NOVABOT, THE ROBOT ASTRONOMER AND GEOGRAPHER (Individual Extension)

---

Project Description

Nova is a chatbot that is passionate about teaching astronomy and geography to users. Her implementation of natural language processing in python allows her to understand and answer user questions related to astronomy such as stars, galaxies, black holes and planets, and questions related to geography such as country capitals and geographical records. She was created with the intent to spread basic knowledge about astronomy and geography to individuals who are interested in these fields.

Software Development Life Cycle

We chose to use the incremental development life cycle. This life cycle is ideal because we would like to continuously gather user feedback and make corrective changes throughout development. As part of the agile process, we will also be participating in peer programming since we all have limited experience creating chatbots and want to support each other. Our incremental life cycle will include a project planning phase, a researching phase, a development phase, a review phase and then a presentation phase. We aim to develop a basic chatbot in the first major increment then develop a GUI for the chatbot in the second major increment.

How does Nova work?

Nova was developed in python and uses natural language processing to understand and converse with the user. She takes the input from the user and cleans the sentences into a condensed and easily-readable format. As part of the cleaning, Nova checks if the users question is in English and if not, translates it into English so she can understand it. She then compares the input to the questions she has in her questionbank and gives a response based on the similarity between the input and questionbank questions. All the questions and responses are found inside the corpus.txt file and may be expanded or changed as more astronomy or geography questions are thought of. Nova also has the ability to query wikipedia when she doesn't understand a question and directs the user to the relevant wikipedia page for more information. Once Nova chooses a proper response she checks whcih language the user asked the question in and translaters her question into the target language.

Requirements

You will need to download and install NLTK and Spacy in your python IDE before using Nova. You will also need to download en_core_web_lg for Spacy. Once that is complete, you can run main.py and begin using Nova.

• pip install nltk
• pip install -U spacy
• python -m spacy download en_core_web_lg
• pip install pyspellchecker
• pip install beautifulsoup4
• Sadly the translation functions aren't available on other machines as the API key used is stored in a local .json file which can't be published online for security reasons

Sections

• Corpus: Contains a compilation of questions and responses that Nova uses to converse with the user
• Preprocessor: This section is in charge of formatting the user input into a more readable format for the system
• Processor: This section takes the preprocessed data and tries to match it with its accurate response
• Main: This section introduces the user, takes their input, generates the GUI, and manages how the program executes
• norm_punc: This section was taken from the Phrasal library. It normalizes sentences by removing elements such as special characters, extra spaces and apostrophes.
• translation: This section handles requests to translate text into a target language or to identify the language of a query
• wikipedia: This section handles requests sent to wikipedia for information Nova doesn't know

Corpus --> Preprocessor --> Processor --> Main

Capabilities

• Nova utilizes natural language processing and pattern matching effectively so the user input does not have to match the predefined questions exactly to get an accurate response
  • Input: what is nuclear fusion?
  • Nova: It is a processes by which the Sun fuses hydrogen atoms to form helium
  • Input: nuclear fusion
  • Nova: It is a processes by which the Sun fuses hydrogen atoms to form helium
  • `Input: what is the capital of Canada?'
  • Nova: Ottawa
• Nova cover a wide range of topics in astronomy, astrophysics and geography
• Nova is easy to refactor and reuse since its structure is very basic. The corpus.txt can be modified to suit any topic of interest
• Nova can recognize inputs even if they have simple spelling errors
  • Input: what s the capitall pf jamaica?
  • Nova: Kingston
• Nova can recognize user inputs even if they contain special characters
  • Input: what i$ the c@pital of India?
  • Nova: New Delhi
• Nova can recognize the parts of speech of words in the user input
  • Input: what is a white dwarf?
  • Nova: Parts of Speech: [('white', 'JJ'), ('dwarf', 'NN')]
• Nova recognizes synonyms of the word "bye" and uses them to find out when the user is done asking questions
• Nova speaks every language that Google translate understands, just speak your native language to her and she'll respond in kind
• Nova can look up information on Wikipedia when she doesn't understnad a question and provide the user with a short snippet and a link to the full article

Limitations

• Nova can take a while to process user input and output the correct responses since there is a lot of conversion and formatting to be done. The processing time could be lowered by using more efficient functions.
• Nova is somewhat limited to the predefined questions and responses in the corpus.txt. The corpus.txt will have to be expanded to allow Nova to answer more questions.
  • Input: what is a megabyte?
  • Nova: I do not understand your question
• Only English language Wikipedia is supported, so queries Nova doesn't understand in any language other than English will return the query from English Wikipedia

New Features

• Added a GUI to make the NovaBot more user friendly. This is an improvement because the user had to previously view all the logs that were displayed on the console.
• Added a new topic. Geography was chosen as the new topic since it seemed like the closest topic to astronomy. The addition of the new topic opens up a lot more conversations between the user and NovaBot.
• Added a feature that gives five different responses if NovaBot does not recognize the question. This feature makes NovaBot seem more realistic since she no longer gives the same answer to every unrecognized question
• Added a feature to correct spelling. This was a very important addition since previously, NovaBot would fail to recognize user input even if it had simple spelling errors. Now the user will not have to worry about minor spelling errors since NovaBot will correct spelling automatically.
• Added synonym recognition. NovaBot now will close whenever a user inputs synonyms of the word "bye". Previously, the user could only terminate the program is they typed the word "goodbye" so this addition allows them more flexibility.
• Added POS-tagging. This feature was to be used to extract words that seemed unnecessary during the preprocessing. However, the feature was not used since the alternative NLP tools that we implemented already took care of those words.
• Added a function from Phrasal that normalized sentences. This feature allows NovaBot to recognize user input even if it contains special characters, apostrophes or extra spaces. It is just another feature to help NovaBot recognize more user inputs.
• Added the capability to understand every language that Goole Translate supports and respond in that language
• Added the ability to look up information on Wikipedia to support Nova's knowledge base