Query Segmentation via RNNs Encoder-Decoder Framework.

Query segmentation is the task of segmenting a Web query into adjacent phrases, typically keywords that form noun phrases or concepts that are relevant to the search task. In this paper, we describe a research study and some preliminary experiment results for query segmentation via a Recurrent Neural Network encoder-decoder framework on a public benchmark dataset (Webis-QSeC-10). The resulting segmented queries can be used for several downstream tasks such as improving the performance of relevance ranking in search, better understanding of the query intent, and suggesting queries for auto-completion.