Breaking Barriers

In 2018, Dialpad acquired TalkIQ to combine its best-in-class automated speech recognition (ASR) and natural language processing (NLP) capabilities with our leading communications solutions. The technology from TalkIQ plays a huge part in Dialpad’s native Voice Intelligence (Vi) technology, which provides users with intelligent conversation insights like real-time coaching and sentiment analysis.

Over the past few years, our team has analyzed over a billion minutes of voice to further improve our ASR technology. As part of this process, we’ve benchmarked our incremental updates against key competitors in this area and found that we scored lower on transcription accuracy... until now.

How We Measure

To measure how our technology has improved, we send a test set of audio files to each competitor and compare the resulting transcripts with the ground truth. Then, we tally up the number of errors in each transcript to determine the accuracy percentage. While our test sets don't remain static, we’ve been able to determine that, for the first time, our accuracy percentage is now higher than all other major competitors on our customer's audio.

Specifically, Dialpad’s transcription technology is the top provider for both general and keyword accuracy for our customer's calls. General accuracy looks at the entire audio clip, while keywords focus on words our users care about such as company, people and product names.

We’ve found the difference in performance is statistically significant for five of six domains: Contact Center, Talk, Sell, Dialpad Meetings and Australia/New Zealand. We have achieved these gains by implementing experimental models and architectures, building product-specific models (ie. one for Talk, Sell & UC), and performing specialized processing for companies and users.

Why Does This Matter?

From live coaching and suggestions to sentiment analysis and note-taking, accuracy is key if you want useful takeaways from your conversations. We strive to make the most accurate real-time speech recognition, so that the insights provided by our Vi engine are helpful and time isn’t wasted trying to decipher poorly transcribed notes and action items. We want your team to be able to focus on the task at hand by trusting our real-time speech recognition and voice analytics technology to take care of the rest.

How Can My Team Leverage Vi?

Vi must be enabled at the Company and Office level, then any call will have the option to take advantage of its features. Additionally, you can configure Vi preferences for Individual Users, Main Line, Departments, Call Centers, and Coaching Groups. Learn how to enable Vi here. For more information on our privacy policy please visit our Security & Privacy page.