Nanosemantic: DialogOS (Dialog Operating System)

The main articles are:

• Speech Recognition (Technology, Market)
• Speech technology: On the path from recognition to understanding
• RPA (Robotic process automation)

DialogOS is a dialog platform, a tool for managing dialog robots (chat bots, voice and text robots). DialogOS allows you to create and train dialog robots used to process user requests in linked dialog mode.

2025

Finalizing System for Voice Control in Robotics

Nanosemantic has completed a universal software complex that allows you to control robots using natural speech without the requirements for high equipment power. At the heart of the ‒ DialogOS complex: the platform provides a voice interface, recognizes and synthesizes speech, integrates with external systems through the API. A device with a microphone and speaker is enough to embed. The company announced this on December 2, 2025.

𝓲

Фото: Nanosemantic

Neural networks allow robots to perceive speech as it sounds in a regular conversation. Where strict command compliance was previously required, the system now correctly processes free formulations, relying on meaning and intonation.

The complex includes the company's own solutions: DialogOS dialog platform for creating voice and text robots, NLab Speech ASR for accurate speech recognition and NLab Speech TTS for its synthesis. The current version of DialogOS is compatible with Gemini and the domestic generative model, while the architecture of the complex is designed in advance to connect any other LLM. Nanosemantic tested all major LLMs from leading developers, and depending on the specifics of the task, the most current model can be connected to the complex.

{{quote 'author=said Yegor Kirillov, business analyst, Nanosemantic Laboratory LLC. | The key task in the development of this complex was modularity and independence from the hardware platform. We implemented a client-server architecture where only an ultra-light client is launched on the target device ‒ be it a robot dog or an industrial manipulator ‒. All complex processing, including KWS, VAD and LLM integration via DialogOS, occurs on the server. This approach allows us to guarantee high performance and recognition quality regardless of the computing power of the robot itself, as well as centrally update and scale the "brain" of the entire system,}}

The complex has already been tested on a robot dog. She knows how to maintain dialogue, executes voice commands and responds to her name ‒ "Quantum." For this, a combination of KWS and VAD was configured, allowing the robot to track the call in real time. Quantum speaks with the synthesized voice of Leo from the company's library, although third-party options can also be connected if necessary.

Шаблон:Quote 'author=noted Sabina Spirina, General Director, Nanosemantic Laboratory LLC.

Service for accurate evaluation of virtual assistant responses

Nanosemantic has significantly expanded the functionality of the DialogOS platform: among the latest updates is a service for accurately assessing the responses of virtual assistants. The tool for marking individual replicas (steps) and the entire chat (session) will allow the platform user to point through the dialogue and make it more efficient.

DialogOS is a professional dialogue platform that combines a variety of tools and technologies necessary to create, monitor and support dialog robots of various levels of complexity. It is also used to build a communication architecture in robotic systems and digital avatars.

Now users of the AI system can assess the effectiveness and appropriateness of both the specific response of the chatbot in the question-response bundle and the dialogue session in full, putting special marks on them (positive, neutral, negative). Filters allow you to select sessions according to certain parameters and set a time interval for them, and after the markup is completed, the system generates a report with a metric of the quality of chat bot communication.

The service will be useful to business analysts, dialog scriptwriters and all DialogOS users who actively track recognition indicators and the level of satisfaction with chatbot responses. Based on the statistics received, they will be able to adequately assess how well the assistant is configured and copes with the task, as well as "catch" weaknesses and form a list of necessary improvements.

Also in development is another analyst and screenwriter tool ‒ Dialogue Map ‒ which will help visualize a set of dialogs as a built graph. It clearly demonstrates in which places of conversation there is a fixation and outflow of customers, and other states in which they fell into dialogue. When analyzing the graph, you can understand which nodes to pay attention to and which dialogs you need to view in more detail.

The marking functionality is based on the principle of qualitative analysis: random sessions for a certain period are included in the sample, forming a representative set of data. This approach allows you to set and test hypotheses, draw conclusions based on statistics and predict the further behavior of the assistant, ‒ said Grigory Shershukov, Product Director of Nanosemantics.

Session markup will be available to Nanosemantic customers on the DialogOS platform in beta as early as November 2025.