qna-builder 0.1.3

Similarity-based conversational dialog engine for Python.

QnA Builder

Introduction

QnA Builder is a simple, no-code way to build chatbots in Python. It provides a similarity-based conversational dialog engine, QnA Bot, which makes it easy to generate automated responses to input questions according to a set of known conversations, i.e., question-answer pairs, stored in a knowledge base. QnA Bot relies on a collection of question-answer pairs to generate answers for new inputs.

Install

The easiest way to install the qna-builder is by using pip:

pip install qna-builder

This library is shipped as an all-in-one module implementation with minimalistic dependencies and requirements.

Getting started

A QnA Bot can be set up and used in four simple steps:

1. Import QnABot class

from qnabuilder import QnABot

2. Initialize a bot

bot = QnABot()

3. Fit the bot engine to a knowledge base

bot.fit(kb="knowledge_base.json")

4. Generate answers

bot.answer("Hey. What's up?")

"All good. What's up with you?"

Algorithms

Currently, QnA Bot engine supports the following algorithms for similarity-based answer generation:

• TF-IDF Vectorization ('tfidf')
• Murmurhash3 Vectorization ('murmurhash')
• Count Vectorization ('count')

Supported similarity metrics are as follows:

• Cosine similarity ('cosine')
• Euclidean distance ('euclidean')
• Manhattan distance ('manhattan')

Knowledge base editor

By calling run_editor() method of QnAKnowledgeBase class, the knowledge base editor window will open up in your web browser and allows you to edit your knowledge base by adding, removing, or modifying questions/answers.

from qnabuilder import QnAKnowledgeBase

kb = QnAKnowledgeBase('my_knowledge_base.json')
kb.run_editor()

Here, you can see a screenshot of the knowledge base editor:

Note that you need to install the optional requirement streamlit to be able to use the knowledge base editor.

Tests

To run the tests, install development requirements:

pip install -r requirements_dev.txt

Then, run pytest:

pytest