Why are stenographers still employed and present within many governmental chambers?

Question

Yesterday a news story broke about a stenographer on the US Senate floor fainting which prompted a brief recess while she was being medically attended to.

Hearing this, I started to wonder why stenographers are still being employed for this position, working from the chamber floor, when we live in a day and age where technology could do a better job than having someone listening and hand-keying in what they are hearing. Furthermore, there are already numerous recording devices in place throughout the chamber floor of the Senate, House of Representatives and probably most other state and international governmental chambers.

Specifically in the case of the US Senate, there are more than a few microphones and it also has multiple video feeds recording what's taking place. Modern technology can easily convert spoken audio to text, and even though the process isn't perfect it would probably be more accurate if it was automatically converted to text and then reviewed by a stenographer who reviews the recordings of each session against the computer-generated transcription.

I understand the importance of keeping thorough records and minutes of governmental meetings but in a place like Capital Hill, it seems to me that a better, more accurate approach to keeping such a records would to implement effective technology.

Is there a good reason why stenographers are still employed to hand-generate meeting transcripts?

UPDATE

I threw this question out with the very basic assumption that human-generated transcripts are not necessary when technology is already recording all of this material and auto-dictation software could certainly transcribe what's being said with moderate accuracy.

After reading the comments below, I further recalled that stenographers actually DO NOT the record words they hear, in the typical sense. Instead, a stenographer is trained to phonetically type what they hear so that they can type at speeds of over 200 words per minute!

I think this makes my question all the more relevant. It is true that speech recognition software is still imperfect but if the system is designed to dissect sounds, rather than words, the computer could quite possibly do a much better job at transcription that a human! This is because simple sounds, even coming at fast speeds would be easier to recognize than entire words.

Answer 1

I wasn't able to find an answer online, so I called the Office of the Clerk of the House to ask why they use stenographers instead of electronic recording devices. They gave me a list of reasons:

1. To preserve the decorum of the House.
2. House members like having a stenographer they can talk to about the record (to have something re-read, or to have their assistance locating materials in the record).
3. The use of electronic devices is limited in Congress. This is due both to security concerns and infrastructure.
4. With stenographers they have instant access to the record. With recording devices this may not be the case.

Some of these reasons seem better than others, but there they are "from the horse's mouth".

Answer 2

The answer is simple.

1. Because the machine cannot guess who is talking. The stenographer sees who talks and writes down the name. (this should be enough)
2. Because, lots of times, the speakers speak without a mike (especially during a disagreement) and the machine doesn't record anything. The stenographer is there and listens and writes down.
3. Because the stenographer has to give the overall picture of what is going on in the parliament (voices, objection, applause, etc.). The machine cannot do that.
4. Because the minutes have some standard format in some cases. For example for the voting, you don't write what is being said, but you use a specific form to make clear what every party votes.
5. Because the speaker doesn't always speak perfectly and sometimes makes serious mistakes. Someone has to check the grammar mistakes or other mistakes in order to protect the speaker etc

But I agree that there is no reason for stenographers to keep the minutes by shorthand since there are recorders. Shorthand should be used only for interventions without a microphone.

Answer 3

If I understand correctly, you are asking why isn't voice recognition software used instead of manually writing what one hears.

Although voice recognition is very advanced nowadays (and available for various types of devices) there are still some technical problems related to correctly identifying all the words.

This article tells us about some of these issues:

• Difficulty identifying the correct word in the context of the sentence in which it's being used when there is another word that sounds similar (e.g., whether the user means "wear" versus "where")
• Environmental variables such as background noise levels, machinery and other noises present in the workplace
• Speaker variables including stress, emotion, speech quality and health
• Differences such as accents and dialects

Of course, some of these problems can be tackled by using artificial intelligence / machine learning techniques (e.g. properly identification of word based on context), but some doubt still exists.

Another approach is to configure a very low tolerance (do not make assumption when the word is not very clear) and allow the computer to interrupt the speech and ask to repeat a phrase. Although this might provide very high confidence in speech recognition, it can be confusing for some persons.

This article also mentions some difficulty in properly recognizing fast speech, when words tend to be united (continuous speech as opposed to discrete speech):

Continuous speech recognition operates on speech in which words are connected together, i.e., not separated by pauses. It recognizes speech by analyzing words and the relationships between words (context). This is in contrast to discrete speech recognition where each word must be identified as an entity, i.e., as an individually identifiable unit without consideration for relationships between words. Because the latter requires that the speaker pause between each word, the words-per-minute output (approximately 150 per minute) is too low for use in many environments.

There can also be a juridical problem if the transcript is wrong in a way that clearly distorts the message. You cannot put the blame on the machine.

A possible approach is to have voice recognition operate with very few interruptions and have a human being review the text afterwards. Although a human being is able to review the text, there is a lack of feedback (cannot ask for clarification in a real-time fashion).

Conclusion: Possible reasons for not using voice recognition in official meetings are: technical limitations and/or juridical issues.