OUR TECHNOLOGY

AUTOMATE COMPLEX HUMAN CONVERSATIONS WITH OUR ADVANCED CONVERSATIONAL AI PLATFORM

LEARN

DOMAIN ENTITY MODELING

DIALOG FLOW DESIGN

DATA EXTRACTION & INTEGRATION

USER LOG ANALYSIS

UNDERSTAND

SPELLING CORRECTION

LINGUISTIC DATA ENRICHMENT

NAMED ENTITY RECOGNITION

INTENT DETECTION

TALK

USER CONTEXT MGMT.

PARALLEL DIALOG HANDLING

DYNAMIC ANSWER GENERATION

CHAT PLUGIN FOR MOBILE & WEBSITE

INTEGRATE

ON-PREMISE OFFLINE CAPABILITY

DOCKER DEPLOYMENT

REST & GRPC APIS

BUSINESS PROCESS INTEGRATION

ARCHITECTURE

CONVERSATIONAL AI PLATFORM

LEARN

DOMAIN ENTITY MODELING

Modeling of knowledge domains

-

Definition of entities and synonyms

-

Description of entity relationships

CONVERSATION FLOW DESIGN

Visual modeling simple and complex conversation flows

-

Automatically learning conversation models from user data

-

Measuring conversation quality via graph-based “model checking”

DATA EXTRACTION & INTEGRATION

Extension of ultra-small data sets with external data sources

-

Advanced web scraping and crawling technologies

-

Integration and enrichment of existing training data

USER LOG ANALYSIS

Visualization of user conversation flows

-

Analysis of triggered user intents

-

Evaluation of used entity vocabulary

UNDERSTAND

SPELLING & TYPING CORRECTION

Correction of spelling mistakes for multiple languages

-

Correction of typing errors e.g. on mobile phone

LINGUISTIC DATA ENRICHMENT

Lemmatization based on multiple thesauri

-

Stemming of words

-

Decompounding of words

-

Chunking of long sentence structures

NAMED ENTITY RECOGNITION

Hybrid approach with neural networks, classification and rule-based techniques

-

Machine-learning & deep learning based entity model training

-

Highly performant named entity recognition of domain specific custom entities

INTENT DETECTION

Hybrid approach to intent detection considering the context of the user at any time in the matching process

-

Ensemble-based intent detection with 12 performant intent detection algorithms for an even higher accuracy

-

Performance analysis based on standard machine learning key performance indicators (KPIs) such as accuracy, f-score, precision, recall, etc.

TALK

USER CONTEXT MANAGEMENT

Management of user context information

-

Detection of the user context based on unstructured user input

-

Capability to add external user context (e.g. sensor from robots or smart homes)

MULTIPLE CONVERSATIONS HANDLING

Multiple parallel conversations supported

-

Conversation flow switch detection with machine learning algorithms

-

Switch between conversation flows with keeping (reinstating) previous context of the user

DYNAMIC ANSWER GENERATION

Enrichment of pre-modeled answers with dynamic content (e.g. from databases, business logic systems, ERP or CRM)

-

SEQ2SEQ model for automatically generating answers (only large datasets of user interactions)

CHAT PLUGIN FOR MOBILE & WEBPAGE

Highly-customizable web chat plugin with full access to source code (colors, size etc.)

-

Secure, fast websocket based chat plugin to easily add to mobile application or webpage

-

Floating icon capability of web chat

-

Fully Facebook Messenger compliant data format for rendering of visual elements, postbacks and webviews

INTEGRATE

ON-PREMISE & OFFLINE CAPABILITY

Installable and deployable on linux and windows server infrastructure

-

No dependencies to cloud or SaaS providers

-

Full offline-capability (no internet connection required)

-

Full control of your data in your data centers

DOCKER DEPLOYMENT

Easy and secure deployment via best-practice container technology based on Docker

-

Highly scalable due to Kubernetes and Docker

REST & GRPS APIS

REST API for training, testing and operational tasks

-

gRPC based on Protocol Buffers Version 3 for training, testing and operational maintenance tasks

-

Fully compliant data format and protocols to Google Dialogflow v2 and v1

BUSINESS PROCESS INTEGRATION

Easy APIs to integrate, control and adjust

-

Easy adding external information from business applications and workflow engines

TECH STACK

INTEGRATION &
DEPLOYMENT

  • GRPC
  • PROTOBUFFER 3
  • REST:API
  • DOCKER
  • KUBERNETES

MACHINE &
DEEP LEARNING

  • TENSORFLOW
  • KERAS
  • SCIKIT-LEARN
  • FASTTEXT
  • SPACY
  • DUCKLING
  • NLTK

DATA
MANIPULATION

  • PANDAS
  • NUMPY
  • DASK
  • MULTIPROCESSING
  • VARIOUS OWN
  • ALGORITHMS & MODELS

PROGRAMMING
LANGUAGES

  • PYTHON
  • NODE JS
  • TYPESCRIPT
  • JAVASCRIPT
  • HASKELL

EXAMPLE TECHNOLOGY USE CASE

One of our showcases is a B2B marketplace for real-time services. This marketplace enables consumers to book professional service providers (e.g. hairdresser, massage, handy man etc.) through social messaging apps and voice assistants (e.g. Facebook Messenger, Amazon Echo etc.) and via a chat on the website. Therefore we developed an intelligent digital assistant (similar to a chatbot but much more intelligent) that matches the service demand of the consumer with a specific service providers’ offer. The interaction with the digital assistant is fully in natural language (German), text and/or voice based.

AI DEVELOPMENT INFRASTRUCTURE

All of our developers are equipped with cutting-edge, high performance and custom built laptops from Schenker Technologies, our close partner for high performance computing. Check out the specifications below. For large AI models we use AWS Amazon IT infrastructure with Tesla graphic cards as GPU computing infrastructure.

ONDEWO AI powered by

TALK TO OUR EXPERTS

We are here to help

We find the right solutions for you

We give you insightful information and explain our pricing

We connect you with valuable and helpful resources

There is something missing. Please fill.
Please hold on while we submit your answer!

ALWAYS UP TO DATE!

Subscribe to our newsletter to recieve information and updates about ONDEWO and advances in Conversational AI.

There is something missing. Please fill.
Please hold on while we submit your answer!