RT-04 STT Transcription Guide for Transcribers
The Philosophy of Part-time Transcription
Occasionally, annotators will become antsy, and will begin to rush through files in order to get them done more quickly. In general,this is not helpful to the project. Sloppy work must be corrected at some point, and the time that an annotator can save by first-passing a file quickly is usually outweighed by the time it takes to go behind them and fix mistakes in second, third, or even fourth-passing the files. The same is true of the segmentation, second-passing, and quality control steps of the transcription process.
In addition to this, rushing through a file does not serve the annotator well either, as there are always more files to transcribe. By all means, work quickly, but do so at a pace that still allows for the highest-quality transcription with the fewest possible errors.
The Philosophy of Segmentation
Segmenting is the most intuitive part of the transcription process, particularly in the case of CTS (telephone) files. Annotators simply break up the speech into manageable segments. The definition of manageable is a little subjective, as some annotators will like longer segments than others, and certain speech participants are easier to understand and transcribe than others. But in general, segments should be between two and eight seconds long. Eight seconds is a long segment, particularly if it is densely packed with speech. A good rule of thumb is to put in breakpoints at every obvious opportunity. Every major pause or breath is a good spot for a breakpoint.
If, when looking back, certain sections of the file appear to be too clipped, it is easy to go back and combine them with SHIFT + BACKSPACE. Annotators should be very careful of breaths and muttering when segmenting. It is very easy to end a segment in the middle of a breath. If an annotator is segmenting quickly, listening and creating segments with the RETURN key, they should go back over each segment carefully to ensure that no sounds or breaths are truncated by the segmentation. At the same time, they check to be sure that only a small amount of blank space precedes and follows each segment.
The Philosophy of First-Passing
First-passing is the most important part of the transcription process. It is where the bulk of the work gets done, and it is the first level of quality control, as the annotator doing the first-passing should simultaneously check the quality of the segmentation. The specific pitfalls that annotators can fall into when first-passing have been detailed elsewhere, so this document will address topics more general.
First-passing must be of the highest quality. It is not a draft or a rough. It should be a stand-alone product in which the annotator has made a good effort to render the transcript perfect. Because, even when this standard is adhered to, mistakes will be made. It is often quite difficult to understand what the speech participants are saying. Difficult segments are heard in a different way by each annotator, and it can be tough to correct a segment correctly once an annotator has already misheard it. Thus, the tricky parts of a transcript must be done exactingly the first time, in order to allow as many levels of quality control as possible.
The Philosophy of Second-Passing
This step of transcription is in place as the most important line of quality control. It is important that it not be the step where all the difficult bits of the file are transcribed for the first time. Instead, annotators doing second-passing should be checking to make sure that all of those difficult parts were done properly in the first pass. This includes things like counting the number of stutters and hesitations, making sure that difficult or unintelligible regions are transcribed correctly, and that the file overall makes sense. This is also the chance to catch any missed breaths, laughs, coughs, etc. and to make sure that all sections of speech are properly accounted for. In broadcast news, annotators should pay careful attention to the speaker notations, and ensure that all of the speakers, and the spelling of the speakers' names, are correct.
The second pass is a crucial stage, as even the most scrupulous of annotators will make mistakes on the first pass. At the end of the second pass, the annotator should be confident that the file is perfect, and that they would stand by their work if it had to be shipped out immediately following the second-pass.
The Philosophy of Third-Passing and Transcription Check
There is work to be done even at this stage in the transcription process. This is the step when spelling is checked, when the transcript is checked again to make sure that the second-pass did not miss or create any new errors, and when the contractions are expanded. A word on this. While expanding the contractions is a fairly straight-forward task, it is not without the potential for error. It is important not to expand possessives (e.g. "^Mary's hairbrush" or "^John ^Travolta's increasingly depressing career"). Also, an annotator should check for commonly missed contractions at this stage, searching the transcript for its instead of it's, your instead of you're, their instead of they're, etc.
While receive a final reading after the second-pass, they will not generally be read after the contractions are expanded, as the notation for the contractions breaks up the flow of the file and makes comprehension very difficult. Therefore, contractions must be done properly the first time.
Philosophy: Final Thoughts
There must be a level of pride in an annotator's work. It is not a life-saving job, true. But can still provide satisfaction when done properly. More importantly, LDC has rigid standards of quality that must be adhered to for a project to be considered ready to ship. Any mistakes that are left in the transcript will have to be corrected before it is sent out. This means long, late hours, especially for the full-time staff and project supervisors, who will also have their supervisors waiting anxiously for the finished product. This sort of pressure does not make anyone happy or well-disposed towards their fellow annotators. Therefore, a job done properly the first time is sine qua non for a happy work environment and the continuation of the flexibility that everyone finds so appealing about employment at LDC.
It is also worthy of note at this point that nothing done on LDC projects is anonymous. Annotators leave their digital signatures on every file they touch. Thus, exceptionally sloppy work can easily be traced back to a particular annotator, simply by the nature of the computer system. Annotators are personally responsible for the work that they produce.