ConnExt: Brain Region Connectivity Search Engine

Given a Brain Area or a pair as input, the engine searches a corpus of around 55k full text neuroscience articles for evidences about connected brain areas.

Various algorithms have been trained for extracting connectivity. Following is a brief explanation of each of the models used in the engine

• RegExp patterns : A set of popular connectivity patterns applied to the corpus, with the hypothesis that a sentence matching any of the regular expressions, contains a connected pair of Brain areas.
• Conn Word presence : Presence of connectivity indicators such as the words, Afferent, Connect, Innervate, Originate, Pathway, Project, Receive, Input etc denote connectivity between the pair of brain regions in the sentence.
• BoW clustering : A supervised learning model using Bag-of-Word feature representation.
• Conn Word clustering : A supervised learning model using binary feature representation based on the popular Connectivity Words.
• LPBridge clustering : A supervised learning model using Link parse based bridge as features.
• All Models : Includes results from all the models.

The models based on clustering indicate a confidence score associated with every prediction.