This is a simple Chatbot written in Python 3.5 with a MySQL database backend. The code builds on my other SimpleBot demo (https://github.com/edbullen/SimpleBot) and introduces some NLP with Python NLTK and basic Machine Learning capabilities to demonstrate Sentence Classification using the NLTK and scikit-learn. The Stanford CoreNLP package, written in Java, is also used to parse grammar and extract sentence topics, subject, object etc.

The ChatBot conversation operates in 3 modes

Chat (just learned responses from previous exchanges)
Statement (receive a statement of fact and store it away)
Question (receive a question and attempt to answer it based on previously stored statements)

As it currently stands (May 2017), the functionality is far from perfect, but it demonstrates the concepts of Natural Language Processing, Sentence Classification and a very basic level of Natural Language Grammar processing.

This version is still at a basic experimentation level - there is no concept of authentication, security etc.

Python Library Dependencies

pymysql
nltk
numpy
pandas
scipy
scikit-learn
Stanford CoreNLP Parser This is a Java package that needs to be download and located in suitable dir for future ref
Java - tested with Java 8, java version "1.8.0_131"

Files and Components

Core Functionality

chatbot.py - main ChatBot library
utils.py - generic function utilities used by ChatBot (config, DB conn etc)
features.py - library for extracting features from sentences using NLTK
botserver.py - Multi-Threaded server to allow multiple clients to connect to the ChatBot via network sockets
simpleclient.py - Simple network sockets client to connect to botserver

Default botserver.py logging location is

./log/bostserver.log

Setup and Test

./config/config.ini - template config file, requires editing before starting the chatbot for the first time.
pwdutil.py - store an encoded password for connecting to the database schema.
setupDatabase.py - drop and recreate the database tables (existing data gets lost).
pingDB.py - test the database configuration: create test table, insert data, query it, drop the test table.
mlClassGenerateRfModel.py - generate a scikit-learn Random Forest Model for sentence classification based on input CSV. Default output file-name is "RFmodel.ml"

Tools and Utilities

All Dump and Load utilities use the ./dump subdirectory by default.

dataDump.py - Dump out a database table in CSV format
dataLoad.py - Load a database table in CSV format
featuresDump.py - read in a CSV of sentences, dump out features using features.py into a CSV
testClassifyModel.py - test the Sentence Classification logic and Model for a given sentence (using the pre-built Random Forest RFmodel.ml model)
testGetAnswer.py - test retreiving an Answer from the chatbot for given sentence
testGetGrammar.py - extract grammar structure for given sentence (using Stanford CoreNLP package)
testStoreStatement.py - parse and store a given Statement sentence in the database.

Install and Setup

Details of installing dependancies for the NLPBot to function are documented in the python_server_config.md note in this repo.

In summary, for a new install, the following steps are required:

Install Python 3.5
Install pyMySQL
Install NLTK
Install Machine Learning Libs
Create the Linux BotUser
Install GIT
Install Java 8
MySQL Database Server Configuration
Install Bot Code and Configure
Install the Stanford CoreNLP Package
Configure the botuser ./config/config.ini file
Start BotServer

These steps are documented in detail here

Sample config.ini Configuration File Contents

[MySQL]
server: localhost
dbuser: nlpbot
dbname: nlpbot
dbcharset: utf8mb4

[Server]
listen_host: 0.0.0.0
tcp_socket: 9999
listen_queue: 10

[Java]  #required for Stanford CoreNLP
bin: /usr/bin/java

[StanfordNLP]
corejar: /home/botuser/StanfordParser/coreNLP/stanford-corenlp-3.7.0.jar
modelsjar: /home/botuser/StanfordParser/coreNLP/stanford-corenlp-3.7.0-models.jar

[DEBUG]
assoc: False
weight: False
itemid: False
match: False
server: False
answer: False

Starting the BotServer

The chatbot can be started with a multi-threaded server scheduler (botserver.py) that listens for connections on a TCP port. This is a very simple "bare-bones" multi session framework with no authentication and just relying on TCP sockets for connection.

Remote TCP Socket Connection requests are given a thread and their own session connection.

The botserver gives each session a connection to the shared database server.

nohup python3.5 botserver.py &

Logging

Bot server output is logged to

~/NLPBot/log/botserver.log

Stopping the Server

$ ps -ef | grep botserver
botuser  **22915** 22854  0 08:15 pts/0    00:00:01 python3.5 botserver.py
botuser  23165 22854  0 08:54 pts/0    00:00:00 grep botserver
kill -9 22915

Local Client Connect

python3.5 simpleclient.py -a localhost -p 9999

Remote Client Connect

Make sure the botserver port is allowed through the firewall.

python3.5 simpleclient.py -a 192.168.10.101 -p 9999

Sample Conversation

This sample conversation shows the capabilities of the NLPBot ChatBot as of 1 May 2017. There are some good examples of limitations which are most noticeable around the very basic Question mode (looking for a previously learnt Statement). It does however generally determine whether it is "just chatting", being given a Statement to store or being asked a question.

$ python simpleclient.py -a 192.168.10.1 -p 1234
Connecting to server 192.168.10.1  at port  1234
>>> Hi
Bot> hello
>>> The latest Python source distribution is always available from python.org at https://www.python.org/downloads/
Bot> Shall I store that as a fact for future reference?  ("yes" to store)
>>> yes
Thanks, I've made a note of that.
>>> There is a Python newsgroup comp.lang.python and a mailing list: python-list                          .
Bot> Shall I store that as a fact for future reference?  ("yes" to store)
>>> yes
Bot> OK, I've stored that information.
>>> Where can I get the latest Python source code from?
Bot> Sorry, I can't find an answer to that.
>>> Where is the latest Python source distribution?
Bot> The latest Python source distribution is always available from python.org at https://www.python.org/downloads/
>>> Thanks.
Bot> How are you?
>>> I am fine thanks
Bot> How are you?
>>> Good
Bot> Please train me - enter a response for me to learn (or "skip" to skip)'
>>> skip
Bot> OK, moving on...
>>> Is there a Python newsgroup or mailing list?
Bot> There is a Python newsgroup comp.lang.python and a mailing list: python-list.
>>> quit
Connection closed

PyData 2017 Slides

GitHub Preview Viewer for HTML Slides