Colleen Hunter, Channel Partner Manager, at the start-up Yewno discusses technological innovations that can vastly improve the research experience across the sciences, social sciences, and humanities. Using machine-learning techniques, Yewno offers a new kind of search experience.
2. Yewno 2Strictly confidential, not for distribution without the prior consent of the Yewno
A profoundly new approach of knowledge discovery
for people who like to ask “why?”
4. Yewno 4Strictly confidential, not for distribution without the prior consent of the Yewno
Knowledge Singularity
“What distinguishes knowledge from information is the way in which
knowledge empowers actors with the capacity for intellectual or
physical activity. Knowledge is a matter of cognitive capability and
enables actors to do and reflect. Information, by contrast, is passive
and meaningless to those without suitable knowledge. Knowledge
provides the means by which information is interpreted and brought
to life.”
The Work Foundation’s Knowledge Economy Program interim report (Brinkley 2008)
5. Yewno 5Strictly confidential, not for distribution without the prior consent of the Yewno
The human brain & learning
Initial processing & feature
extraction
Further abstracting &
understanding the “What?”
Semantic content arises &
understanding the “Where?”
Facial recognition requires
simple & complex aggregation
6. Yewno 6Strictly confidential, not for distribution without the prior consent of the Yewno
Yewno understands meaning
Machine Learning &
Computational Linguistics
analyses of raw data on
ingestion
Neural Nets
Topic models
Stochastic learning
Knowledge Extraction occurs &
Concepts are semantically
projected onto a Knowledge
Network representing
correlations
Good afternoon. I’m Colleen Hunter, a Channel Partner Manager with Yewno.
Today’s volume of information is ENORMOUS, which makes so much of it undiscoverable. When you have an idea what you’re looking for, it’s overwhelming. When you don’t know what you’re looking for, it can be nearly impossible.
Yewno is different. The discovery environment is complementary to existing search and discovery layers as it provides graph visualization allowing users to immediately understand correlated topics and concepts, by providing context.
It was noted by The Work Foundation’s Knowledge Economy Program, ”What distinguishes knowledge”…
In the quest towards Knowledge Singularity — Yewno enables a day when all the world’s knowledge is easily accessible and at the fingertips of everyone!
Let’s consider how humans process information and learn. Before being able to recognize the content of an image, the brain extracts features with different levels of complexity:
1. Processing starts at Visual Cortex
2. It gets further abstracted at the IT-Cortex (understand what is in an image)
3. IT-Anterior is where the semantic content arises (spatial orientation and attention)
In case of face recognition, the brain aggregates simple edge-like features to form more complex face-like ones that trigger the recognition
Left image taken from https://grey.colorado.edu/CompCogNeuro/index.php/CCNBook/Perception;
About the two streams (dorsal/ventral) you can read here: Gazzaniga, Ivry & Mangun: Cognitive neuroscience, 2002; Leslie G Ungerleider and James Haxby: ‘What’ and ‘where’ in the human brain, 1994 (http://psych.colorado.edu/~kimlab/ungerleider_haxby.94.pdf)
How does Yewno do it?
We mimic how the brain reads and analyzes information. Our technology ingests full text and applies machine learning and computational linguistics to…
CLICK: extract concepts, create neural networks and topic models
CLICK: to uncover knowledge and create unbiased inferential correlations.
Yewno’s continuous semantic analysis against the knowledge network detects emerging phenomena and presents the most updated knowledge and relationships.
Yewno creates inferences from millions of diverse sources of unstructured data for users. The system can trigger automatic discovery of potential paths (between any number of concepts), increasing the possibility for interdisciplinary exploration.
This is Yewno’s discovery environment.
CLICK: You can explore multiple concepts simultaneously on the graph.
CLICK: You’ll also see a web of connecting lines and nodes between them. These map out correlations between the concepts. Blue nodes have a 1-to-1 relationship with a concept and the orange nodes depict a connection between more than 1 concept.
CLICK: Along the right side, you’ll see there are summaries of related documents and links to full text.
We were thrilled to get this feedback from a Medical faculty in Germany - “I loved to see the connection between blood sugar levels, the circadian rhythm and cancer visualized. To see how they influence each other via different pathways is just great. Plus having the sources directly linked is phenomenal. This would be a great tool to teach students about the complexity of the body and how everything is interconnected.”
Additional feedback came after a professor shared a Yewno graph in advance of a classroom discussion, challenging students to discover something new and present it.
So you know,
CLICK: double clicking on a node explodes that concept to trace correlations between that and the original query/queries.
CLICK: By clicking on a line between any 2 nodes, you can explore the connection to learn more relationships, context, and discover documents that include both concepts.
A student commented, “The graph allowed me to explore both nutritional and ethical related topics to veganism. I learned that a supplement of B-12 is required since it is only found in animals. I also noticed that related journals came up when I clicked on certain terms - very helpful for my research.”
On a similar note, we’re using our technology, applying topic models to all of the full text content we’ve ingested. During the content acquisitions earlier this spring, MIT was eager to see the results it showed of institutional collections and MIT Press books, to compare and enhance the probability of discoverability. This gave way to our newest service, Yewno Unearth, as our algorithms revealed new topics, new subtopics, and concepts at the article and book chapter level to look at the content in a whole new way.
Our Topic Models are being added across all of the content we have from these providers and more, and Yewno welcomes other publishers and aggregators to join us in delivering new possibilities for knowledge discovery. Universities can benefit from sharing OA or proprietary content with users, as Stanford, MIT, and Harvard already have.
There is no ‘one size fits all’ but some combination of Yewno Discover, Unearth, and other tools adds an extra layer of value and discoverability. So, if you don’t know, now Yewno.
Thanks again!