Workflow

When undertaking any task, scientific or otherwise, it is always a good idea to have a plan that states which tasks you are going to do, and in which order it would be most convenient to do them. But there are different levels of detail in such plans, or lists of  tasks; some take a bird eye's view and helps get a look at the entire state of things and to see if anything important is missing somewhere; I like to call this bird eye's view plan my Game Plan. At a different level of detail we have the kind of plan that states every little detail that should be taken care of - this kind of plan almost tells you when to breathe and when to go to the little linguist's room and will take you though all the steps from first sitting down with the speaker(s) who was agreed to work with you and until your data are safely stored on your labtop and backed up on your external HD; this kind of detailed and operational plan I like to call my Work Flow. Traditional linguistic fieldwork workflows are somewhat simple since they basically only involve the linguist asking questions to the speaker, receiving an answer which is then written down in the notebook before asking a new question etc. A modern workflow may also involve such tasks as these but would involve many other important tasks in addition, such as:

• informing the speaker of the purpose of the research 
• the inteded uses of the data
• the speaker's rights related to the data and their use
• obtaining the speaker's signature on an informed consent form 
• explaining how elicitation works
• explaining what kind of questions you will be asking and their purpose 
• turn on recording equipment
• checking recording equipment is working properly by listening with headphones
• properly position external microphone
• document the circumstances surrounding the recording, i.e. write meta-data
• transferring raw recordings to labtop
• enter elicited data into dictionary database and text editor for analysis

If you plan to record naturally occurring speech apart from elicited speech you further need to:

• make any necesary revisions to recordings for transcription
• link media files to annotation files in a transcription program
• transcribe, translate and interlinear glossing

The most time consuming and tedious of all of the above mentioned task is without any doubt transcribing and glossing. On the other hand these are also the most helpful and valuable kind of annotations you can make because they essentially make the data you collected available to other non-speakers, laymen and researchers alike.

So, how do you transcribe a stream of speech without speaking the language - the answer is of course "with the help from a speaker". Of course, if the speaker is willing to learn how to use a transcription program then this eases the process because he just needs to be equipped with a computer and a set of headphones. But in many cases this is not possible; there might not be enough time for him to learn how to transcribe, there might not be any orthography used for writing the language, there might not be any litterate speakers, the speaker you are working with might be an elderly person who might not be geared towards computers. So, in these cases the transcription setup needs to enable two people (you and the speaker) to listen to the recording at the same time in order to arrive at a common understanding of what is being said.

I have done some experimenting with different setups for this kind of work:
      
One involves two high quality headsets (such as the Sennheiser HD-280 pro or the Beyedynamic DT 770 pro) and a splitter that channels the sound into both headsets; soundwise this setup is ideal because it employs high quality closed headsets which is always the best way to transcribe. On the downside, I have found that some people get uncomfortable wearing headphones for extended periods of time, especially if they're not used to wearing headphones (many speakers of endangered languages have never even wore headphones before in their life). I have also found that wearing closed headphones makes communicating difficult - and you need to both hear what the speaker is saying to be able to type it in and to be able to make analyses while transcribing and glossing. 
     
The other setup simply involves a set of speakers allowing the both of you (and everybody else in the vecinity) to hear the recording. The obvious advantages of this setup lies in that you avoid the above mentioned problems. The serious downside is that you just can't get the same kind of accurate accustic transfer from sound file to your ear as when you use closed headphones - a milder downside is that good quality loudspeakers such as the Roland Cakewalk MA-15DCW weigh almost 7 kg and they are even compact versions capable of the same output as larger stationary studio equivalents.

It may be hard to believe that there is such a big difference between closed headphones and loudspeakers; recently a colleague objected to my insistence that one of the above two setups is strictly necesary in order to do proper transcription suggesting that "you might as well just use the built-in speakers of your labtop when you don't need to do any fine tuned phonetic transcription and analysis". To this I would reply "not true!" - you really need to set down and hear the difference in order to really appreciate it. And of course, for a large part transcribing with loudspeakers, or perhaps even low grade headphones is not a problem - it is when you get to that final maybe 10 % of instances when you, and even the native speaker, as hard as you try are unable to make out what is being said. I've had this experience transcribing English, and also Danish, my native language, let alone when I was transcribing Nawatl, Tol and Garifuna which to me are exotic languages that I mostly have non-speaker knowledge of.

For the current project I will be using the second setup, the one with the loudspeakers, but I will also have by Sennheiser 280s with me in the field in order to better resolve doubts and for when I'm listening on my own.