Hearables Challenge

The National Science Foundation sponsors the Hearables Challenge to develop algorithms or methods that could improve hearing in a noisy setting, especially in the very challenging situation of sustaining conversation where the background noise also includes conversation.

LDC supports research into processing speech in complex noisy environments by developing the Conversations in Restaurants corpus containing multiple, multiparty, goal oriented and informal conversations in a variety of noisy conditions characteristic of restaurants including background conversation. 

The pilot corpus for this effort, LDC2017E31, is available to participants in the NSF Hearables Challenge. It consists of a single continuous segment of 30 minutes and 27 seconds extracted from a larger recording of normal conversational speech in a noisy restaurant environment.  

Challenge participants can obtain this corpus by taking the following steps:

1. Sign the LDC User Agreement and return it to LDC as directed.

2. If you do not already have an LDC user account, register for one.

Once LDC receives the signed agreement and confirms the existence of a user account, we will contact participants with instructions for accessing the data set.