The Explore tool technology is deliberately designed as an interdisciplinary research tool and will present you with articles that you will perceive as noise.

It is built this way by design, in order to increase the chances of serendipity and that you just might come across a new idea, a different way of thinking, or a rabbit hole to jump on for inspiration.

If you find a map has too much noise for your taste, you may be able to improve this by:

• Modify your problem statement by removing some of the more general background information about the problem.
• Setting the relevance score filter (usually set to 40-95%) to a higher range.
• Setting the repository filter only to repositories within your field of study. 
• Using the hierarchy editor to remove the irrelevant concepts. 

The Explore tool is built for quantity of maps as much as quality of maps. 

What we mean is rather than spend a lot of time perfecting one map, we recommend building multiple maps with the most interesting findings from the first map. 

• From our Scithon™ tests, we see that teams who build 15-20 maps over the course of a few hours end up being best positioned to draw conclusions and build a comprehensive overview over spot on papers. 
• If you create your first map and you feel that there is one concept/area missing, you can of course go back and edit your problem statement. However, you can also pick a related paper from the map and build a new map. 

We recommend going slow and doing smaller changes initially using the hierarchy editor. 

The hierarchy editor is a helpful way to edit the output of your map. Keep in mind that using the editor is not the same as changing a set of keywords for a search; what you are doing here is fundamentally altering the fingerprint of the input text. Merging and nesting a lot of concepts may lead to very limited results, and if no papers match well with a section of your fingerprint, it will disappear during the map recreation. 

.

The most interesting relevance score range will vary from project to project. 

The relevance score is set to 40-95% by default. This is because for most common research topics, anything over 95% will be a duplicate, and we keep it to 40% to include some serendipity of noise. Experiment with your relevance filter to find a good range for your current project.

.

Don’t let an incomplete problem statement prevent you from getting started. 

Keep in mind you can always make the starting point a relevant research paper. If you want to work with your own text, we have a lot of recommendations on how to write a good problem statement - for example, 300-500 words being the “sweet spot” and describing the background of the problem as well.  However, the absolute technical minimum is 100 words, and the maximum is 5000 words. Our recommendations are defined in order to get optimal results - but obviously, it’s better to get started than to be perfect. If a comprehensive 500-word problem statement seems daunting, just write up a quick 100 words to get started - or just pick a relevant research paper and use the URL to start.

.

Duplicates happen. 

If a paper is highly relevant in two different subcategories, you might find it duplicated in both. We wouldn’t want you to miss it!

Occasionally, a paper will be published in different repositories with different DOIs and even slightly different titles. We continuously try to improve this without risking removing papers that are not the same but are just very similar. But every now and then there will be duplicates.

.

There are synergies between Explore and Focus

Go here to learn about them.