Natural Language Processing Engine

Convert natural language questions and commands to formal queries a computer can understand

Alterra Deep NLP engine

Deep Learning inside – no traditional coding required – just feed the training corpus into the artificial neural network

Powered by Alterra’s phrase2vec phrase embedding and slot filling algorithms

  • Question answering
  • Intent classification
  • Paraphrase detection
  • Natural language command interpretation
  • Phrase embedding
  • Slot filling, entity extraction

Applications

  • Conversational virtual agents, chatbots

  • Call center automation

  • Alexa skills

  • Voice control for mobile apps

AI as a service

APIs exposing Alterra’s Deep Learning AI algorithms

  • 01

    Phraser API

    Semantic classifier for questions and commands. AI to power intelligent agents, Alexa skills and IoT devices.
    Learn more
    API documentation

  • 02

    FAQ API

    Semantic question answering. AI to power your own FAQ bot, knowledge base search or call center automation.
    Learn more
    API documentation

  • 03

    Slot filling API

    Extract entities from phrases. Convert free-text user queries to SQL-like structured queries that can be submitted to databases. Fill out forms via chat, with infinite context, random access navigation, etc.
    Learn more

  • 04

    Time and date API

    Extract time and dates from free text; returns them in a structured form (e.g. “what if we go there next Sunday morning and return between Nov 11 and 13”). More comprehensive and accurate than most other solutions on the market. Pre-trained – no further training required.
    Call us

  • 05

    Alterra Travel API

    Sell flights and hotels via chat. The API takes free-form user query, extracts all parameters such as locations, dates, target prices, class of service, amenities, etc. and returns them in a structured form. Pass these structured queries to a GDS or your booking engine. Pre-trained RNN inside – no further training required.
    API documentation

  • 06

    Destination recommendation engine

    Recommends leisure travel destinations by activities or themes, aka “Pandora for travel” (e.g. “7 day vacation for two in unspoiled nature, with good biking”).
    Call us

01 Phrase embedding

Like word2vec but for phrases

word2vec: word ↔ vector

Semantically similar words cluster together

phrase2vec: phrase ↔ vector

Semantically similar phrases cluster together

Vectorization

Alterra Phraser API

Semantic classifier for questions and commands

AI to power intelligent agents, chatbots, Alexa skills and IoT devices

Convert

natural language questions and commands to the canonical form a computer can understand

Deep Learning

no rules-based coding required

1,000x faster than RNN, and yet even more accurate

Features

  • Intent classifier (paraphrase detector) for natural language questions and commands
  • Users may ask the same question in a multitude of semantically equivalent ways - Phraser will reduce these paraphrases to one canonical form (assign to a pre-defined class)
  • Your program may reply with text or call any other function
  • Powered by Alterra's phrase2vec sequence embedding algorithm - it's like word2vec, but for multiple-word questions and commands
  • More accurate than competing intent classifiers and chatbot frameworks: P@1=90%
  • 1,000x faster than a typical RNN
  • Runs on plain vanilla CPU – no GPU/TPU is required
  • Humans may train the neural network, to make it even more accurate
API Documentation Get you API key

02 Question answering

Convert your static FAQ to a question-answering system

AI to power conversational virtual agents, knowledge base search or call center automation

Alterra FAQ API

Semantic question answering
API access to Alterra FAQ bot

Automate

sales, support, internal IT, onboarding

Deep Learning

no rules-based coding required

Features

  • Natural-language search for FAQs, help centers, knowledge bases, user guides, canned responses
  • Use fully autonomously or pair up with live agents
  • Users may ask questions in their own words
  • Search by meaning, not keywords
  • Powered by Alterra's phrase2vec sequence embedding algorithm
  • More accurate than Google custom search engine and other FAQ search solutions
  • Humans may train the neural network, to make it even more accurate

FAQ API

AI to power your own FAQ bot, site search or call center automation

API Documentation Get you API key

FAQ editor

Create and edit your FAQ file
Label the training set, to improve ML quality
Optional – you can also do everything via the API

More about FAQ editor Open FAQ Editor

03 Slot filling

Extract entities from phrases

Convert free-text user queries to SQL-like structured queries that can be submitted to databases

Deep Learning (LSTM) inside – no keywords, no regexs – just feed your training corpus into the neural network

98% per-label precision

Learn more

Alterra slot filling API

Extract entities from phrases

Convert free-text user queries to SQL-like structured queries

Extract

entities and parameters from natural language queries

Convert

free-text user queries to SQL-like structured ones

Deep Learning

no rules-based coding required

Features

  • Slot filler for natural language queries
  • The API takes free-form user query, extracts all parameters and returns them in a structured form that can be submitted to databases
  • Powered by Alterra's recurrent neural network (two bi-directional residual LSTMs)
  • Supervised learning: a labeled training set is required (e.g. a chat or call log)
  • More accurate than most competing slot fillers: ~98% per-slot precision
  • Example: Alterra travel API – parses flight and hotel reservation requests
  • Upload your training set here or contact us