Synapse is an orchestration service provided by UneeQ that simplifies the deployment of digital humans. It connects a conversation engine to a digital human and prepares the text output from the engine as a spoken response. Synapse also includes features such as logging, language detection/translation, sentiment analysis, behavior automation, moderation, conversation state management, and more. It cannot be customized by customers, but it eliminates the need for manual preparation of conversation for spoken interaction. It can be set up for a customer in minutes.
Synapse is a professionally-managed orchestration service provided by UneeQ that dramatically reduces the effort involved in deploying a digital human to production.
Orchestration, also known as 'middleware', is the service which connects a conversation engine (i.e. NLP, NLG, LLM, etc.) to a digital human.
Synpase - a junction between two nerve cells, consisting of a minute gap across which impulses pass by diffusion of a neurotransmitter.
While digital humans do not have neurons, by way of analogy, the "thought" coming from the conversation engine needs to junction with the rendering of the digital human.
Most companies have explored or invested in a Natural Language Processing service (aka chatbot) to provide automation flows for business processes. This can be as simple as an FAQ on a website or as complicated as a creating a travel itinerary entirely via a chat interface. What they all have in common is they are essentially text-only. While some bots support images or emojis, and some bots have been integrated into phone systems, at their core, they are systems that receive text inputs and produce text outputs.
These text outputs are intended to be read, not spoken aloud.
Synapse exists to prepare the "thought" (aka the text output of a conversation engine) as a spoken response. This process involves a variety of steps entirely separate from the standard writing and editing processes. A simple example is a number written as 2023 needs to be spoken aloud as 'twenty twenty-three' casually or 'two thousand twenty three' more formally. When should we use informal speech? When should we use formal? How do you signal the system to express the written value correctly when spoken aloud? These are all the concerns of Synapse.
Syanpse additionally provides for behavior automation and coordination between multiple engines, making run-time decisions about how best to respond to a user's inquiry. For instance, a user speaking with a digital human trained by a financial institution may ask the question 'what's the weather today?' as a random thought within the stream of a normal conversation. A human bank employee might answer this question reasonably with information they know from a weather report or observed themselves that day. A chatbot would likely reply "I don't understand." This gap in ability to process any input naturally frustrates users and leads to low engagement with automation.
If we expect people to talk to machines... The machines better be able to talk like people.
In no particular order -
- Logging to Customer database or logging service
- Language Detection / Translation
- Coordination of multiple conversation services
- Sentiment Analysis
- Behavior Automation
- Moderation (Input and Output)
- Conversation 'guardrails' (preventing abuse)
- Maintaining conversation state
- Location awareness (requires frontend permissions)
- Fact-based domain (publicly-available, non-prioprietary knowledge)
Synapse is a managed service provided by UneeQ. It requires a separate license. Synapse is configured according to a customer's specifications by UneeQ employees.
No. Synapse, like SynAnim, is a global service focused on normalizing inputs so that they can be correctly spoken and animated by UneeQ digital humans. It is not the service where a customer would input anticipated inputs (intents) and expected outputs (fulfillments). UneeQ may provide baseline knowledge domains or personality configurations with all digital humans, but these services are not customizeable, they are either enabled or disabled by customer request.
For successul deployments, the amount of time we have observed customers invest in not only writing the conversation, but also in preparing it for spoken interaction is considerable. Synapse exists to eliminate the need to prepare conversation for spoken interaction as it automates the process entirely. Synapse also provides a mechanism for data retention to customers which did not previously exist in a single-step.
Synapse can be configured for a customer in minutes and made instantly active.