/310-Group-Project

Primary LanguagePython

310-Group-Project

Project Description

Our project is going to be an interactive chatbot that takes on the role of a would-be Atlantis explorer. The chatbot will only talk about things regarding Atlantis and will excitedly talk to the user about his upcoming adventure. The link to our GitHub repo is as follows https://github.com/edouarde1/310-Group-Project.

How to Run the Chat Bot

  1. Download our GitHub repo
  2. Open the repo using VSCode or Terminal
  3. run run_me_first.sh (If you are uncomfortable running a bash script refer to dependencies.)
  4. Run gui.py
  5. Insert google API key in api_key = "YOUR API KEY HERE" in api_scrpits.py

Dependencies

This bot requires nltk, pyspechecker, and stanza in order to run properly. If you do not run first-time-setup.sh you will have to do the following

pip install nltk
pip install pyspellchecker
pip install stanza
pip install googlemaps
pip install geocoder

Then you have to run py_lib_install.py so python can install the additional libraries. Make sure you install all of nltk.

Dataset

This bot pulls from a json file with specifically designed responses for who, what, where, when, why style responses. Essentially, this bot smartly translates user text into queries to generate aproprait responses.

Classes and Functions

aBot.py

This directory includes all the files for the bot.

get_response

This function recieves the keyword dictionary, asks for user input, and returns chat bot responses. User input is processed using get_query_objects(), which extracts nouns and proper nouns. A for-loop iterates through each processed noun in a list and detects if the word exists in entity_dict.json. If there is a match, that means there is a chat bot response for the keyword. If there is no keyword detected in the user response, then the bot returns "Sorry can't help provide any information that relates to [whatever related noun the user entered]".

Parameters:

  • response: a string input by the user acting as the key for keywords

Returns:

  • output: a string containing the bots response

spell_check

Takes a string and returns a string with closest related permutation that is part of the english language.

Parameter:

  • input: a string input

Returns:

  • correct: a corrected string output

gui.py

BotGUI(tk.Tk)

The class which handles window switching and displays the alternate screens

HomeScreen (class)

The main title screen that shows when loading the application.

ChatSreen (class)

The class that manages the chatscreen.

retrieve_user_message(inputField)

gets user input, and then passes the input to the bot for processing. Recieves the updated bot response, and updates the GUI with the new response.

Parameters:

  • inputField: a string which determins what field to be inputed

show_help_popup

Just passes a string along to the responseLablel in order to display the basic welcome prompt

data_load.py

For managing data and searching

data_load

Loads text from a file.

Parameters:

  • filepath: file to path to retrieve text from

Returns:

  • contents: string of raw text

preproc

Extracts the nouns and proper pouns from the user query. Takes a user query and runs string through Stanza's Dependency Parser. More information about this library is found here: https://stanfordnlp.github.io/stanza/depparse.html

Parameters:

  • query: a string

Returns:

  • obj_list: a list nouns and proper nouns from the user query

dependencyParser

Helper utility used to extract the depencies from a sentence. Runs the dp pipeline object in order to run depparse, lemma, and pos tagginig.

Parameters:

  • sentence: a string

syn_detection.py

detect_syn

This function takes a word and uses wordnet in order to return a list of asociated synonyms with our word,.

Parameters

  • input : word to get inputs from

Returns

  • synonyms: list of synonyms associated with said words.

search_json.py

search_noun_quest

takes a noun and searches for related question style keywords.

Parameters

  • noun, question: noun and question. Question is mostly used as an adverb in this case.

Returns

  • data[noun][quest]: returns a string associated with indexed value

get_nouns

returns a list of all nouns within json file

Returns

  • nounList: list of nouns

get_verbs

Parameters

  • noun: a string

Returns

  • questList: list of adverbs associated with said noun

unitTest.py FIX THIS

test_spell_check

Tests functionality of spell_check() function in bot.py

test_get_response

Tests functionality of get_response(query) function in bot.py

test_load_data

Tests functionality of load_data()

api_scripts.py

wiki_response

takes a string and uses wikipedia API to search and retrieve an excerpt if found on wikipedia

Parameters

  • subject: a string

Returns:

  • text : a string containing 150 characters of information, or a string returning a not-found style message to the user.

get_latling

returns a string of latitude and longitude associated with users IP address. It won't work as intended if you're using a VPN.

get_address

uses get_latling method and performs a reverse geo lookup in order to get the name of the address associated with your IP address.

get_directions

uses get_address to provide a starting address and then takes user input. It then makes a request to places API. If failure it returns no suitable directions found, if success it returns a list of instructions to navigate to the following location.

Parameters:

  • destination: a string of where the users wishes to go

Returns:

  • results: a list of directions, or a string indicating no suitable directions where found.

check_if_directions

Uses a naive aproach to check a string is meant to be a list of directions.

Parameters

  • query: a string

Returns either true or false.

nnp_extract

Uses fancy pants POS tagging extract an address

Parameters

  • query: a string meant to be extracted from

Returns

  • text: a string containing address

Features

Spell Checking

Spell checking was used in order to improve the overall robustness of our chatbot. Incorrectly spelt words are recognized by our system, and replaced with what is assumed to be the most likley word that the user intended to input. This allows our bot to handle a wide variety of cases that it previously was unable to.

Synonym Detection

Synonym detection was used in our chatbot in order to improve the greeting function. Rather than hardcode a large if statement containing some number of standard greetings, we used synonym detection to recognize when a greeting word was input, in order to then output the bots standard greeting.

POS Tagging, Tokenization & Segmentation

In order to better suit the keyword style bot that we have created, we needed to implement a way of understanding sentences that were input, and extract the necessary information required to generate a query that the bot can use to get a response. To do this, we used POS tagging, tokenization, and segmentation. These systems work together to process questions asked, and provide us with a query in the form of <Noun> <Descriptor> which we use to generate much more intricate and conversational dialogue.

APIs Used

Wikipedia

This API was used to provide text exerpts for the bot. More details can be found in get_response in api_scripts.py

Google Geocode

This API was used to provide an address after an IP address's longitude and latitude was found. More details can be found in get_address in api_scripts.py

Google Places

This API was used to generate directions to destinations. More details can be found in get_directions in api_scripts.py