Voice Solution Building

NOIMAN Consulting

The major workload when building a voice solution comes from customizing a voice platform to the features of your product or serviced, and to the language of your users. Let us do that for you!

We build your voice solution on these platforms

Amazon Alexa / Lex

Google Actions

Samsung Bixby

Apple Siri

Nuance Vocon

Softbank Choreographe

Line Clova

Which platform is best for you? Our innovative approach allows easy scaling of a solution built on a specific platform to any other platforms and languages, so you can build on one platform now to get your foot into voice, and then port to a new platform later for a fraction of cost and time.

We also do the following:

  • Quick migration of an existing voice solution to a new platform (e.g. port an Alexa skill to Bixby)
  • Building your own voice platform from scratch. This option brings voice to products, services, hardware, software or languages which are not supported by the big voice platforms.
  • Any Natural Language Processing (NLP) tasks such as parsing, grammar building, machine translation or information retrieval.
  • Design a natural-language chatbot along your specifications and implement with a botkit like DialogueFlow or wit.ai.

Modules of Building

Building a powerful voice solution means completing several essential procedural steps and modules. We can take of all of these for you, or help you with individual parts:
  • Rapid Prototyping: A working proof-of-concept will show you early in the project, how voice can change the feel of your products and services.
  • Use Cases and Requirements: Define scenarios and user stories how voice will help people to use your products and services, and convert them into technical specifications.
  • System Architecture: Connect voice seamlessly into everything around: hardware, servers, databases, APIs, content, IoT, graphics and existing user interfaces.
  • Design of Language Model, Grammar and Dialog: This is the core task of any voice solution: What will people say in their own words to get something done by your products and services, and how do the machines understand this? On some platforms, this will also include: carefully defining a set of natural sample sentences, creating a pronunciation dictionary for unknown words like brand names, setting format standards or writing linguistic rules.
  • Localisation: Launch your voice solution in multiple languages, or add languages to your existing voice solution. Each language presents individual challenges, so mere “command translation” is not enough.
  • Data Collection: A corpus of linguistic data helps to get a robust range of how people say the same thing differently (and they will). We can draw on our own databases, or organize individual new data collections with native speakers.
  • Coding: Convert dialog flow and architecture efficiently into code as required by the platform, e.g. in node.js, C++ or Python.
  • Quality Assurance: Run a scalable suite of tests and tools to establish an objective KPI showing performance and accuracy of your voice solution.
  • Legal: Manage licensing, copyright and privacy ramifications of your voice solutions.

Can’t we build it ourselves?

A frequent yet risky assumption is that nowadays, voice has basically become a free commodity and that anybody can voice-enable their products or ideas more or less simply by using voice “off the shelves”: All that remains to do is to supply a few sample phrases and to modify code templates, and then the voice solution is ready to go live. This “quick and dirty” approach is risky: After a first bad impression, people tend not to come back to voice; some might even post funny videos about their short-lived experience with a bad voice solution.

The most popular Alexa skills like the Jeopardy quiz skill (produced by Sony) or the Chase Manhattan banking skill are developed by dedicated in-house voice teams of more than a dozen specialists each who build skills for a living. These companies understand that user experience and usability of voice rises and falls with how carefully a voice solution is customized for specific requirements, and that it needs to be well supported also its after release.

Careful customization of the building blocks that the big voice platforms provide is the key to using the voice commodity smartly – every use case is different because people will say different and new things to your product or service, and they might expect a different response from the device . The process of identifying and designing these individual sets of linguistic data and then coding them efficiently is the core work of customizing voice.


If you build a voice solution, let us make your work easier by guiding you with best and proven practices and by providing you with an objective framework to measure progress and performance. Or just leave some of the specialist work to us.

Switch The Language