Speaker Identification in Q&A

One problem often faced by transcribers and editors is Speaker Identification, particularly in the case of corporate participants, while they answer questions posed by Analysts. To learn how to identify speakers correctly, we should first know about the different elements that come into play in a person’s voice or speech pattern.

Basically the acoustic pattern of a person depends on his anatomy, i.e., size and shape of mouth and throat, and learned behavioral patterns, i.e., voice pitch, speaking style, et cetera. Just like every person has a different fingerprint, every person has a different voice and speech pattern. No one has a completely neutralized and perfect speech pattern, and if somebody does then he or she is unique and would be easy to identify anyway.

Let’s learn about some of the elements that will help you identify speakers:

1)     Accent: Every person has a particular accent, which is a mixture of three elements: intonation, word connections, and pronunciation. To know and learn more about Accent please visit Understanding Accent webpage on this site.

2)     Density of Speech: Some people talk fast and say more words per minute, while other people talk slowly and hence less words per minute.

3)     Usage of Fillers: Speakers use different fillers according to their preference. E.g.: “you know,” “basically,” “like,” et cetera. Also see Speech Disfluencies.

4)     Sentence Structure: Some speaker tend to start every sentence with an “And;” some end every sentence with a “so.”

5)     Grammar: The grammatical patterns used by speaker also might differ from one person to another.

6)     Voice Quality: Some speakers are loud, while others are not.

7)     Type of Answer: Some speakers like giving long-winded answers, while others like to be short and to the point. This is a personal preference, but helps in speaker identification.

8)     Area of Expertise: A CFO will talk about financials, a COO will talk about the day-to-day operations of the company, and a CEO will talk about administration and overall company performance and operation.

Note: Speaker Identification should be done using not one, but all or few of these elements into consideration.

One of my former employers always used to quote “Never Assume Presume.” But in the case of speaker identification making a reasonable educated guess taking into consideration the elements explained above is the best one can do, unless of course the speaker identifies himself/herself before answering the question.

Now let’s learn the steps you have to take for speaker identification:

1)     First and foremost, you should list all the corporate participants and their designations that are present on the call.

E.g.:   Mr. X, CEO

          Mr. Y, CFO

          Mr. Z, COO

2)     Then try to create “Voice Prints” for each participant, i.e., audio snippets of each participant should be noted.

E.g.:   Mr. X, CEO          =       03:20

          Mr. Y, CFO          =       24:35

          Mr. Z, COO          =       32:55

If you use Express Scribe Transcription Playback Software for transcription, then you can use the inbuilt feature of audio bookmarks available in the software for creating Voice Prints.

3)     Listen to each “Voice Print” intently taking into consideration the above mentioned elements of speech. If you can attribute specific quality to each speaker.

E.g.:   Mr. X, CEO          =       03:20  = Nasal, Loud.

          Mr. Y, CFO          =       24:35  = “you know” smooth

          Mr. Z, COO          =       32:55  =  Heavily accented, slow.

4)     This might seem a bit time consuming, but considering the fact that you don’t have to go back to the “Voice Prints” time and again to match specific audio patterns, it is more convenient and less time consuming. Secondly, with practice the whole exercise shouldn’t take you more than a few minutes.

5)     If more participants are there on a call than the ones presenting in the call, and you come across a sentence like “Other members of the management team are present here.” Then you need to be extra careful while identifying speakers.