On-demand webinar: navigating the challenges to AI success
Resources
Blog

How a Tier 1 Automotive Software Provider Creates Smarter, More Natural In-Car Infotainment Systems

Published on
May 21, 2019
Author
Authors
Share

The Company

A leading provider of vehicle electronics software approached Appen to collect audio and linguistic data to help develop automatic speech recognition (ASR) capabilities for its in-car infotainment system.

The Situation

For an in-car infotainment system — or any ASR system — to recognize and correctly process voice commands, it must be trained on speech data that accounts for a broad range of inputs and all possible variation in how people speak. There are countless different verbal commands a driver might use to adjust the climate control, radio, navigation, phone, and other settings in an automobile. Training these systems to understand multiple dialects and various speaker categories poses an even bigger challenge, requiring many thousands of utterances in each of the targeted languages.

The Solution

Appen provides services to collect natural language data and text data, covering all the scenarios and variation that the system might encounter in the real world. Working with in-market, on-demand crowds of native speakers, we are able to rapidly expand ASR capabilities in new locations and languages, for any given scenario. And because the company has strict standards for audio recording quality, Appen replicates the same advanced recording procedures across different locations and studios, and supervises them to comply with quality standards for a range of languages used in the automotive industry.Services included:

  • Spontaneous, unscripted speech data collection in which native speakers are given a set of scenarios (i.e., How do you ask for temp lowered? Put your favorite music on? Change the radio station?), and must generate various responses
  • Text data collection using similar scenarios as for speech, but aimed to obtain larger volumes of data and a broader variety of speakers
  • Scripted speech data collection for short, fixed utterances
  • Test driving simulation to mimic the cognitive load of driving, so speakers come up with more natural, real-world responses
  • Country-specific studio data collection with specialist equipment to ensure that different studios are calibrated for precision and compliance to strict audio standards

The Outcome

Working with Appen for more than six years, the company has created a smarter, more connected and more natural in-car experience — with systems that are able to recognize natural spontaneous responses. With our data collection and annotation services, the company has rapidly expanded the system in over 20 new languages. And because our linguists have deep expertise in both creating and localising scenarios that mimic real-world driving conditions, the Tier 1 provider knows that it is receiving the high-quality speech and language data it needs to train its ASR systems.

Benefits

  • Spontaneous speech data that fits users’ natural behaviour
  • Rapid deployment in new languages & locations
  • Strict audio quality compliance across a range of over 40 languages

Related posts

Insights from the International Conference on Acoustics, Speech, and Signal Processing

Appen recently sponsored the IEEE International Conference on Acoustics, Speech, and Signal Processing (iCASSP) in Brighton. Our VP of Business Development in Europe, Dorota
Read more

Appen Becomes Leading Language Service Provider; Maintains Leading Position in APAC

Appen is excited to announce our official ranking as one of the largest language service providers (LSPs) in the global translation and interpreting industry. Issued May 2019
Read more

What is Human-in-the-Loop Machine Learning?

Human-in-the-loop (HITL) is a branch of artificial intelligence that leverages both human and machine intelligence to create machine learning models. In a traditional
Read more

Deciphering AI from Human Generated Text: The Behavioral Approach

One of the most important elements of building a well-functioning AI model is consistent human feedback. When generative AI models are trained by human annotators, they serve
Read more