The Challenges of Mixed Audio Transcription

In today’s article, we are going to focus on the challenges of mixed audio transcription.

An audio which contains two more than two languages is called as a mixed audio. Here, we will be focusing on audios having two languages. For example: transcription of an audio which contains English as well as Japanese languages.

There are three types of mixed audios. They are as follows:
1. Consecutive interpretation.
2. Simultaneous interpretation.
3. Random mix of both languages.

Consecutive Interpretation: In this case, the two different languages (English and Japanese) are spoken one after the other. Here, the same audio gets allocated to both language transcribers. It is important for both the transcribers to listen to the entire audio, mainly because in consecutive interpretation, the rate of language change is very frequent. This process takes up a lot of time. Next, compilation of the two different final transcripts is required which adds up to the time taken. Frequent speaker interruptions also make it difficult to compile the transcript.

Simultaneous Interpretation: This signifies that two different languages are spoken at the same time. In this case too, the same audio gets allocated to both language transcribers. Compilation is not required as both the languages are heard throughout the audio file and can be transcribed in separate documents. In most instances, both the languages are heard in different channels (for example: Japanese in the left channel and English in the right channel). In cases where just one channel is available, it is very difficult for both language transcribers to work on the audio as it is not possible to clearly hear only one language at a time. Since both languages are heard throughout the audio, the transcription process is challenging and increases the time and effort utilized in the process much more than a normal single-language audio. Owing to this, clients are charged a premium rate, which they are made aware of before taking up the assignment.

Random mix of both languages: This means that there is a random mix of more than one language in the audio file; there is no pattern in terms of the frequency of language change. One needs to listen to the entire audio carefully to figure out the breakup of both languages. After that, the job can be allocated to both language transcribers. In this case, compilation of the files can be done by someone who is familiar with both the languages.

In all the above scenarios, it is clear that working on a mixed audio transcript consumes almost twice the time than that of a normal English transcript (single language audio). Allocation of mixed audio transcription jobs also takes significantly more time than the standard transcription job, right from listening to the audios to compilation of the final transcripts, as well as passing on specific instructions and references (if any) to both language transcribers.

Why are ‘Batch Deliveries’ not feasible/advisable in Transcription Industry

Definition: Whenever the client requires the delivery of transcripts in batches as opposed to a single delivery of the entire job, it is termed as Batch Delivery.
Batch Deliveries are not our standard service and we do not encourage them, the reasons for which will be clear as you read through this blog.

Batch Deliveries usually are favoured by clients in rare cases where;
1. Clients need urgent delivery of the entire job and it is not possible for us to process that job in a day. Thus, sometimes clients request us to split the job up in batches so that each batch will have a separate delivery schedule.

2. In some cases where clients have a very large job, for example, an entire Conference/Symposium which spans over an entire week, the clients upload the audios on a daily basis and hence the delivery is then scheduled in batches as per the turnaround time (TAT) for that particular file.

Disadvantage of Batch Deliveries:
1. Serious Quality Implications: With regards to batch delivery, there are multiple audio files with urgent TAT. These files are usually processed as single entities, not an entire job, and can get allotted to different Editors. This raises the issue of quality/consistency in the overall output of the job. If it is a single delivery, then we always try and give the entire job to a single Editor (owner) so he has a good understanding of the subject matter of the job, the speakers in the files with regard to identification, etc., and hence the final transcripts are accurate and of the best quality.

2. Consistency Issues: In batch deliveries, there are multiple files and multiple Editors work on it at the same time which usually results in issues of inconsistency in formatting/speaker identification, etc. For example, there may be instances where the Main Speaker has identified himself/herself at the start of the symposium/interview in the first audio file. As it is a batch delivery, the Editor doing the second or subsequent files might not be the same person who worked on the first file. Hence, the second Editor might miss out on that information given in the first audio. This will result in incorrect or no speaker identification in the subsequent files which leads to major consistency issues in the job.

3. Delayed Deliveries: Sometimes, as there are three to four batches of the same job that have to be delivered over a span of 3–4 days, there may be a high chance that some batches may get delayed. This is because it is very challenging for the Process Manager to keep a track of these batches as compared to a single delivery for the entire job.

Thus, we always encourage our clients to avoid Batch Delivery Schedules and to keep a single delivery schedule for the entire job so that we can assure our clients an unclear-free, optimal quality output with timely deliveries.