Enterprise - Class Conversational AI Platform

With Natural Language Processing (NLP), Virtual Dialog, Omnichannel interface, Building Tools and Programmable Communications

  1. Architecture & Components

    Overview of the Technology

InteliWISE delivers the Comprehensive Conversational AI Platform, for developing, curating, and delivering interactive enterprise-class Virtual Assistants and Chatbots.

The high-level Architecture graph is presented below:

Our Platform consist of 3 layers:

  1. The Omni-channel UI: the omni-channel User Interface - the front-end module or API / SDK, applicable to web services viewed on desktop, mobile or social media.
  2. The AI-Engine which is powering the man-to-machine conversation; this module enables understanding consumer’s question, its intent, its context - with the use of NLP (natural language processing) technology, learning from past conversations, and providing instant, automated answers.
  3. Administration, integration and AI - Content - management tools; that enable brand’s teams to augment, scale and integrate with external IT systems, incl. CRM, eCommerce, Marketing Automation etc.

The Conversational AI consist of intertwining, yet independent modules, that provide with full capabilities for powering corporate Intelligent Automation initiatives, specifically in Customer Service, eCommerce or Service Desk.

At the heart of the system there is the US-Patented NLP (Natural Language Processing) engine, powering fully automated conversations for enterprise use. It helps to detect and classify intent of a user question. It then generates fully automated, scripted responses, using the Rule - based Virtual Conversation/ Dialog module. The system enriches customer experience by making man - to - machine communication natural.

For maintaining, developing content and conversational capabilities, it deploys a number of administration and content management tools. After analysing the Reporting section, one can add new sentences or change the existing content.

It also has a robust, programmable API that enables integrations with external systems. For supporting Omni-Channel, there’s a module that can escalate conversations to live agent in omni - channel model, whenever AI can not handle right responses.

InteliWISE solutions are patented in the United States covering systems and methods for generating and implementing a computer generated human – a machine web interface based on natural language processing and avatar virtual agent based characters.

Key Components

The key components/ modules of our technology:

Deployment

Both on-premise and managed (cloud) is available. Our standard solution is a Cloud-based service but our competitive advantage is that we deploy 1:1 version on - premise (on customer’s infrastructure).

InteliWISE NLP (Natural Language Processing) Engine

Intro | InteliWISE NLP

InteliWISE provides all the necessary tools to build and deliver intelligent, enterprise - class Conversational Agents (CA), also called AI Chatbots. Our comprehensive platform enables to understand user intents and respond with natural feel. It is powered by several powerful machine a learning technologies and a combination of multiple knowledge sources.

The foundation of those technologies run on the Natural Language Processing (NLP) - the industry-leading tech used today as a man-machine interface. In a nutshell, it translates text or spoken user questions into actionable responses or actions.

The platform is powered by hybrid NLP intelligent automation technology, that combines

- the full predictability and control of US-patented, InteliWISE rule based NLP,

- with the scalability of the Machine Learning (ML) approach.

This combined approach guarantees top-of-class quality of language understanding (intent, query) and faster and more cost-effective process of Chatbot training.

The platform is powered by hybrid NLP intelligent automation technology, that combines

- the full predictability and control of US-patented, InteliWISE rule based NLP,

- with the scalability of the Machine Learning (ML) approach.

Untitled

Input | Query Handling

‘Scripts’ are on the entry level of processing of the question. They drive user actions, like tracking and user behavior.

Scripts allow for:

  • execution of automated tasks

eg. when visiting specific pages or sequence of pages or even counting number of pages and firing

tasks depending on those counts - after meeting some defined conditions we can navigate to

selected page with additional content or present topic from knowledge base or just connect with live-assisted chat - we can hook product presentations too basing on dynamic content delivered by company and

present it depending on section of page (promos);

  • and defining conditions
    • eg. whether to show the end-user module,
    • whether to start Live Chat session,
    • whether to launch contact form.

The Appserver (backend logic module for website module) tracks user actions (pages, urls, tasks) and decides if conditions are met. On the bottom level Appserver can decide to use AI Engine when user action is typed by hand or user picks suggested item. AI Engine can exploit scenarios defined or match most appropriate answers from Knowledge Base.

  1. Query Processing & Response Generation

The InteliWISE NLP (Natural Language Processing) is powered by:

(1) statistical classification with (2) text correction and (3) intent matching, (4) Machine Learning (ML) tools for training and response augmentation. The Engine runs:

  • compound classifiers and spanning - meaning weights and text distances;
  • first stage text corrections basing on our own dictionaries + we can attach external engine for example Google Search for text correction suggestions
  • prefix simplifications for variety of language constructions with the same meaning eg. you can ask “what you know about X” or “tell me about X” - so we simplify it into one variant - all those prefix variations simplify thousands of variations into one simple case,
  • Lemmatization - we simplify words into base forms, for better matching,
  • Synonyms of words, to allow millions of combinations to flatten into single case,
  • Keywords important from designer perspective to boost their weights in the matching algorithm,
  • Statistically weight and match and then score millions of solutions to pick the best one or judge if the input was not close enough to store knowledge,
  • Finally it uses semantic relations such as synonyms, antonyms, etc. InteliWISE NLP Engine is also able to handle misspellings and grammar errors.

High level query processing by inteliWISE Rule - based Engine is presented below.

  1. Dialog Management

A flow of questions and responses can be designed around the specific business process.

The system enables to convert traditional FAQ or static search into natural Conversation. An example can be step - by - step guided workflows, or a true natural conversation.

Key Features of NLP

Handling questions in natural language

After quickly and accurately determining the subject and intent of an issue, our NLP-powered virtual agent understands:

  • slang,
  • misspellings,
  • acronyms,
  • and grammatically incorrect words or phrases;

● Deploys the right social protocol for human – like behavior incl. greetings and farewells, criticisms, compliments;

● Delivers social engagement capabilities beyond business centric dialog.

Handling spelling/grammatical mistakes

At input level, data is preprocessed with spelling correction. 2 methods of spelling correction are deployed:

  • using external systems e.g. google search and its ‘suggestions’ feature

  • or using our stored data (text corpuses based on proper data / e.g. BBC text corpuses or client provided text corpuses). When using stored text corpuses, system compares user typed text with stored structure to check the spelling and correctness of text and suggest corrections for further investigation of internal engines (as alternative variants of user input).

Synonyms and prefixes

After text pre-processing the Engine composes alternative variants using synonyms and prefixes (e.g. there are many ways to ask something about - all variants are meaning the same, thus have strictly different language constructs).

Synonyms and antonyms give our solution the power of generalization over many language constructs - and do not change the meaning of user input (positive and negative weighting).

After creation of many language constructs the system removes from each sentence as many parts as possible without changing the meaning of it. It has been observed that many language constructs are very redundant and do not change the meaning and the quality of matching process (not removing them could make the process of factorization of results the impossible or asymmetric).

At last stage our technology analysis all statistically matched rules to choose the best chosen Response.

Hand over, handling “don’t knows”

Contextual auto completed

The “auto-complete” suggests one of best matched responses from the knowledge base. It narrows down selections and improves the quality of search.

“Don’t know”/ “Query miss” templates

Whenever NLP can not find the right response match for the question, the InteliWISE solution provides has number of ways to handle them:

  • Varied “don’t know” templates, tracking how many times user confused the system to change the system reaction,
  • Present “auto-complete”
  • Hand - over - automatic switch of the user to the live agent
    • on chat
    • or ticket
    • or call.

Broad Coverage - Contextual Auto-complete

Broad coverage refers to having a large range of domains covered by a system, for example attempting to cover all possible search topics. Using Contextual Auto Complete we maximize the success of a first hit response.

Above is the schema of how InteliWISE Engine processes the query

Handling clarifications and scenarios

Scenarios - free and guided dialogue

When user input is not clear enough or client needs the process separated in multiple stages (queries), the scenarios can be used. We can provide the net over stored knowledge to drive the user through the factory process or ask him additional queries to clarify the intention. We can lock the user on the path or provided the “quit” option. User is presented with suggested next steps and creates the feeling that system is driving him in the right direction. There can be as many scenarios built on the same data as we need. We can reuse once entered data to design many processes - thus not replicating stored knowledge and making it possible to fast correct and edit.

  1. Use of context

Different contexts can be analyzed before the answer is delivered, incl.:

  • History of the current conversation
  • History of prior conversations by current user (useful for personalization/preference setting over time)
  • User interaction data – collection of interactions across many user
  • How and Where the interaction started on a page
  • Answers from multiple resources

  1. Multi-language recognition of chosen languages
    1. Application of full semantic analysis of queries, dictionaries and databases of ready-made answers in a given language (selected languages, incl English, Polish)
    2. Use of a combination of automatic translators and dedicated semantics:
      1. Each user input is translated into knowledge base language e.g. from Spanish into English,
      2. then user input is processed internally as usual, but all system outputs are back translated to the user language e.g. from English into Spanish. Such solution operates only 10-15% less reliable than originally translated knowledge base. There is no time waste for translation services, no time spent on setting up the services - all works on the same knowledge base serviced in one language and serves support in multiple languages (as many as needed).

InteliWISE Programmable Escalations (Hand-over) Module

The module enables to fully pre-program Virtual Agent/ Chatbot behavior whenever the end-user question can not be handled automatically (by AI/ NLP or automated answers).

This powerful feature enables to patch the conversation with Chatbot | Virtual Assistant to a live agent, through chat, voice or video. This forward is seamless for an end - user, but the live agent is feeded with the context of the previous conversation with Chatbot. This feature maximizes the productivity of the intelligent automation software while optimizing the end-user experience (who is not left without response).

The module enables to program switching to the Live Chat software, Video Chat, or Contact Center voice system.

Knowledge Base is utilized by users for self-service. It can address most common and less complex questions. However there are situations which require human interaction. Thus Inteliwise System also provides seamless switch between Virtual Agent (automated solution) and Live Chat operator (human being). The conversation can be switched between Knowledge Base and Live operator many times, as well as transferred between multiple operators.

The Live operator is also able to search through the Knowledge Base for faster responses. The system even enables the operator to manually switch the conversation back to the Virtual Agent (Knowledge Base) if the topic is well covered by the Automated Solution.

InteliWISE NLP Content Development, Administration & Management Module, NLP Training

Our team & partners successfully deployed more than 150 enterprise virtual assistants, and this was feasible through our comprehensive, structured approach:

  • Delivering Tools for deployment, administration and management;
  • Delivering structured set-up (incl. content development) methodology via our Consulting & Professional Services
  • Leading projects and providing technical support.

Our clients and partners are able to fully manage our Chatbots, Live Chats and other solutions with the use of comprehensive administration tools:

  • InteliWISE NLP Engineer - Advanced Content Authoring Tool; enable full range of adjusting semantics, content authoring changes and system modifications; it combines features that enable Editors to create, secure, and manage every Knowledge Bases designed by or for the client. Our tool enables to edit, add, or delete Q&As, scenarios, and scripts on number of different levels.
  • InteliWISE Panel - enables management and customization of GUI, customer engagement, Reporting, set-up; web - based tool.

InteliWISE NLP Engineer | Builder (developing Content/ teaching Chatbots)

The tool for the comprehensive Virtual Assistant knowledge management is InteliWISE Engineer. It is also the base where the automated Q&A are stored and the content management system

Knowledge is structured in separable knowledge bases with given weights. We can deploy new solutions as mixture of existing data properly weighted (trained). We can make some language constructs more important than other by intelligent weighing (by statistical learning or by weighting by humans).

We organize data in “usage clusters”:

· spell correction data (built by hand or automatically using high quality text corpuses);

· synonyms and antonyms (built from generally available sources e.g. thesaurus or built upon client request);

· personality knowledge (built to give small talk feature and to appeal user when using our platforms);

· client knowledge (built by hand from clients’ solutions or automatically from user given sources: FAQ, Web Page, Corporation Documents);

The example view of knowledge builder is below.

The complete List of features - on request (sales@inteliwise.com).

Full manual on NLP Content Development and Training is available on request.

AI Training, Chatbot learning

There’s a number of ways in which the Chatbot knowledge & content is being augmented - from previous conversations, on

All the query miss - questions which were unanswered by Chatbot, are:

  1. Recorded and stored in Reporting section of NewPanel
  2. Processed by Level 2 AI algorythm; suggestions are send to Content Engineer tool for approval of Content Manager

InteliWISE Administration Panel

The unique, InteliWISE Panel enables administration and customization of almost every element of the InteliWISE Virtual Agent Solution.

Any authorized project team member can run manage changes in:

  • The Look & Feel of the solution
  • The moment of engagement (i.e. when Virtual Assistant shows up to a user)
  • The first message/ welcome message
  • The customized script.

Examples of views from the Panel are below:

The complete List of features - on request (sales@inteliwise.com).

InteliWISE Voice & Messaging API (comm.AI)

Our API cloud communications platform enables customers and developers to embed PROGRAMMABLE digital telephony and cross-channel messaging within web applications.

The Java - Script powered engine enables programming.

InteliWISE Omni-channel End-User Interface (UI) Module

The user-interface of InteliWISE platform is a full omni-channel, including Web - based module (Widget, pop-up), Mobile, In-app, Messenger (incl. Facebook Messenger).

From the brand’s perspective, the look and feel of the User Interface can be fully customized through Administration Panel.

Welcome messages and initial dialogs can be programmed or can show - up contextually to a specific customer behavior with the use of rule-based Engagement Triggers.

The content may be text - based but also may feature multimedia or emoticons, movies, pdf, schemes.

The example of the web - based Widget:

InteliWISE Engagement Triggers & Scenarios Module

InteliWISE Reporting Module (Real-time analytics and performance dashboards)

The InteliWISE performance reporting tools provide critical customer information, including the full transcript on exactly what your customers are talking about – and how. All the reporting is presented in the web-based Control Panel. The access to the Panel and Dashboard is given to our customers, or InteliWISE team sends the reports in the form of e-mails. The performance monitoring can also be integrated with Contact Center Solution, or other systems, that would enable monitoring of important customer information from Virtual Agent without leaving company’s preferred analytic systems.

Our reporting tool will report on information like:

  • What exactly did customers ask for?
  • How they formulated questions in their words?
  • For which questions they did not find answers on site?
  • Interaction recording (chat history) - you can review any of the past conversations. Although it provides a lot of useful information, going through all of the conversations may be a very time-consuming task.
  • Feedback management - the list of questions that have been asked most frequently. This information lets you evaluate what the customers are interested in and which knowledge areas should be expanded and enhanced. Users’ queries that have not been recognized by the system are listed in the “Missed Questions” section. It is an invaluable source of information in terms of continuous expanding and improving of your Virtual Agent’s knowledge.
  • Escalation management - the performance of scripts having live chat invitation included shows escalation between Virtual Agent and Live Chat operator. Integration with external analytic systems.

Selected views from our Reporting module is presented below:

InteliWISE Integration (API) Module

Integrations with external IT systems are essential for Chatbots to provide the ‘Conversational experience’ in the problem solving or troubleshooting flow.

Value proposition / use cases for integrations are:

The main value proposition is to

  • Feature dynamic data in the Chatbot conversation, examples:
    • account balances, recent transactions, reservations and other end - user relevant, specific data from IT systems,
  • Pull-out data from other IT systems that are relevant for conversation context,
  • Push data and transcript from Chatbot conversations to external, 3rd party reporting systems.

Integrations allow for personalized responses and dialog, based on the use of information about the customer, existing service, status, balance and other. That means that they require integrations with 3rd party IT systems. The

Integrations offer support for sales and marketing efforts, examples:

· show external reports;

· show personalized data appropriate for customers who are currently logged in;

· build scenarios which collect data forms and pass them “on-the-fly” to external data sources and subsystems;

· fetch and store parameter values;

· call external functions to fetch customized values;

· show data which actually depicts dependency by creating “collecting scenarios” and then fetching and storing values that some other values depend on;

· connect to any type of SOAP based WEB service (upon configuration).

InteliWISE Platform provides rich API that enables integrations with 3rd party software, across channels (omni-channel.

There’s number of options for integrations:

  1. The use of our API and documentation, including manuals,
  2. Using ready integrations, including off the shelf (OTS) and custom integrations,
  3. Using our experienced professional team for ‘turn-key’ integration services.

Standard integrations

Our Platform enables integration layer using REST or SOAP services.

Based on integrations, one can combine complicated procedures and interchange protocols with a simple abstraction of “parameters”. In our subsystems we define the “parameters” which trigger external procedures in the External Data Source. Our subsystem maintains dependency relations between the “parameters”.

For example if a procedure call requires to pass 2 “parameters”, the collecting scenario should collect those “parameters” before calling that special procedure. By collecting these “parameters” the scenario satisfies the requirements of the External Data Source. The “parameters” can be placed in the display templates presented to the end-user (together with static text).

Prior to using the “parameters” one has to define how data is fed and stored in terms of external data sources and what functions or procedures are responsible for data management. Some choices include:

  • data type,
  • possible ranges
  • and visual controls that will drive the interaction with the end-user.

When the “parameters” are placed in the display template they are transformed into interactive controls that are fed with the data stored in the External Data Sources. Possible values can be obtained when showing these controls to the user – in order to minimize the chance of erroneous choice or value. When the user changes the value (this option can be granted to the user) of the visual control, it is stored in the External Data Source for further usage or simply as a collection managed in the current conversation session.

Nowadays information systems have to be based on an “open architecture” approach. Every system uses its own data but in many cases it needs to use external data and store data externally. Our scenario driving mechanisms can be extended by using external data stored on subsystems of our partners and clients. Many types of information can be mixed in order to form consistent data exchange scenarios. Our system is based on the idea of “parameter” abstraction. Those simple abstractions are transformed into complex and powerful data inter-exchange scenarios which enable you to interact with various External Data Sources.

The whole process is shown in the diagram below

.

Subroutines

Subroutines are a light and well isolated integration layer allowing us to integrate any arbitrary process or data flow into a conversation with the AI Chatbot.

Subroutines, unlike a chat with Chatbot, are a strictly controlled (but not necessarily linear) conversation units that may span across one or more VA responses. Every unit is independently developed to suit very specific customer needs when it comes to deeper integration with customer systems, for example ordering a ticket through conversation with VA, or automatically checking account or case status.

Every subroutine can posess its own configuration space, securely stored server side. It is a general purpose storage that allows to change parameters that may change over time, like access tokens or API keys.

Once developed, subroutines can be attached at any number of VA conversation nodes. Once VA node with subroutine is reached, the system passes control flow to the subroutine. From this point, until subroutine flow ends, conversation is strictly controlled by its process. Despite that, conversation with its flow is registered as a regular VA chat for the statistical purposes.

Customer metadata

The personalization of Chatbot’’s responses requires user identification.

During a standard interaction, every ‘public’ user is anonymously identified by our systems in the scope of being new or returning user.

In case when user is identified, for example users logged-in to services, our system offers additional API that allows to mark a user with a freely structured data object. This data object can be later used to identify user in the Agent’s console or in the transcripts, or to present a personalized response.

InteliWISE Omni-channel Agent’s Portal | Console

Agents in Call center or help desk handle user questions via our intuitive Agent’s Portal.

It enables live agent to answer consumers on many channels, incl.

  • Chatbot - Agent can preview the entire conversation with AI-powered chat
  • Live chat,
  • Calls, including mobile and VoIP
  • Tickets
  • Facebook (wall and messenger)
  • Video call, screen-sharing

For more information, look at https://inteliwise.com/products/ai/chatbot-for-helpdesk/

Supplementary documents

Ask us about detailed documentation or product sheets, i.e.

InteliWISE system Feature List

InteliWISE AI Chatbot Content Development Manual

InteliWISE Cloud-based Set-up

InteliWISE On-Premise Deployment

On-line: DOCS.inteliwise.com

AOB