Thin paper 0070

A short history of speech recognition

There has been more progress in speech recognition technology in the last 30 months than in the first 30 years.

Introduction

Computing power and artificial intelligence are largely behind the advances in this space. With massive amounts of speech data combined with faster processing, speech recognition has hit an inflection point where its capabilities are roughly on par with humans. The graph below is from Mary Meeker’s 2017 Internet Trends report. It plots Google’s word accuracy rate which recently broke the 95% threshold for human accuracy.

Mary Meeker's Internet Trends Report 2017 -- Slide 48

While there have been a ton of strides lately, voice recognition dates back to the early 1950s. Below are some of the key events that shaped this technology over the last 70 years.

1950s and 60s

The first speech recognition systems were focused on numbers, not words. In 1952, Bell Laboratories designed the “Audrey” system which could recognize a single voice speaking digits aloud. Ten years later, IBM introduced “Shoebox” which understood and responded to 16 words in English.

Across the globe other nations developed hardware that could recognize sound and speech. And by the end of the ‘60s, the technology could support words with four vowels and nine consonants.

1970s

Speech recognition made several meaningful advancements in this decade. This was mostly due to the US Department of Defense and DARPA. The Speech Understanding Research (SUR) program they ran was one of the largest of its kind in the history of speech recognition. Carnegie Mellon’s “Harpy’ speech system came from this program and was capable of understanding over 1,000 words which is about the same as a three-year-old’s vocabulary.

Also significant in the ‘70s was Bell Laboratories’ introduction of a system that could interpret multiple voices.

1980s

The ‘80s saw speech recognition vocabulary go from a few hundred words to several thousand words. One of the breakthroughs came from a statistical method known as the “Hidden Markov Model (HMM)”. Instead of just using words and looking for sound patterns, the HMM estimated the probability of the unknown sounds actually being words.

1990s

Speech recognition was propelled forward in the 90s in large part because of the personal computer. Faster processors made it possible for software like Dragon Dictate to become more widely used.

BellSouth introduced the voice portal (VAL) which was a dial-in interactive voice recognition system. This system gave birth to the myriad of phone tree systems that are still in existence today.

2000s

By the year 2001, speech recognition technology had achieved close to 80% accuracy. For most of the decade there weren’t a lot of advancements until Google arrived with the launch of Google Voice Search. Because it was an app, this put speech recognition into the hands of millions of people. It was also significant because the processing power could be offloaded to its data centers. Not only that, Google was collecting data from billions of searches which could help it predict what a person is actually saying. At the time Google’s English Voice Search System included 230 billion words from user searches.

2010s

In 2011 Apple launched Siri which was similar to Google’s Voice Search. The early part of this decade saw an explosion of other voice recognition apps. And with Amazon’s Alexa, Google Home we’ve seen consumers becoming more and more comfortable talking to machines.

Today, some of the largest tech companies are competing to herald the speech accuracy title. In 2016, IBM achieved a word error rate of 6.9 percent. In 2017 Microsoft usurped IBM with a 5.9 percent claim. Shortly after that IBM improved their rate to 5.5 percent. However, it is Google that is claiming the lowest rate at 4.9 percent.

The future

The technology to support voice applications is now both relatively inexpensive and powerful. With the advancements in artificial intelligence and the increasing amounts of speech data that can be easily mined, it is very possible that voice becomes the next dominant interface.

At Sonix, we can thank the many companies before us that have propelled speech recognition to where it is today. We automate transcription workflow and make it fast, easy, and affordable. We couldn’t do that without the amazing work that has been done before us.

Other Sonix Articles 📃

Tips on how to capture great audio

Hear from other Sonix users about how they record high quality audio

Why should you transcribe?

Five reasons you should be transcribing your audio and video files

History of speech recognition

How did we get to where we are today in speech recognition? Sonix explains

How to remove the metallic sound

The metallic, tin-like sound you hear in your audio is an unwelcome annoyance

Six best tips for transcriptionists

Helping transcriptionists work faster and be more accurate

AI, ML, and NLP

Artificial Intelligence, Machine Learning, and Natural Language Processing

Is voice the next major UI?

We think that it will change how we interact with technology

Word error rate

How do you judge accuracy in the realm of speech recognition?

A comparison of automation services

Independently reviewed and Sonix scores the highest among automated services

Start transcribing with Sonix 🚀

Sonix transcribes, timestamps, and organizes your audio and video files so they are easy to search, edit, and share. Start your free trial today—all features included, no credit card required.

Try Sonix for free

Your first 30 minutes of transcription are free