This month in Letter to nature https://doi.org/10.1038/s41586-019-1335-8 the researchers demonstrated the possibility of extracting knowledge and relationships from the massive body of scientific literature in a collective manner mining the scientific literature.
They used an algorithm named mat2vec (in python).
Can we create a workflow in KNIME that really automated this?
The code is at: https://github.com/materialsintelligence/mat2vec
Hi @karelman, welcome back to the forum!
From what I gathered so far, this mat2vec is a more specific use case of word embedding for material sciences, right? I think there would be two options: you can get started with the deep learning extensions (here is a useful blog post about word embedding in KNIME: https://www.knime.com/blog/word-embedding-word2vec-explained) or, if you want to use their specialized mat2vec library, use the Python Integration in a KNIME workflow.
Was this useful? Did it answer your question?
Hi @ana_ved mat2vec is not a specific algorithm for material Sciences, the authors use this topic as a probe of concept.
The beautiful thing with mat2vec is that could extract knowledge and relationships for any kind of topic.
Word2vec has a different goal, different functions.
I agree with you, could be used as a Python code, but I have the filling that if we can reconstruct it with Knime nodes could be more understanding and usable for a biggest population.
This topic was automatically closed 182 days after the last reply. New replies are no longer allowed.