LogoLogo
  • Dhee.AI
  • Concepts We Work on
    • What is NLP?
    • Natural Language Parser Pipeline
    • Word Embeddings
    • Textual Entailment
    • User Intent Recognition
    • Document Reading
  • Getting Started
    • Train your bot
      • Create your bot project
      • Train for Intent recognition
        • Define the Slots
        • Train the Triggers
        • Train for Stop-Triggers
      • Configure Agent Responses
      • Write your REST Endpoints
      • Configure Endpoints
      • Configure Workflow
        • Read Inputs
        • No Code Approach
        • REST Endpoints Approach
        • Acknowledge User
    • Build and Deploy
      • Build
      • Test
      • Deploy
        • Embedding Widget in Apps
  • Platform Reference
    • Manage Projects
    • Agent Settings
      • Basic
        • Language
        • Domain
        • Voice
        • Avatar Settings
      • Widget
        • Widget Theme
        • Widget Label
      • Advanced
      • Team
        • Development Team
        • Support Team
      • Import Export
      • Emailer
      • Billing
      • Botstore
    • Knowledge Management
      • Document Reading
      • Frequently Asked Questions
    • Intents and Automation
      • Intents
        • Slots
        • Triggers
        • Stop Triggers
        • Special Intents
      • Skills/DSM
        • Dialog Actions
        • Dialog State Transition
          • Slot State
        • Dialog Workflow
        • Skill API
      • Backend API
    • Extended Message Types
    • Entities And Other Data
      • Entities
        • Multilingual Entities
        • Language Specific Entities
        • Custom Entity Types
      • Agent Responses
        • Multilingual Responses
        • Missed Query Responses
        • Support-Unavailable Responses
      • Directive Utterances
        • Customize Inputs
        • Customize Outputs
      • Translations
      • Query Substitutions
      • Abbreviation Texts
    • Test & Deploy
      • Build
      • Test
      • Deploy
        • Launchpad
        • Widget
        • Signal
        • Telegram
        • Google RCS
        • Facebook
        • Alexa
        • Whatsapp
        • Custom App
        • Voice
        • Telephony
        • Email
    • Reports
      • Statistics
        • Summary
        • Daily Reports
          • Conversation Analytics
        • Weekly Reports
        • Monthly Reports
        • Output Spreadsheet
      • Conversations
        • Conversation Logs
        • C-SAT
      • Generate Report
        • Lead Report
        • Category Report
        • Device Demography Report
        • Utterance Report
        • Missed Query Report
        • Location Report
      • Report Settings
        • KPI
        • Schedule Report
      • Personnel Audit
        • Development Team
        • Support Team
          • Supervisory Sessions
          • Login Information
      • Bot Store
        • Reviews
        • Reported Issues
  • The Social Desk
    • Whatsapp
      • Customer Chat
      • Manage Template
      • Outreach Campaign
        • Create new
        • Manage Campaign
    • Reports
    • Settings
      • Team
      • Whatsapp
  • Extras
    • External Links
  • News
    • Dhee.AI's Edge Server for Telephony Achieves Breakthrough Optimization on Intel Architecture
  • Dhee.AI Telephony Edge Servers: Elevating Conversational AI Experience with Edge Computing
  • Pinnacle Teleservices Joins Forces with DheeYantra to Supercharge Conversational AI on WhatsApp
Powered by GitBook
On this page
  1. Concepts We Work on

Document Reading

I'm a bot. Documents are not boring for me!

PreviousUser Intent RecognitionNextGetting Started

Last updated 3 years ago

Neural Document reading is a task where a Deep learning Model finds an answer to a query from a document/context.

Components of Document Reader

  1. Elastic Search as a Document store

    The documents are preprocessed and are stored in elastic search indexes

  2. Neural/statistical Document Ranker

    Document Ranker retrieves Top n documents out of m documents (m>>n) which would most likely have an answer for the incoming query. The most likely documents that would contain answer are chosen based on:

    • Neural approach: Semantic similarity of query embedding with the already existing Document embeddings

    • Statistical approach: Based on word overlap in question to that of document

  3. Transformer Reader

    • Extractive Reader

      we feed question and context (the list of documents shortlisted by the document ranker) as input to Transformer. The Embeddings generated from the transformer layers are passed through two separate Feed-Forward neural networks. One of the Networks predicts the start token index and the other predicts the end Token index. The Probability distribution over the words in documents (for both start and end token) is used to retrieve the answer

    • Generative Reader

      Generates a novel answer (not necessarily a span of text) from the document

Document Reader