We recently wrapped up developing our own Finnish speech to text ASR model and now it’s ready! Finnish speech-to-text (ASR) is designed to help large companies detect business critical patterns from customer phone calls such as sales leads, efficiency problems, unhappy and leaving customers and customer service quality.
Analysing customer calls helps sales managers find new sales leads (upsell opportunities and win-back leads) from existing conversations in near real-time while also providing customer service managers with a tool to help improve customer service quality.
Transcribing customer phone calls into text is the first important step towards finding business critical customer insight. Do you know why podcasts and phone-calls are different? True, there is a spontaneous speech in phone calls. Meaning that when humans speak over phone, they often pause mid-sentence to reformulate their thoughts, , their ideas fly from one to another, speaker can stop and restart a badly-worded sentence. We wrote all the differences also here.
After transcription, all calls become searchable in Feelingstream. Our AI-based models automatically detect each call’s sentiment, topic etc. This step spawns actionability which could mean contacting clients, improving services, taking advantage of sales opportunities, etc.
There are a couple of reasons why we developed our own Finnish speech-to-text model instead of using cloud-based services:
The best accuracy in the Finnish speech-to-text model
- Higher accuracy. Our speech-to-text (or ASR – Automatic Speech Recognition) model is more accurate because it is focused on customer conversations in Finnish in specific sectors. This was kept in mind while collecting training data for the model. Also adding different dialects to a training data, mean the higher accuracy in transcription.
- Adaptability. Our speech-to-text model is easily adaptable to customer-specific words or expressions. For example, if the model doesn’t recognise specific product names, the model could easily be fine-tuned to increase accuracy.
- Punctuation and capitalisation. To make it easier for business users to read the outcome of transcription, we do sentences. This means we have trained punctuation and capitalisation models for this process. Moreover, it helps data-scientist to fully analyse textual format conversations and prepare for sentiment, topic, etc models.
- On-premises. Our speech-to-text model can be used on-premise without any external connections. This means that you have full control over your data security.
If you are interested in detecting business critical patterns in customer conversations in Finnish, get in touch to see a live virtual demo.