Learning Meta Word Embeddings by Unsupervised Weighted Concatenation of Source Embeddings

Anonymous

Abstract: We propose a method to protect the privacy of search engine users by decomposing the queries usingsemantically \emph{related} and unrelated \emph{distractor} terms. Instead of a single query, the search enginereceives multiple decomposed query terms. Next, we reconstruct the search results relevant to the originalquery term by aggregating the search results retrieved for the decomposed query terms.We show that the word embeddings learnt using a distributed representation learning method can be used to find semantically related and distractor query terms.We derive the relationship between the \emph{obfuscity} achieved through the proposed query anonymisation method and the \emph{reconstructability} of the original search results using the decomposed queries.We analytically study the risk of discovering the search engine users' information intents under the proposedquery obfuscation method, and empirically evaluate its robustness against clustering-based attacks.Our experimental results show that the proposed method can accurately reconstruct the search results for user queries, without compromising the privacy of the search engine users.