Audio Recording of Workshops and Seminars

As part of its E-Science Scoping Survey, the AHDS recently ran a series of Expert Seminars, designed to encourage subject specialists to debate and discuss different possibilities for the use of e-science within their particular fields.

As part of this process, the AHDS made audio recordings of the seminars with the aim of transcribing the recordings, and presented them to seminar chairs to facilitate their task of completing reports on each event.

This case study looks at some of the issues that occurred as the AHDS recorded and transcribed the material from these seminars. While its findings are based on roundtable seminars, some of them may also be of use to those doing other types of audio recording - interviews, field notes etc.

Equipment

The environment, the number of people speaking and the way discussion is structured (e.g. question and answer, lecture, discussion, debate) all determine how best recording equipment should be set up for audio recording.

In this instance, the E-Science seminars had on average 12 participants sitting around an table around four by three metres.

Initially, the AHDS used digital voice recorders to record the seminars. Three microphones were attached to the recorders and placed along the table. After two of the events it was clear this solution was not particularly suitable. On the digital file, the words were audible but they were muffled, and the volume altered according to the distance of the speaker from one of the three microphones. It was clear that a digital voice recorder would work well for a one to one conversation but for a group meeting was not appropriate. Noises from coffee cups, other crockery and other background noises also affected the quality of the final recording, although it was discovered that placing mouse mats under the microphones alleviated some of these noises.

For the rest of the series, the E-Science team then switched to speakerphones, which had a much more powerful recording range (six metres per speakerphone). Because the setting was reasonably compact and the speakers did not change positions during the day, this solution seemed the most appropriate one. Initial testing (ensuring that voices could be picked up and recorded from all seats round the table) proved this to be the case - all organisers are recommended to check a sample of recorded sound prior to beginning the actual event. In other instances, for example if the speakers are moving around larger distances, organisers may have to consider clipping small microphones to clothes and using radio links to record the sound.

In this instance, the speakerphones were connected together in series, with the first being connected to a laptop via a USB drive. This USB connection was sufficient to power the first speakerphone; the other two speakerphones needed to be plugged in at the mains. The E-Science team needed to be sure that they have the sufficiently long extension leads (preferably tied down to avoid anyone tripping up over the cables) so that the position of the speakerphone does not need to be compromised. Some speakerphones can work with batteries, but organisers will need do some investigation to see if they last for the length of the event in question - often batteries do not last more than a couple of hours.

Rather than someone switch the recorder on and off during the duration of the seminar, they were left to run for their full course (editing was done afterwards). Voice-activated recorders were not considered, as they can function poorly when there are softly-spoken voices.

With the aid of a well-regarded open-source software program called Audacity, the AHDS were able to record a high-quality version of the seminar, the recorded sound streaming directed from the speakerphones to the laptop. Audacity has settings that allow for each speakerphone to be recorded in a different file, or all together in one composite file. Various other settings have an effect on the sound file recorded (for example mono or stereo, sample rate)

Audacity also allows for basic editing of the digital files. This permitted, for instance, some noise removal (the sound of fans could be heard in the background of the initial files) and the removal of superfluous conversation at the beginning and end.

Other types of commercial and freely available software are available (A Google search for "audio recording software" reveals the large numbers). The level to which editing is required will again determined on the project's needs. If the recordings are simply done for transcription, then little will need to be done. But the more important the quality of sound, then the more editing will have to be done.

So if a lecture or conference paper by a distinguished guest is being recorded and then made publicly available, it makes sense to ensure the dissemination version has a good quality sound. Organisers should note that sound editing can be a time-consuming task requiring specialist skills. Likewise, editing may need to be done to split the sound files into different chunks (e.g. presentation and discussion) depending on predicted user need. Some types of audio recording (e.g. Apple's AAC format) allow for markers to be placed within audio recordings so that users can skip easily between sections. Such added features are likely to become more common in the future, but they demand extra editing time.

The file formats of the sound recordings may also have to be altered for dissemination. The original AHDS files were produced in the wav format. This created high-quality files, but they were of a large file size, too large to conveniently send electronically to the transcribers. The AHDS therefore made use of Audacity to convert the files to mp3. In general, if files are to be disseminated, whether over a public or private network, compressed versions of the sound files will need to be created, and if necessary split up into convenient sizes.

Transcription

Unless extensive editing of the audio is undertaken, a recording of a seminar or workshop is not a resource that is easy for the user to manipulate. (Although this may be different for lectures or presentations). Transcribed versions of recordings make it much easier to search for and analyse what was said.

Transcription is usually best dealt with by external agencies. Whilst it is possible to do on one's own, it is a slow, rather boring task, that can be greatly accelerated by the experience and specialist equipment of transcriptions agencies.

While there are plenty of agencies that can do the task of transcription, a surprisingly high number work with cassettes but not with digital files. Asking around may be the best way to find recommended (and non-recommended!) agencies. The ahds-discuss mailing list allows teams to ask about digital work they have out-sourced to other services. Local contacts and other mailing lists may also be able to supply details of services they have used within their area.

When communicating with such an agency, a number of variables will affect the final costs; the required delivery date, the quality and length of the recording and the type of content recorded.

For its E-Science seminars, the AHDS was fortunate in locating a service that did not alter its price according to the quality of the recording made. This was a distinct advantage given the quality of the first two seminars. Other services request a sample of the recordings to determine their quality and then proceed to finalise their rates.

The service used by the AHDS was also convenient in that it charged according to the length of recording rather than the time taken to do the transcription. If the latter, it can be difficult to get a good estimate of how much the service will take overall, and it may be more expensive than one had budgeted for. The former option allows for a fixed price.

The final costs paid by the AHDS depended on the delivery times requested. Some of the seminars required immediate transcription. A 24-hour service was provided which costed GPB 1.35 per minute. Other seminars were not so urgent, and having the recordings transcribed in one week cost around 80 pence per minute. For the rapid service, the files were broken into constituent parts and transcribed by multiple workers. This did not make much difference in this particular context, but other customers should be aware that splitting the task between multiple transcribers can cause problems if there needs to be a high level of continuity and consistency between all parts of the recording.

Rates may change if a team or department can offer a service regular transcription work. Larger transcription companies may also offer regular customers extra services, such as providing recording equipment.

A few other issues are worth bearing in mind. Those transcribing the event are unlikely to be familiar with the voices of the speakers, making it tricky to identify and label speakers in the transcribed documents of group sessions. Often, professional transcribers will circumvent this by, as happened for the AHDS, using the title 'female voice' or 'male voice'. If the commissioning team do want the transcribed version to have identified speakers, it is worth getting each speaker to clearly identify themselves at the beginning of the recorded session.

While some agencies may charge more when the content of the material to be transcribed increases in complexity, it is worth shopping around. Expert transcription services should be experienced in dealing with such material, and should be able to offer their services without extra cost.

Good transcription services will also ask for the level of precision required. On relistening to a recorded session, one will be surprised by the amount of dead time, unfinished sentences, pauses and coughs that are part of a recorded group session. Judging that such elements were not important to the final recording, the AHDS requested that such elements be omitted from the transcriptions giving a shorter, more readable transcription.

In other circumstances, however, it may be completely appropriate to have a high level of precision in the transcribed document, containing all the apparent gaps and dead ends in the conversation - this depends entirely on how the project intends to exploit the recorded material.

Consent

Before recording any event, some kind of consent is required from the participants. In this particular case, verbal consent was considered enough. At the beginning of the session, each participant was asked if they agreed to being recorded and their permission was recorded on the digital file.

Such an informal process was fine for this type of recording. The recording and transcriptions were for private use only; sensitive personal information was not included; the material was not going to be released publicly. But in cases where these are relevant issues, organisers should ensure that they gain written consent before beginning recording. How a consent form should be structured depends on the precise needs of the work being undertaken but at the very least it should indicate how the the recorded information is being used, stored and being made available to others.

Further discussion about consent and examples of consent forms (external link), used in various different recording contexts, are available from the Economic and Social Data Service.