What’s New in Omilia Cloud Platform®

Posted: Dec. 14, 2021

Always keeping up with our clients’ needs and demands, we’ve worked hard to create a Conversational AI platform that does the work for you, makes the hard stuff easy, and steps in right at the time you think you have an unsolvable task, an alphanumeric ID that you can’t capture or a continuously failing authentication process.

We’re excited to announce that the Omilia Cloud Platform® has been upgraded and polished to be your best bud on Conversational AI and automation projects.

Here’s what’s new:

Console

You now have an all-in-one, uniform UI that hosts all the OCP services and gives you a great navigation experience and quick and seamless access to all the OCP services and more!

No more stand-alone UIs for each of the services or switching tabs and pages to land in the right service.

OCP Console offers a unified UI for all the OCP services, with the same look & feel and better UX for the users. Users don’t need to visit each service’s stand-alone UI anymore.

In addition to this, the technology behind Console (micro-frontends) allows teams with maximum control of their service UI to have independent development cycles and deploy their work to OCP Console during runtime. The purpose of the Console-UI component is to provide the main layout of Console (landing page, one-stop shop for all OCP Service) by utilizing server-side rendering techniques. In addition, console UI provides the routing engine and the placeholder for OCP Services which provide a UI (frontend) part as micro-frontend. Console UI identifies which micro-frontends to fetch and then render and serve them through a unified user experience.

Console-UI also provides some common information across all micro-frontends, such as authentication data. Finally, since Console-UI is a wrapper application for other (possibly external) UIs, it follows a level of abstraction to be as much as possible decoupled with the frontends it serves

OCP Console’s key features include:

  • Uniform UI.
  • Accessibility through keycloak.
  • OCP Service Integration:
    • miniApps
    • Billing
    • NLU
    • Orchestrator
  • Support Center.
  • Learning

Bot Orchestration

We’ve created a single manager to help overcome the issues that emerge in a multi-model AI world; orchestrator. Orchestrator oversees the conversations and assigns tasks to the highly mission-focused bots, OCP miniApps®.

Orchestrator allows users to create and manage complete dialog applications through a simple and intuitive user interface. It’s a powerful tool that leverages the simplicity of miniApps and enables the creation of powerful dialogs without the need for training or extensive domain knowledge.

Orchestrator features:

  • OCP miniApps integration: Users can directly manage any of the miniApps that are available to them inside Orchestrator UI. They can also navigate to the miniApps UI and manage their already created miniApps, while their changes are being reflected in real time inside the Orchestrator UI.
  • Reusable Flows: Flows allow users to create self-contained and reusable dialog steps. This design pattern enables the creation of simpler and more testable dialogs and allows collaboration and parallel development between members of the same team.
  • Conditional Dialog Logic: Different paths in a dialog App’s logic can be created by using condition blocks.
  • Transfers: From the context of each App, users can create Transfer blocks, where they can specify several parameters like the transfer target and attached data. These blocks can then be used as parts of the App’s Flows.
  • Error Handling: Specific Flows can also be created and used to handle specific cases like, for example, by assigning them to specific dialog errors.This way, no matter where an error occurs, the specified Flows will be called to handle it.
  • Dialog Deployment: At any point during development, the App or Flow currently under development can be deployed with the click of a button and be ready to accept calls or chat sessions.
  • Chat Simulator: An embedded chat window allows testing the Apps in real time without navigating away.

NLU

Conversational Natural Language service offers state of the art Natural Language Understanding for Internal (Omilia) and External customers, fusing different Technologies to achieve Enterprise-Level Performance. Through its powerful API, the user can get Semantics like Entities, Intents and Contextual information, using input utterances originating from different channels (Voice, Chat, Mobile App, etc).

The Service offers Natural Language understanding to internal or other 3rd party customers outside of Omilia, providing Intent, Entity recognition by taking into consideration Contextual Information that might be present. Last but not least, the service also offers Automatic Correction and/or Normalisation of textual input.

Features:

  • pre-trained Ontologies per Industry
  • Ability to define custom intents by supplying phrases as training data
  • Ability to define custom Intents by leveraging Semantic Relationships between Entity Types (at the semantic level, not literal word level)
  • Ability to define new Entity Types and Entity Instances
  • Ability to define Context Rules, which operate on Intent and Entities to Infer higher-meaning (a simple example is "if you see a Number Entity (e.g. "100") next to a Currency Entity (e.g. "dollars") then mark that span as a structured Amount Entity Type (e.g. Amount.value=100, Amount.Currency=USD) )
  • Intents' grouping
  • Intent Dismambiguation: mark Intents as ambiguous, and define disambiguation strategies
  • Multiple Intent classification within the same utterance: "I would like to pay my utility bill, but first can you tell me my balance?"
  • Multiple Intent classification within the same utterance: "I would like to pay my utility bill, but first can you tell me my balance?"
  • Anaphora Resolution: "User: I would like to order a hamburger and a diet coke", "Bot: OK, anything else?", "User: actually can we make it a cheeseburger?". The bot will resolve the anaphora "it" to the previous order item "hamburger" and make an order modification rather than adding a new cheeseburger ontop.
  • seamless Fusion of ML and algorithmic-ontology models

OCP miniApps

OCP miniApps are task-specific bots that possess both intelligence and expertise around a specific type of information. This way you can easily pick and choose the best components to include in your customer service automation project.

OCPminiApps are zero-coding, zero-maintenance, instantly deployable and configurable natural language dialog components that handle a single task, such as soliciting a US address or credit card number, or even negotiating an appointment.

OCP miniApps are powered by the Omilia Cloud Platform (OCP) and bring Omilia Conversational AI experience to any call center platform provider. Pre-trained and pre-tuned by Omilia leveraging vast amounts of data, miniApps can accurately handle exceptions and real-world conditions, complete with data validation, disambiguation and error recovery strategies. OCP miniApps® can be quickly tailored to each application’s requirements, significantly limiting a substantial part of delivery time.

Types of OCP miniApps include:

Alphanumeric

  • Flexible options for minimum and maximum length
  • Ability to define regular expressions (RegEx) to increase accuracy in the speech recognition
  • Ability to compare against a predefined list of acceptable utterances (increased accuracy)
  • Pre-built ZIP Code grammar for US and Canadian ZIP codes
  • Note: alphanumeric accuracy is increased by the use of, regular expressions that describe “known good” formats (e.g., “all account numbers are in the format of abc-##), pre-defined lists of accepted utterances, and limiting string capture length.

Date

  • Takes dates from callers in a flexible fashion
  • Returns those dates in a desired format (e.g., mm-dd-yyyy, yyyy/dd/mm, etc…)
  • Options to return Day of Week, Day of Month, Month, and/or Year
  • Determine what date ranges are acceptable to return, including “future dates”
  • Ability to compare against a predefined list of acceptable utterances

Alphabetic (only)

  • Flexible options for minimum and maximum length
  • Ability to define regular expressions (RegEx) to increase accuracy selection via n-best list
  • Ability to compare against a predefined list of acceptable utterances (increased accuracy)
  • Note: alphabetic accuracy is increased over alphanumeric accuracy by elimination of numeric entities from n-best list

Numeric (only)

  • Flexible options for minimum and maximum length
  • Ability to define regular expressions (RegEx) to increase accuracy selection via n-best list
  • Ability to compare against a predefined list of acceptable utterances (increased accuracy)
  • Pre-built Credit Card Number grammar
  • Note: numeric accuracy is increased over alphanumeric accuracy by elimination of alphabetic entities from n-best list

Intent OCP miniApps

The intent miniApp provides a building block to handle intents. Mainly, it can understand an intent and map it to a prompt. There are several different domains available:

  • Banking
  • Car Retail
  • COVID
  • Insurance
  • Telco
  • Universal (generic intent domain)

In each of the miniApps you can always group the different existing intents together or add custom Intents on top of that using machine learning.

Corpus Collection

You want to create your custom model for your domain for your organization? You need data!

Set Up a corpus collection application in minutes, collect caller utterances by asking a simple “How may I help you question”. At this point the system has no understanding and will reply something like: “Thank you for your response. Since I am still learning I will transfer you to the keypad selection system”.

This way you can:

  • Learn what your callers are actually asking for
  • Get all the data you need to train your own model

In each of the miniApps you can always group the different existing intents together or add custom Intents on top of that using machine learning.

Announcement miniApp

Ability to create tailored announcement messages with dynamic values from other miniApps. This means you can specify different announcements based on information like where the customer is calling from, what time of day it is, or what is the nature of the customer’s inquiry.

In User Functions you can use a JavaScript function to do computations with data that you are going to use in the announcement.

YesNo miniApp

In a YesNo miniApp you can:

  • set a welcome prompt, as well as the initial question prompt and two rephrase attempts
  • control how to manage Agent requests
  • set up error handling and customize how and when the call ends in case of errors in the dialog.

Web Service miniApp

Web Service miniApps allow you to pull data from an endpoint by making API requests, and manipulate that data as you see fit and use it in other miniApps (using Flow).

In this miniApp you can:

  • specify Wait announcements the customer will hear while the data is being pulled
  • make API requests using GET, POST, PUT or DELETE
  • set API request timeouts and fallbacks
  • configure custom JavaScript functions for data processing and manipulation
  • specify values the miniApp will return as output

Intelli miniApp

Intelli is a simplified version of Web Service, so it allows you to process data without hitting an endpoint.

Additionally, in all of the above miniApps you have the ability to choose the voice of the announcer, as well as the rate and the volume.

Flow miniApp

A Flow miniApp is a DiaManT application that can combine several miniApps in order to implement an effective business logic "flow". At the moment, it is not available in the free trial mode; if you would like to build your own Flow application, please contact Sales.