Graphlogic.ai documentation

Additional information

Functional and non-functional capabilities of the Platform

Functional capabilities of the Platform

Clustering of historical data to identify frequent topics and enable automatic learning.

Multi-language recognition.

Recognition of natural language input produced by a person in an electronic chat - using closed and open grammar to understand the meaning and comprehend phrases expressed in free form by the interlocutors.

Intent detection (recognition of the subject matter of inquiries, determination of the main topic, and focusing on it, assigning the main topic to the dialogue).

Recognition of multiple intents within a single statement.

Handling of specific text input features (spelling errors and typos, abbreviations and numeral expressions, free word order).

Initiative: clarification of details regarding the problem or situation, assistance in formulating the problem, fulfilling requests.

Sending messages on a timeout.

NLU Supervised ML. Machine learning based on annotated examples (instead of describing rules in scripts and regular expressions).

Entity recognition. Identification of variables from the inquiry, such as dates, card numbers, etc.

Dialog scenario configuration through a visual interface, without the need for programming: adding new dialogue branches with branching based on parameters extracted from the interlocutors' responses or obtained from external systems.

Making changes to responses.

Adding new data sources.

Creating new dialogue topics.

Creating libraries of keywords and stop words.

Configuring new interaction channels with external systems.

Built-in visual tools for testing recognition accuracy.

Ability to integrate with cloud-based ASR/TTS.

Ability to use different recognition models at different stages of the dialogue.

Support for scenario versioning, ability to roll back a scenario to an earlier version.

Ability to import/export dialogue scenarios for transfer between isolated instances of the system and for creating backup copies of scenarios.

Mechanisms for ensuring scenario correctness (checks for infinite loops, correct parameter naming, etc.).

Ability to depersonalize dialogues using a separate microservice.

Ability for the interlocutor to interrupt the bot's response. Should be determined by the scenario configuration, separately for each response.

Dialog configuration tools do not impose restrictions on the number of dialogue steps or branches.

Ability to use regular expressions instead of NLU (for rapid scenario testing).

Ability to generate statistical reports displaying all dialogues between the client and the bot in terms of time intervals and interaction channels.

Ability to provide customized analytics for each individual bot as an additional service upon request.

Export of statistical data to Excel, CSV, or via API.

Ability to integrate with backend systems without programming.

Ability to integrate with backend systems directly from the scenario.

Determination of the tone of the inquiry (sentiment analysis) to transfer complex cases to an operator.

Determination of the stylistic component of the conversation, identification of conflict situations, and logic for resolving them.

Ability to integrate with external systems (sending requests, receiving responses, generating responses with variables obtained within the response).

Informing and routing the interlocutor when no response is received from external systems.

Ability to integrate with a knowledge base with answers via API through the built-in external request interface.

Ability to manage multiple projects (for different user segments, e.g., individuals, legal entities, internal support).

Non-functional capabilities of the Platform

Ability to deploy on customer servers.

Ability to deploy in a virtualized environment.

Horizontal scalability to ensure linear performance growth.

Ability to deploy in a high availability configuration.

Support for modern scaling and monitoring tools: Kubernetes, Grafana.

Built-in mechanisms for monitoring the operation of services and solution modules.

Ability to train/annotate/design dialogues by the vendor.

Access segregation by departments (each administrative department has its own settings and can only modify its own objects).

Ready integration with chat platforms (Enda, Jivo, Webim).

Technologies

Platform technology stack:

JavaScript/TypeScript, ES2015, React.JS, JSS, Styled components, GraphQL, Blueprint.js (frontend)

Python, asyncio, aiohttp, GraphQL, PostgreSQL, Redis, RabbitMQ, Docker, Kubernetes, Nginx, Prometheus, Gitlab (VCS, CI/CD, Registry), Pytest, Selenium (backend)

Integration with an audit/logging system (Kibana).

Integration with system monitoring tools.

Maximum number of concurrent sessions - 500+ messages per second on GPU.

Maximum response time - 3 seconds excluding external request processing time.

Ability to scale horizontally and vertically.

Functional capabilities of the Platform

Non-functional capabilities of the Platform

Technologies

Want to print your doc?
This is not the way.

Try clicking the ⋯ next to your doc name or using a keyboard shortcut (

CtrlP

) instead.