Transcription Shortcuts

You didn’t get a degree in Speech-Language Pathology without at least hearing about language sample analysis (LSA). For those of us who had to learn to analyze a language sample by hand, I feel your pain. In another post, I share why I believe the labor-intensive, prone-to-error, by-hand method of LSA is outdated, unrealistic, and destructive to the process. I digress. Today’s entertaining read is on the topic of transcription. And yes, I know you are groaning, rolling your eyes, and feeling a slight pang of distress. The word “transcription” conjures the same negativity as language sample “analysis by hand.” But it really isn’t so bad. In this post, I am sharing some transcription shortcuts using SALT software. I explain when you can take a shortcut in transcription and what those shortcuts affect in your analysis. Transcription can be complicated, but it does not have to be. By applying SALT’s standard codes and conventions, the software will produce data describing transcript length, intelligibility, syntax, semantics, verbal facility, pragmatics, and errors in production. Further outcomes can be obtained depending upon the sample context and if higher levels of coding are applied to an existing transcript, for example, narrative scoring scheme and subordination index coding. Generating all that data can make the transcription process more time consuming. But there are times when an assessment doesn’t call for outcomes on every facet of spoken language. Fortunately, there are shortcuts you can take to streamline the process and get only the information you need for your specific case. This is why understanding how transcription codes and conventions impact the analysis outcomes is useful. |: ; ( ) ~ [EW:] (( )) = + $ %* <> !.? \ Huh? These characters are how SALT knows what is an error, maze, pause, interruption, omission etc. Every convention/code drives the analysis outcomes. When you know what they do, you learn what you can eliminate and still get results you are interested in. Key to this is knowing the aims of your assessment: what linguistic features do I want to look at, and what features are not relevant? Let me explain. If you do the minimum - transcribe only the words of the target speaker, segment the utterances (C units), and mark elapsed time of the sample - you already get a lot:

• The legend/transcript to share with family, educators, etc.
• All words produced (total number of inflected words). Look over SALT’s list of all words spoken (in alphabetical order). This list can be quite revealing. Are there lots of words, but low vocabulary diversity? Were the conjunctions simple or higher order? Was pronoun use appropriate? Were there mental state words in the sample? Were there question words?
• Words per Minute. If you only type the target’s words, you get a gross measure of Words Per Minute (WPM).¹ WPM can be an indicator of excessive pausing, a slow or fast rate, and a strong predictor of language impairment.
• Sample length, or how much talking occurred. This can be very informative, particularly when you think about why the outcome occurred. For example, was the sample short because of low language, reticence, the topic, the context, or examiner influence?

This is the minimum. When determining what to add to your transcription, think in terms of building on this minimum in order to meet the aims of your analysis. Below is a chart explaining  SALT’s transcription conventions and their effect on language sample analysis outcomes. For example, look what happens when you add macrostructure coding (NSS) to the legend of target speaker words: you now have data that describes the overall coherence and organization of the narrative, crucial information regarding language abilities and how those abilities relate to academic performance. Or, type the examiner utterances for tremendous information about discourse (often useful for analysis of speakers on the spectrum), as well as a more precise WPM. Again, I remind you to be mindful of the aims of your analysis. For example, if you want a gross measure of intelligibility - either for an initial evaluation, or to assess progress - discover how marking unintelligible words and segments provide the data of interest. Ask yourself, “What elicitation context(s) will answer my diagnostic questions?”, “What measures will support my questions?” and, “What conventions/codes do I need to type to get those measures?”  

  _____ 1. Do note, however, the gross measure of WPM is calculated by taking the sum of ALL words spoken – by both the target speaker and the examiner – divided by the total time elapsed. In order for this measure to be useful, you as the examiner must either be mindful to limit your own productions or you’ll need to do more work on the transcription end.