Can x2vec Save Lives?

According to Graphika Labs team member Alex Ruch in his latest article published in the Journal of Physics: Complexity, the answer may be yes.

As graph and language embedding models become increasingly standard in large scale analyses, Alex Ruch explores how integrating these models’ complementary relational and communicative data may be used to overcome past hurdles and prove particularly helpful in predicting rare events or classifying members of hidden populations. In his article, Alex shows how merging graph and language embedding models (metapath2vec and doc2vec) extracts unsupervised clustering data without domain expertise or feature engineering.

What does this mean? Take for example mental health support groups, which form in amorphous communities online due to social stigma and comorbidities. Predicting suicidality in the context of prevention among these individuals is typically very difficult due to resource limits as well as other challenges. Ruch’s method of merging graph and language embedding models may offer a better solution to identifying at-risk individuals and even help save lives. 

“One of the things I enjoy most about working at Graphika is having the freedom to work on projects like this and knowing that the models and tools I develop can be used to help with real-world problems while also advancing data science and machine learning methods!” Ruch had to say.

Read the full article here: