| CPC G10L 15/01 (2013.01) [H04L 5/14 (2013.01)] | 10 Claims |
|
1. A method for testing a full-duplex speech interaction system, applied to a test device, the method comprising:
determining a scene mixed corpus set by mixing a valid corpus set related to a test scene with an invalid corpus set which is unrelated to the test scene;
playing each corpus audio in the scene mixed corpus set to a speech interaction device under test equipped with the full-duplex speech interaction system;
acquiring a work log of the speech interaction device under test, the work log comprising at least a first log and a second log, wherein,
the first log records valid/invalid attributes recognized for each corpus audio and corresponding corpus text, and
the second log records a decision result for each corpus audio, the decision results comprising response result and discarding result;
determining a rejection rate based on a counted number of corpus audios having invalid attribute in the first log and a counted number of discarding results in the second log; and
obtaining number of false responses results according to a counted number of log entries which have false response result in the second log, the false response result referring to the response result that is expected not to be output but is actually output from the speech interaction device under test, and determining a false response rate based on the number of false responses and a total number of the played corpus audios in the scene mixed corpus set.
|