Description
difficulty: easy
duration: 350/170 hours
mentor: @kudep
requirements:
- docker - basic
- python - advance
- REST API - basic
useful links:
- Description of the framework https://github.com/deepmipt/dialog_flow_framework
- Repository of Dialog Flow Engine https://github.com/deepmipt/dialog_flow_engine
- Documentation of Dialog Flow Engine https://dialog-flow-engine.readthedocs.io/en/latest/index.html
- Repository of Dream https://github.com/deepmipt/dream
tl;dr: Dialog Flow Framework (DFF) is a free and open-source software stack for creating chatbots, released under the terms of Apache License 2.0. DFF can be extended by add-ons, these add-ons are implemented as python packages and distributed by PyPi.
In this task we are going to develop Python packages for dialog data exploration and visualization utils based in DFF to analyze user behavior by using ML annotators, statistics methods based on data from databases.
Idea Description:
Dialog Flow Framework (DFF) is a free and open-source software stack for creating chatbots, released under the terms of Apache License 2.0.
DFF is developing an open source community for writing conversational systems using artificial intelligence. We promote the use of Machine Learning in modern conversational systems. One of our projects is Dream Dream Socialbot was originally developed for participation in Amazon Alexa Prize Challenge, and then it was converted to an open-source platform DeepPavlov Dream for building modular dialogue systems.
Common description is available here
A presentation about Dialog Flow Framework is available here
A YouTube video about Dialog Flow Framework is available here
Currently, more than 50 dialogue services have been based on DFF and their number is growing at an accelerated pace. The proposed tasks for this project will give more opportunities to those who want to create a new dialogue service using machine learning.
More examples of Dialog Flow Framework you can see here:
DFF can be extended by add-ons, these add-ons are implemented as python packages and distributed by PyPi.
In this task we are going to develop Python packages for dialog data exploration and visualization utils based in DFF.
Examples of visualization:
**- User behavior graph
- User path**
Scenario node counters
You can check add-on examples for DFF here.
You can check the earliest version of visualization tools here.
During the task, you will need to write a Python package that will allow you to analyze user behavior by using ML annotators, statistics methods based on data from databases like Postgresql or ClickHouse, etc.
Coding Challenge
Write a simple backend service that will respond to templated http-pages based on FastAPI, with the use of info from postgresql and wrapped in docker containers.
Coding Challenge
Write a simple visualization service that will be developed based on streamlit with using pandas, Postgresql as data storage and wrapped in docker container