Chatbot Instructions

User Manual: Pdf

Open the PDF directly: View PDF PDF.
Page Count: 2

DownloadChatbot Instructions
Open PDF In BrowserView PDF
Chatbot assignment
Deadline: March 22nd (Monday)

Goal:
The goal of this project is to write a working chatbot by applying some of the techniques
that we have seen (or will see) thoughout the course. This means that your chatbot should
not only look for predefined keywords and use them to query a database but should rather
implement, at least, one more “intelligent” feature (such as synonym detection using
Word2Vec, text generation, sentiment analysis, formal-informal speech detection, etc.).
You might well not even set up a database.
Your chatbot should be able to:
•

Do some basic chit-chat: greetings, introduction (of abilities) and goodbye.

•

Answer at least two types of domain specific questions. For example, if you chose
to focus on movies, your bot might be able to give a quote from a given actor (or
movie), tell you when an actor was born, or list a director's films.

•

Do something else than just querying a database. Use any NLP technique, such as
Word2Vec, to try and extend your functionality.

I'd like the whole thing to be turned in as a GitHub repo, with a complete README.md
describing what you've done and giving an example dialogue demoing it.
The use of any external API is allowed. For example, you could use an API to retrieve
information about the train services (https://www.ns.nl/reisinformatie/ns-api) or the
weather (https://openweathermap.org/api).

Proposed work plan:
For week 1, I recommend starting by getting a Telegram bot set up, playing with the basic
Markov Chain text generator (markov_norder.py), deciding on your basic domain,
coming up with a list of possible query types, and look for a dataset.
For week 2, you should be implementing at least some basic queries (i.e. mostly
substitution-based), and starting to integrate smarter functions (e.g., similarity-based
substitution for recognition or answering).
In week 3, you should finish your remaining functionality and try to improve usability if
possible: have some people use it and see what modifications are needed to make it more

natural.

Resources
Chatbot ideas and information

pattern matching bots in the NLTK api

basic concepts: retrieval
http://www.nltk.org/api/nltk.chat.html
http://www.wildml.com/2016/04/deep-learning-for-chatbots-part-1-introduction/ vs. generation
http://www.wildml.com/2016/07/deep-learning-for-chatbots-2-retrieval-based-modeltensorflow/
LSTM-based retrieval system

Corpora
http://www.cs.cornell.edu/~cristian/Cornell_Movie-Dialogs_Corpus.html
https://www.reddit.com/r/datasets/comments/3bxlg7/
i_have_every_publicly_available_reddit_comment/
https://machinelearningmastery.com/datasets-natural-language-processing/
Any of the datasets included in Blackboard.
The presidential speeches included here.
Scrapping resources
Need some more data? Web scraping resources1:
https://first-web-scraper.readthedocs.io/en/latest/#act-3-web-scraping
https://doc.scrapy.org/en/latest/intro/tutorial.html
There are now actually good scraping tutorials, libraries (rvest), and text manipulation
libraries (tidytext) for R, too:
https://blog.rstudio.org/2014/11/24/rvest-easy-web-scraping-with-r/
Source Exif Data: 
File Type                       : PDF
File Type Extension             : pdf
MIME Type                       : application/pdf
PDF Version                     : 1.3
Linearized                      : No
Page Count                      : 2
Has XFA                         : No
Title                           : Chatbot instructions copy
Producer                        : Mac OS X 10.12.6 Quartz PDFContext
Creator                         : TextEdit
Create Date                     : 2018:03:08 07:46:09Z
Modify Date                     : 2018:03:12 23:50:26+01:00
EXIF Metadata provided by EXIF.tools

Navigation menu