Knowledge Based Chatbot

Knowledge based chatbot to find apps with desired features. Currently working on a small subset of applications. To adapt this chatbot simply follow the structure in the corresponding knowledge base file and re-train the model.

Description

Uses JSON files to provide knowledge base to the chatbot. Using the Rasa interface this chatbot allows smart user-machine interaction. This chatbot allows user to select a distinct application based on their preferred features.

Sample Story -- find app by features

Your input ->  hello                                                                   
Hello! I am a mobile app assistant. Which features do you need help with?
Your input ->  I want to activate GPS Navigation                                       
Sure! I see you have multiple apps with this feature:
1. OsmAnd 
2. Organic Maps 
Do you wish to use any app in particular?
Your input ->  Whichever has Real-time traffic                                         
Great! Then let's launch OsmAnd!

Sample Story -- Generate

Your input -> hello                                                             
Hello! I am a mobile app assistant. Which features do you need help with?
Your input ->  I want to create a calendar event for my route
Let's get started! The following information is required:
  - start date/time
  - end date/time
  - invites
  - name
  - description
Please provide the 'start date/time':
Your input ->  August 25th
Please provide the 'end date/time':
Your input ->  August 26th
Please provide the 'invites':
Your input ->  ccg.campas@gmail.com
Please provide the 'name':
Your input -> Route Event
Please provide the 'description':
Your input -> If sunny we go on a hike!
Thank you got your information!
Please confirm the following is correct:
  - start date/time: August 25th
  - end date/time: August 26th
  - invites: ccg.campas@gmail.com
  - name: Route Event
  - description: If sunny we go on a hike!

File structure

• \actions
  • ActionQueryKnowledgeBase.py: base actions that can be done using the current knowledge base
  • actions.py: actions that our chatbot can
  • KnowledgeBase.py: Connection with our knowledge based used to extract and manipulate data from it.
  • EventHandler.py: Class controls input requirements from user for feature integrations between two different types.
  • Attention: obtains the attention on input for disambiguation of types
• \components
  • DateComponent: Detects dates in messages (including 'today', 'tomorrow', etc.) and standardizes date format
  • EmailComponent: Detects emails in messages using regex
  • FuzzyWuzzyComponent: custom components used within the chatbot, current FuzzyWuzzyComponent used to identify synonyms and correct spelling errors
• \data
  • nlu.yml: Training data stores structured information about user messages.
  • rules.yml: Training data used to train the chatbots dialogue management model, short pieces of conversations that should always follow the same path.
  • stories.yml: Training data used to train the chatbots dialogue management model, can be used to train models that are able to generalize to unseen conversation paths.
• \models: contains all currently trained models
• \scripts
  • json-to-nlu.py: Convert JSON (rasa_knowledge_base.json) file to NLU file that the chatbot can train with.
  • nltk_data.py: Download required NLTK package.
  • proactive-bot.py: Starts conversation with rasa chatbot externally.
• \test_python: scripts to generate tests with current data (specifically date tests)
• \tests
  • test_stories.yml: validate and test dialogues end-to-end by running through test stories
  • test_data_component.yml: test correct function of date identification component
  • test_email_component.yml: test correct function of email identification component
  • test_fuzzywuzzy_component.yml: test correct function of fuzzywuzzy component (input spelling errors fixed).
• \
  • config.yml: defines the components and policies that your model will use to make predictions based on user input.
  • credentials.yml: credentials for the Rasa X channel
  • domain.yml: universe in which the chatbot operates. It specifies the intents, entities, slots, responses, forms, and actions the bot should knows about. It also defines a configuration for conversation sessions.
  • endpoint.yml: contains the different endpoints the bot can use
  • markers.yml: conditions used to describe and mark points of interest in dialogues.
  • rasa_knowledge_base.json: knowledge based to be used by our chatbot

Used technologies

Libraries, frameworks, engines, tools, third-party services...

Component	Description	Version
rasa	Open source framework for NLU, dialogue, and integrations.	>=3.0.0
fuzzywuzzy	uses Levenshtein Distance to calculate the differences between sequences in a simple-to-use package	0.2.0
NLTK	provides easy-to-use interfaces to over 50 corpora and lexical resources	3.6.6
re	Python regex library
datefinder	Finding and reformatting input dates

How to install

1. Install Rasa [https://rasa.com/docs/rasa/installation/]
2. Clone project
3. Enjoy!

How to use

1. Enter rasa train to train new model
2. Enter rasa run to interact with chatbot externally
3. Enjoy!

How to use (from CLI)

1. Enter rasa train to train new model
2. Enter rasa shell to interact with chatbot from CLI
3. Enjoy!

How to use alongside chatbot_client

1. Enter rasa run --enable-api --cors "*"
2. Run chatbot client
3. Enjoy!

Notes for developers

Check this for more information on rasa:

• https://github.com/RasaHQ/rasa/blob/main/docs/docs/command-line-interface.mdx

License

Free use of this software is granted under the terms of the GNU General Public License v3.0: https://www.gnu.org/licenses/gpl.html