4. DeepPavlov.ai
Good*
Bad
Bad
Bad
Terrible
Soon to be great
Hard to do
Limited by design
Super hard to do
Needs lot of work
UI Technique Status Potential
* For Non-native English speakers it’s still a pain
Voice Input
Natural Language Understanding
Voice Output
Intelligent Interpretation
Agency
5. DeepPavlov.ai
Bad
Bad
Terrible
Hard to do
Super hard to do
Needs lot of work
UI Technique Status Potential
* More of an engineering problem
Natural Language Understanding
Intelligent Interpretation
Agency*
9. DeepPavlov.ai
What Is Alexa Prize?
The Alexa Prize is a global competition for university students
dedicated to accelerating the field of conversational AI.
Teams of university students will develop a socialbot, an Alexa skill
that converses with users on popular societal topics.
The grand challenge for the Alexa Prize is to create a socialbot that
can engage in a fun, high quality conversation on popular societal
topics for 20 minutes and achieve an average rating of at least
4.0/5.0.
10. DeepPavlov.ai
DREAM socialbot
architecture
1. Multiple Annotators are used to
extract information from the
user input.
2. Skill Selector defines a subset of
active Skills based on the
extracted information.
3. Selected Skills propose their
response candidates.
4. Finally, Response Selector picks
a response to be sent to
Response Annotators and,
eventually, to the user.
All elements of the pipeline are
running asynchronously with two
points of synchronization: Skill
Selector and Response Selector.
Dialogue State serves as a shared
memory.
11. DeepPavlov.ai
DREAM socialbot
infrastructure
1. The core of the DREAM socialbot is implemented with
DeepPavlov Agent (DP-Agent) framework. It
orchestrates services for Skills, Annotators, Skill
Selector and Response Selector, and is located on AWS
EC2 instances with Docker Swarm.
2. Dialogue State history is stored on a separate instance
with MongoDB.
3. AWS Lambda performs HTTP requests to the DREAM-
agent by sending ASR tokens.
4. Testing infrastructure consists of Telegram bots for
interacting with the dev version of the socialbot or with
selected conversational skill only.
5. Dialogue analytics tool and dashboard were located on a
separate EC2 instance.
6. Cluster and application monitoring had configured alerts
to email and Slack.
12. DeepPavlov.ai
DeepPavlov
Ecosystem
▪ Open repository of NLP models
and pipelines
• easy to find and reuse NLP
components for development of new
skills or extension of existing
▪ Open repository of conversational
skills
• alternative implementations of the
most popular skills
▪ Open hub for AI Assistant
distributions
• general and domainindustry
specific distributions of skill sets
13. DeepPavlov.ai
Conversational AI
DeepPavlov Stack
Multiskill
orchestration
Conversa-
tionalskills
NLP
frameworks
ML platforms
Proprietary Open Source
▪ Multiskill orchestration
• DeepPavlov Agent is an engine for
conversational skill deployment and
orchestration
▪ Conversational skills
• DeepPavlov Dream is a collection of pre-
build conversational skills and a default
distribution package for Dream AI
Assistant
▪ NLP frameworks
• DeepPavlov Library provides pretrained
models and simple declarative approach
to build NLP processing pipelines
▪ ML platforms
• TensorFlow and PyTorch as backends
14. DeepPavlov.ai
HOW TO BUILD MULTISKILL
AI ASSISTANT WITH DEEPPAVLOV
LIVING IN PANDEMICS
MAKES US LONELY
Wouldn’t it be nice to have a friend to care about us?
15. DeepPavlov.ai
MEET GERTY 3000
Can help Sam with problems on
the Moon Base Sarang?
Can entertain Sam? Yes ✔
Yes ✔
Can we emulate it with
DeepPavlov DREAM?
Yes ✔
Main Question
Functionality Analysis
Case: Gerty
In the Moon Movie
19. DeepPavlov.ai
Deepy:
HarvMaint.: Configs
What is (all) harvesters’ status?
Intents
What is harvester status?
Prepare rover for a trip
domain.yml
intents:
- all_statuses_request
- status_request
[..]
- trip_request
responses:
utter_status_request:
- text: "The harvester {harv_id} is {harv_status}.“
[..]
nlu.md
## intent:all_statuses_request
- What is the harvesters status?
- What is the combines status?
[..]
stories.md
## harv_status + prepare_trip
* status_request
- utter_status_request
stories.md – training for dialogs
RASA Configs
nlu.md – training for intents & slots
domain.yml – basic ontology for skill
Simple and easy to use
20. DeepPavlov.ai
Deepy:
Chit-Chat Skill (AIML)
<?xml version="1.0" encoding="UTF-8"?>
<aiml version="2.0">
<category>
<pattern>I AM ^ TIRED</pattern>
<template>
🙁
<random>
<li>Get some sleep<get name="name"/>.
You're very tired.</li>
<li>Have a rest and be happy! How can
I help you?</li>
</random>
</template>
</category>
[..]
</aiml>
</xml>
Assistant Profile (Name, Place, etc.)
Patterns
Greeting scenario
Topics
Looks up for patterns
Dialog Processing
Picks random pre-defined response
If not sure, confidence is low (0.2)
Returns response + confidence
21. DeepPavlov.ai
Deepy 3000: Demo
A prototype of a fictional Moonbase A.I. Assistant, inspired by the Moon Movie
(2009) made by Duncan Jones