Figure 4From: Care episode retrieval: distributional semantic models for information retrieval in the clinical domainTraining the RI-Note model. Word w4's context vector, C w 4 ⃗ , is updated by adding the index vector I n o t e 2 ⃗ of the note it is part of. The same update is applied to all other words in the note. As a result, context vectors for words co-occurring in the same note become more similar.Back to article page