Supervisor: Wei Liu

UTS FEIT Chatbot is a chatbot directed for prospective high school students in search of universities to get into. The chatbot is able to answer user queries regarding courses at UTS including details of the course such as duration, credit points, as well as answering queries regarding the structure of courses and subjects at UTS.

A live instance of the chatbot is live on WebChat

As a bare minimum for chatbots, the bot is able to say hello and good morning, and respond to thank you. In addition, the bot would tell users what it is able to do when asked.

The bot is able to respond with a short description of a course/major/subject when presented with the code or name of the item. The description is followed by an URL to the UTS Handbook.

The chatbot stores the current entity (course) in a slot and can further continue conversations regarding said entity.

The chatbot can retrieve details of the course from the dataset, i.e. credit points, duration, as well as boolean details (is it a combined degree/honours degree or not)

The chatbot is able to retrieve substructures of a structure when asked (what subjects are in a subject, what majors are in a choice block, etc.)

This chatbot utilises Rasa with spaCy for its language processing. Install using pip with:

pip install rasa[spacy]
python -m spacy download en_core_web_md
python -m spacy link en_core_web_md en

To train the bot from the files, use:

rasa train

Then run the bot with:

rasa run

Or to test a conversation with Rasa in the command line, run:

rasa shell

On a separate terminal, run an action server for custom actions with:

rasa run actions

There are a few editable data files to modify the training of the chatbot

There are a few .csv files that contain the structure of courses and subjects at UTS. The dataset consists of recursive relationships of abstracted structures. data/items.csv simply contains the code and name of an entry in the UTS Directory, while data/relations.csv contains the said recursive relations. data/courses.csv is a specialised type of structure for courses since courses have more attributes such as credit points, ATAR, and other details as mentioned above. data/alt_names.csv contains alternate names for courses which is used for searching course names.

data/nlu.md contains training sentences to train the NLU for intents and entities. Below is an example intent:

## intent:intent_name
- sample sentence with [entity](entity_type)

## intent:details
- can you tell me about [c10148](code)

data/stories.md contains story examples which are lists of intents, entities, and responses from the bot, with * denoting user input, and - denoting bot utterance/actions Example below:

## story_name
* intent
- utterance
* intent{"entity_type":"entity"}
- action

## story_details_02
* greet
- utter_greet
* details{"name":"bachelor of science in it"}
- action_details
* thanks
- utter_thanks
* goodbye
- utter_goodbye  

Domain (domain.yml) is described as the ‘universe’ of the chatbot which contains lists of intents, entities, slots, and actions, where slots are saved entities/variables that define the story, while actions are custom responses that are executed after a command from the user. The domain also contain lists of simple utterances (as opposed to custom actions). For example:

utter_greet:
- text: Hi there! I am Stu from UTS!
- text: Hi! My name is Stu!
- text: Hi there! I am Stu. Do you have any questions about FEIT at UTS?
- text: Hey! I am a chatbot from UTS. Ask me what I can do!
- text: Hello! I am Stu and I am from UTS 

endpoints.yml contains endpoints required by the bot (i.e. action server), while credentials.yml are credentials required when running the chatbot on other platforms such as Facebook.

actions.py contains python codes that are run on action calls. Actions are represented as classes which has a function name that returns a string that corresponds to the name on the domain. The method run takes three attributes, a dispatcher, tracker, and domain that are used for the run methods. directory_loader.py is a python script that loads the dataset in an Object-Oriented environment. The script overrides the [] operators for easy access for entries in the dataset. Functions in the file include:

code()                  -> returns full code (e.g. C10219 for 10219)
get_name()              -> returns official item name
get_search_list()       -> returns list of alt. names paired with ID
get_type()              -> returns type (course, major, subject, etc.)
url()                   -> returns url on UTS handbook