Skip to main content

Introducing: HERE Phonetic Training Data

Phonetic Data

“Human” language is full of irregularities and ambiguities that make it challenging to write software that accurately ascertains the meaning of text and voice data. We spend years and even our entire lives understanding the nuance of language while developers must use natural language processing (NLP) tasks to teach applications to understand these subtleties from the application’s inception. NLP as such is a branch of AI that combines rule-based modeling of language with machine learning to enable computers and devices to understand and generate speech and text. Applications dealing with speech that do one or more of the following have NLP driving them and some type of training data to start with:


HERE Technologies has developed a data solution that bridges the gap between accurate speech recognition and real-world application needs with HERE Phonetic Training Data. This is a data product designed to enhance Natural Language Processing (NLP) and speech recognition systems across various industries. 

If you are involved in voice-led application development, then the quest for high-accuracy speech recognition is something you may know very well. Developers have been asking for specific types of data that will help train algorithms, fuel their voice applications, and extend the reach of voice-enabled interactions into more markets. Phonetic Training Data is HERE’s response to that request. Here are few reasons to consider the HERE Phonetic Training Data in your projects:

1. Accuracy: Improving the accuracy of speech recognition systems starts with incorporating location-based phonetic data. This dataset helps ensure that systems both understand and process diverse pronunciations correctly, which is important for applications in regions with varied accents and dialects.

2. Versatility: HERE Phonetic Training Data can be used across various markets (see examples below). From gaming to banking, and from customer service to smart home applications, this dataset can be used to tailor speech recognition to your specific industry need.

3. High-Quality: HERE’s phonetic transcriptions are methodically curated to cover a broad range of content, ensuring that your speech recognition system is trained on the highest quality data available.

4. Customization & Flexibility: The dataset is also designed to be customizable and scalable, making it usable in your entire ecosystem of voice-enabled devices without requiring separate purchases for different applications. 

What is HERE’s Phonetic Training Data specifically?

Within the execution tree, phonetic data is really the starting point prior to speech recognition engine development. The phonetic training data can be used for training NLP engines to recognize and interpret speech accurately. The data consists of phonetic transcriptions that represent how words are pronounced, thereby enabling speech recognition systems to handle a wide variety of pronunciations and accents. It offers location-based phonetic transcriptions for improved speech recognition and covers the following geographic features:

  • Administrative areas
  • Points of interest (POIs)
  • Street names

Key Features and Capabilities

Some additional features that developers should know about when considering HERE’s Phonetic Training Data:

  1. Data Format: The data is provided in CSV format and contains attributes such as Plain Text, Phonetic Text, Preferred Pronunciation, and Phonetic Language. This format ensures ease of use and integration into existing systems.
  2. Standardization: Using an ASCII-based NT-Sampa format, the dataset ensures compatibility and ease of data transfer across different hardware and software environments.
  3. Ease of Integration: Designed with AI and ML in mind, the dataset integrates seamlessly with various platforms, making adoption straightforward and efficient.
  4. Privacy Compliance: Our privacy-by-design approach ensures that all data complies with GDPR standards, safeguarding user privacy while providing high-quality phonetic transcriptions.
  5. Data Freshness: Regular quarterly updates mean the data remains current, allowing your applications to evolve without costly upgrades.

What about coverage?

The HERE Phonetic Training data is currently available in these locations and languages:



Industry Use Cases

use cases

Within these use cases, some of the potential applications of the HERE Phonetic Training Data are: 

  • Voice Training
  • Data/Machine Learning/AI
  • Smart Assistants
  • Voice Bots
  • Voice Solutions

Some examples of real-world use cases for each: 

  • Gaming: Create dynamic dialogues and voice-controlled games that respond accurately to player commands.
  • Voice Platforms: Enhance online shopping experiences, event reminders, and smart home activations with speech recognition. 
  • CustomerService & Retail: Improve chatbot interactions and voice assistant accuracy in retail environments, providing a better customer experience.
  • Banking & Finance: Implement real-time biometric voice authentication and liveness detection to secure transactions and simplify banking operations.
  • Advertising & Media: Utilize dynamic, context-based advertising that responds accurately to voice commands and interactions.

Sample Data Overview

To give you an idea of what the HERE Phonetic Training Data has to offer, we provide sample data in North American - English. This dataset, along with all languages, includes the four attributes for training your systems:

  1. Plain Text: The original text data.
  2. Phonetic Text or string: The phonetic transcription of the text.
  3. Preferred Pronunciation: Indicates whether the phonetic transcription is preferred (Yes or No).
  4. Phonetic Language: The language of the phonetic transcription.

Example CSV using North American English 


How to get it?

HERE Phonetic Training Data available now. Pricing is subscription based to accommodate different needs and a sample data set will be available to help you evaluate its benefits. We invite you to explore the sample data to see how it can enhance your voice-enabled applications and unlock new possibilities for your applications!

Please reach out to us directly using our contact form to obtain the Phonetic Training Data!


As always, thanks for reading and connect with us on our Slack Workspace @heredev!

Aaron Falk

Aaron Falk

Principal Developer Evangelist

Have your say

Sign up for our newsletter

Why sign up:

  • Latest offers and discounts
  • Tailored content delivered weekly
  • Exclusive events
  • One click to unsubscribe