathered all the content—which took only seconds—we asked it key questions. The robot uses a cognitive-pattern detection algorithm across all talks, so it looks for common patterns within the presentations of all speakers. The robot builds patterns and presents its findings about key topic clusters in a colorful knowledge map – TED Talk maps. The size of the cluster represents the number of talks that have been assigned to a given thematic cluster. The most important clusters are shown in the center. So behind each cluster is the respective number of TED talks that deal with the cluster name shown.

Our natural language processing “robot” works like a natural human response would have: our brains use cognitive shortcuts to make sense of our increasingly complicated world, and the shortcuts used here by our machine learning algorithm seem to have the same effect. Out of these 520 hours of video, the robot was able to extract the important shortcuts with cognitive text processing.

Please visit and read the full article on LinkedIn here:

Download the spreadsheet with all details here:

OnlineHelp Settings Map Tab

How to fine-tune the KnowledgeMap cognitive AI clustering algorithm

You can specifiy how the cluster lables should be generated and which lables should be excluded by the algorithm. Please go to the Settings -> Map section.


Strong Cluster Label [Enabled/Disabled]:

This attribute may be useful when certain words appear in most of the input documents (e.g. company name from header or footer) and such words dominate the cluster labels. In such case, enabled strong cluster lables may improve the clusters.

Another useful application of this attribute is when there is a need to generate only very specific clusters, i.e. clusters containing small numbers of documents. This can be achieved by enabling strong cluster lables.


A stopword is a word that has little meaning by itself. For example, the, a, then, and towards are stopwords for all English documents. A stopword can never appear by itself as a cluster label, although it might be used within a label, depending on the stoplabel settings.


If a KnowledgeMap label includes one of the stop labels, the label will not appear on the map of clusters produced by Noggle KnowledgeMap.

A Visual History of Human Knowledge | Manuel Lima | TED Talks

A Visual History of Human Knowledge

Manuel Lima | TED Talks

How does knowledge grow? Sometimes it begins with one insight and grows into many branches. Infographics expert Manuel Lima explores the thousand-year history of mapping data — from languages to dynasties — using trees of information. It’s a fascinating history of visualizations, and a look into humanity’s urge to map what we know.