Page Banner

NIEUWS

How do we find Technology Transfer and Business Potential in Research?

Introduction

ScoutinScience provides AI-powered solutions to our clients through our Machine Learning platform. One of our core values is that we always seek continuous improvement in everything we do as a team. That is why we want to offer a better experience when looking for technology tranfer in research.

One of the most important features of our dashboard is the ability to see the highlights of a research paper from a technology-transfer perspective. For this to be possible, we started with SciBERT.

dashboard-highlights.png Try it yourself with our free DEMO

What is SciBERT?

SciBERT is a "pre-trained language model based on BERT (Devlin et al., 2018) to address the lack of high-quality, large-scale labelled scientific data. SciBERT leverages unsupervised pretraining on a large multi-domain corpus of scientific publications to improve performance on downstream scientific NLP tasks." Source: SciBERT: A Pretrained Language Model for Scientific Text

However, SciBERT is unable to detect the aspect of tech-transfer potential.

Numerous discussions and interviews with business developers over the last two years made it clear that there is no universally agreed-on definition of tech-transfer potential. They all agree that they can recognize it when they see it. Additionally, most of them agree that they can confidently make a decision based on the granularity of sentences. This gives rise to an obvious idea: show the experts something they can annotate.

How did we train it for tech-transfer?

• 1500 sentences were selected for experts to judge and see if they passed an intention check according to strict guidelines

annotator.png

• The filtering is expected to take out sentences that are not relevant from a technology-transfer perspective • The summaries are derived from sentences selected by the classifier trained on the experts' observations • This method turns the extractive summarization into a binary classification task for which the SciBERT model was fine-tuned.

Results

Our predictions are more accurate. Our insights are more accurate. And our recommendations are more accurate—all thanks to the power of artificial intelligence!

Looking through the papers is much faster. The user is able to see the most important parts and the automated summary.

The AI is now able to look for technology transfer potential with the help of our developers' work

We are the first to use SciBERT for this specific NLP task. The actual value lies in the data collected from experts.

Afbeeldingsispd.png Are you curious to see the results live? Click here to test them for free with our DEMO version!

Deel dit item!

We werken met:

Untitled design-4.png

Contact

Connect-U Ariënsplein 1 7511 JX Enschede Nederland

+31 (0)85 06 00 036

info@scoutinscience.com