| commit | bfd55f243feb3f04e26ad07aae035475768ada8a | [log] [tgz] |
|---|---|---|
| author | Marco Nelissen <marcone@google.com> | Tue Mar 18 14:00:39 2014 -0700 |
| committer | Marco Nelissen <marcone@google.com> | Wed Mar 19 09:48:18 2014 -0700 |
| tree | ace7974e03aa66e5858c58e3205171d786e0882e | |
| parent | e848bd9abb3bbbd4c66f9fb9d1442f7663d7ba40 [diff] |
Use more tags to help the ICU detector. The detector only gave non-ascii data to ICU. In some cases that could result in very short data, for which ICU would issue a low confidence level for the actual encoding. By padding the data with additional (ascii) tags, we improve accuracy for such files. Becauses this can reduce accuracy in other cases, only do this when the initial confidence is low. b/13473604 Change-Id: I63d932043155c310b0e358cdf2d37787961e94b7