run
Revised on: May 30, 2018

Let's focus on some core problems faced by an end user that makes existing chatbots inefficient

  • Conversation breaks when the user gets an irrelevant answer from the bot
  • Bot only replies to trained intents, which eventually fails in responding to partial questions
  • The bot is not getting logically trained based on the provided intents
  • Not able to identify user behavior in a conversation

How we gave intelligence to our bot to overcome the existing problem

  • From the provided intents we add additional information to enrich the knowledge of bot. Which in turn increase the accuracy and can provide expected results to the end user
  • We solved the responds of partial questions effectively by Identifying the Synonyms, Hypernyms, Hyponyms and Meronyms of words in a sentence using our active word and sentence API
  • We build our model to understand the connection between the sentences using logical form API
  • Our model can Identify the context of user behavior by analyzing the conversational flow throughout the session and help with quick responses to drive user conversation

What makes Yekaliva.ai a intelligent bot

The brain consists of huge process

Yekaliva ML
Machine Learning

Streaming Machine learning model with classic NLP and statistical functions to classify and segment related intents with questions in order to score dynamically.

Yekaliva Bidirectional Neural Model
Bidirectional Neural Model

A Bidirectional Neural model can convert words into vectors and simplifies it into simulated matrix graph and by using neural classifier, user input matched with specific intent.

Yekaliva Graph Neural Model
Graph Neural Model

We use Graph neural model as it combines vector data and enriched data relationship in a matrix which serves the accurate answer every time by keeping the accuracy high.

Yekaliva Entity Label Graph
Entity Label Graph

Our Entity Label Graph (ELPG) is an independent entity used to identify the independent and distinct words in a sentence and mapping them using answer derivation this is done in order to filter a particular metaphor.

Yekaliva Logical Graph
Logical Graph

Graph can represent NLP data flow through a series of computation in terms of relational and logical dependencies between individual hyponym and hypernym.

Yekaliva Active Knowledge Graph
Active Knowledge Graph

Bot’s Core Knowledge Graph built with existing intent database and external data via data enrichment that collects millions of conceptual and relational data behind the keywords.

Set up your bot today, pick a plan later

Try Yekaliva.ai free for 7-days, no payment information required.

Start Free Trial

Already a customer? Log in