I think this paper, Distributed Representations of Words and Phrases and their Compositionality (Mikolov et al. NIPS 2013), is the best to understand why the addition of two vectors works well to meaningfully infer the relation between two words.
[Read More]
This paper,Centroid-based Text Summarization through Compositionality of Word Embeddings (Rossiello et al., MultiLing-WS 2017), is about text summarization based on cetroid, and then they experiment multi-documents and a multi-lingual sigle document.
[Read More]
After reading this paper,The Distributional Hypothesis (MagnusSahlgren., 2008), the discributional hypothesis mean you can estimate the meaning of word from distribution of words in context.
[Read More]