How we built a transcription product by accident

by | Nov 7, 2017 | Product Launch

When talking to people about Yack.net, I have been asked the same question “If Yack.net is such a good idea, why have you only just got round to building it?”

While solutions to create recordings and transcriptions of your meetings already exist, these existing methods are expensive, time-consuming and may even disrupt the balance of power within your company. Yack.net offers any business the ability to record conference calls where the recording may prove useful, and at a price point that every business can afford.

The value of transcribed meetings has been known for years – but the methods are outdated:

  • Pay a stenographer to sit in on your meeting and take shorthand notes – After the meeting, a transcript is produced from the notes. It goes without saying that this is extremely expensive and can only be used in the highest value meetings such as board meetings, court cases or parliament.
  • Nominate one of the meeting attendees to take notes or produce minutes while the meeting is in progress – The trouble with this is that it interrupts the flow of the meeting while everyone stops and agrees what has just been said in order to get the minutes right.
  • One of the participants on a call will record the audio on a phone or dictaphone – They may then pass the recording through a transcription engine to generate transcribed text. The problem here is that trust may be undermined when only one party has access to a recording and transcript. Also, the recording and consequently the transcription will probably be recorded in a single audio stream, so any transcription is unlikely to attribute the text to the correct participant. If the recording is made without the knowledge of everyone involved in the meeting, there may be a legal issue if the other participants discover they were recorded without consent.

The key use case of Yack.net is to enable participants in more meetings to benefit from recording and transcription than ever before. And it’s far more cost effective than the methods used to date.

So back to the big question: If it’s such a good idea, why have we only just built Yack.net?

The answer lies in the fact that we didn’t set out to build it …

Connected Digital is a small software development company that specialises in developing business systems. We have a development team in Poland, management in UK and we tend to be working for three or four customers at a time. Having built a number of web portals with a common theme around connecting people, it became clear that we could make our own platform and demonstrate our competence as a software development team.

We decided to start with a simple business messaging portal, as this involved implementing product features that were common to a number of our customer systems. We started looking at business requirements to implement a messaging portal that is better than existing products such as Slack or Skype for Business.

Beyond the obvious requirements of being able to chat, call and share files, our first differentiator came from our history of working on a number of software development projects where we become embedded in our customers’ organisations. We regularly represent our customers within their communities and with existing team collaboration software; it can be difficult to be a member of multiple organisations within a single login. It’s time-consuming to keep logging into the same collaboration tools with specific logins for each team in order to keep track of all of your projects. So an early requirement for Yack.net was to develop the ability for a single user to be a member of multiple teams – where all the members of a particular chat room would see you as a member of the organisation that sponsors the chat room. This enables a Yack.net user to represent themselves as members of an appropriate organisation within each conversation in which they participate.

Once the chat rooms were working well with our organisation membership model, we added a search function (because all teamworking collaboration software needs to offer the ability to search for specific content).

We also started developing calls within the chat room. Initially, we experimented with peer-to-peer calls but then, while working on completing the search function, we thought “Why can’t my calls be searchable like everything else in the portal?” We quickly completed an initial prototype using the Chrome browser’s built in ability to perform live transcription of audio at the same time as making a peer-to-peer calls. We equally quickly dismissed this option because it was difficult to share the transcription of each user’s audio with the whole team and we found that reading the transcription of a call without access to a recording of the call proved of limited value due to inaccuracies in the automatic transcription and the fact that people tend not to speak in perfect, fully formed, sentences at all times during a meeting.

Having established a requirement to record the calls at the same time as producing a transcription, in late 2016 we identified Kurento Media Server as a technology that enabled us the host a call where we could produce the recording that we needed. Cloud-based transcription services were available as a beta service in Microsoft Azure and with a great deal of code to implement a number of our own ideas about how to make the idea work, we created a working prototype in early 2017.

Over the past nine months, we have refined and optimised the system. Kurento has been replaced with Twilio (which gives us a scalable cloud-based media server) and Microsoft Media Services are now integrated in the full release.

Having managed a team using working prototypes of Yack.net over the past nine months, I have found that I frequently use the transcriptions to search back over my calls and replay parts of a conversation where specific targets and actions have been agreed. This has enabled me to track both my own targets and those of my team.

So the simple answer to the question “Why Yack.net now?” is that the potential of the system became apparent as the pieces fell into place.

0 Comments

Submit a Comment

Your email address will not be published. Required fields are marked *

Share This