Interspeech 2016 - Experiment results for paper "Error correction in lightly supervised alignment of broadcast subtitles"

Abstract

The files in the dataset correspond to results that have been generated for the Interspeech 2016 article: "Error correction in lightly supervised alignment of broadcast subtitles" DOI: 10.21437/Interspeech.2016-56.

The files in the zip file are of two types:
- .ctm, which correspond to the output of the lightly supervised alignment system.
- .sys, which correspond to scoring of the lightly supervised alignment system and includes the overall F measure as well as the precision and recall of the overall system and of each individual show.

The following is a description about the naming convention of the files:

TableX-LineY: This is the alignment and scoring output corresponding to Line Y of Table X in the article.
TableX-LineY-[dev|eval]: This is the alignment and scoring output corresponding to Line Y of Table X for development (dev) or evaluation (eval) in the article.

All two file types are standard outputs that are recognised by the speech technology community and can be opened using any text editor.