BSS Dialog Composer

The main articles are:

• Speech technology: On the path from recognition to understanding
• Speech Recognition (Technology, Market)
• Speech synthesis

Dialog Composer allows you to develop dialog applications for interaction through phone, chat, SMS, mobile devices or social networks. Supports open standards (VXML, SIP, MRCP, REST), real-time testing and debugging.

2025: Integration with NLU

BSS has released an update to its robot scripting tool. Now Dialog Composer (DC) automatically loads topics from NLU and simplifies the loading of audio materials. Improvements are aimed at speeding up work and increasing the efficiency of user interaction with the system. The company announced this on April 15, 2025.

A key change was the integration of Dialog Composer (DC) with NLU (BSS toolkit for training recognition models through a visual interface), allowing you to select intentions or a list of topics from the model directly during the process of working on scripts. Users will no longer have to switch between systems or remember the topics of cases that models are trained in NLU, which increases the speed of work, the accuracy of scripts and reduces the likelihood of errors. Also, when a new theme appears, it will automatically load into DC and become available when working in the tool.

Another addition is to refine the answers of large model languages ​ ​ (LLM) using code. This opens up opportunities to customize answers, including adjusting words, replacing sentences or searching for information on the customer's website.

There is a function of batch or mass simultaneous loading, several pre-recorded replicas of the announcer in the "Audio files" section, which significantly speeds up the implementation of scripts. This is especially true when working on long scripts with a lot of pre-recorded audio recordings.

Improved the Dialog Composer (DC) interface, making it in the same style as speech analytics (RA). The result is a single UX ecosystem, which provides more intuitive work and improves the user experience. Now you can customize the sequence of menu sections to suit your needs, hiding unnecessary items thanks to drop-down lists.

It is also worth noting the transition to use in databases OpenSearch return. MongoDB This DB is open source, which allows you to freely deploy, improve and expand. software OpenSearch supports distributed processing, allowing users to scale resources horizontally for additional capacity and performance.

{{quote "In this version of Dialog Composer, we focused on improving the functionality, increasing its convenience for users and the speed of tasks. These changes make Dialog Composer an even more powerful tool for automating business processes and improving the quality of customer service, "commented Alexander Krushinsky, Director of Voice Digital Technologies at BSS. }}

2024: RAG bot capability

The creation of the RAG bot is available in the updated version of the no-code platform from BSS. The developer announced this on August 22, 2024.

Dialog Composer (DC) is a tool with which you can independently, without connecting a vendor, create and configure scripts for bots. In version DC-2.11. The service design has been improved and simplified, and functions and generative artificial intelligence have been added to help create more personalized automated customer interaction schemes.

When finalizing the platform, the requests of its users were taken into account. The first thing that becomes noticeable is the new color scheme. Pleasant and harmonious soft shades of the blue palette appeared, helping to work more productively. And by replacing the inscriptions with menu icons, it has become more compact and readable.

In this version, you can create script elements (nodes) with an instant transition to editing mode, which saves time when creating a bot and makes working in a new interface more friendly. Fast text entry for bot responses has been implemented, the ability to work with pre-recorded promptas, search by the name of the desired audio recording, grouping of elements from the drop-down list have been improved. All this allows you to edit scripts in a few clicks and significantly speeds up the entire process of creating a bot.

Also, the customer can independently adjust the height of the bot's voice and the speed of synthesis, deciding what the assistant will pronounce faster or slower. Previously, for such an adjustment, it was necessary to attract engineers from the vendor.

And you can also set communication channels for the text, flexibly configuring responses robot in instant messengers and on customer sites. It became possible for the platform user to independently adjust the sensitivity of Voice Activity Detection (VAD) - a technology that allows you to ignore periods of silence or background noise, saving computing resources and speeding up the data processing process. Users can choose the sound level to recognize to exclude background noise recording.

Alternative texts can be configured for each scenario element. If the client first hits the node, then the bot responds with the first replica, additional options are used when hitting again.

In this version, we collected feedback from our users, and also took into account current technological trends. In my opinion, the update turned out to be a breakthrough in terms of the implementation of new opportunities. I especially want to note the possibility of creating RAG bots that generate responses using LLM models, "commented Alexander Krushinsky, director of the voice digital technologies department at BSS.

He stressed that BSS implements its own LLM models for use in the customer's loop, but can also connect third-party ones, for example, from Yandex or Sberbank. The client receives all possible settings for the correct operation of such a bot: industrial control, model selection, the level of compliance of the client's question and information in the guides.

The bot will definitely not use third-party information: it will answer only on the documents that you show him. We have prescribed in the code a ban on the use of data from other sources. Also, the bot will not give confidential information from the client to the third-party model. This is guaranteed by the fact that we either contact local LLMs without access to the Internet, or remove personal information from customer questions when working with cloud generative models, "said Krushinsky.

2021: Version v. 2.4 with the ability to optimize the architecture of the application of voice and text scenarios

BSS on October 14, 2021 announced the release of updated versions of the Voice Analytics, Dialog Composer and NLU Suite services. Starting with v. 2.4, Dialog Composer allows you to optimize the architecture of voice and text scripting applications using the Decision maker function. Read more here.