alexmotoc / projectaeora

🏆 A virtual assistant to assist traders with financial queries.

Aeora

The ChatBot for traders, providing answers to questions on the FTSE 100.
Built for quick and intellectual conversations.

Installation

Install all the dependencies using the command:

pip3 install -r requirements.txt

This is a Django project therefore the command you need to run it is:

python3 manage.py runserver

projectaeora's Issues

Parse data obtained from the scraper to JSON objects to be passed to the client

Jsonify the relevant data from the response into appropriate objects that will be used in the UI visualisations (cards etc.)

For example:

{
    "speech": "Here are the latest news.",
    "text": [
        {
            "headline": "News article title",
            "impact": "+0.7%",
            "image": "url",
            "summary": "contents...", 
            "redirect": "article_url"
        }
    ]
}

Fix company card price aligning issues

Get rid of tabs and make elements float to the right hand-side of the card.

Add default responses for invalid queries

Define a set of possible replies telling the user what went wrong and suggesting alternative questions.

Compute summary, tags and possibly sentiment analysis for news articles

Create user interface for daily briefing settings

Look in models to see the fields which are required - link with the form in forms.py.

Test UserPreferences model and form

Feature: get stock data based on company's code

Scrape the profile page of the company rather than storing the whole table into a list and searching it.

Add how old the news should be in the preferences form

Create scrollable interactive bubbles

Each interactive bubble should be able to contain multiple elements (e.g. cards or suggestions). When the combined width of these elements exceeds the one of the browser window, the container should be scrollable on the x-axis (with hidden scrollbar obviously).

Fix card information overflowing

Create database model for user preferences

Define the customisable daily briefing fields that can be adjusted from the settings menu.

Define intent and trigger phrases for daily briefings

e.g. "Good morning", "Tell me about my day", "What's on today?" etc.

Asking 'Are pharmaceuticals rising or falling' throws errors and results in infinite spinner.

The error that appears in the console is:

Traceback (most recent call last):
File "/usr/local/lib/python3.5/dist-packages/django/core/handlers/exception.py", line 35, in inner
response = get_response(request)
File "/usr/local/lib/python3.5/dist-packages/django/core/handlers/base.py", line 128, in _get_response
response = self.process_exception_by_middleware(e, request)
File "/usr/local/lib/python3.5/dist-packages/django/core/handlers/base.py", line 126, in _get_response
response = wrapped_callback(request, *callback_args, **callback_kwargs)
File "/media/mattleitzell/Windows/Users/Matt/Documents/CompSci/Term 2/CS261 - Software Engineering/Trader-ChatBot/webapp/bot/views.py", line 75, in chat
if response['type'] not in ('input.unknown', 'incomplete', 'simple.response'):
TypeError: string indices must be integers

Implement functionality for greetings from Aeora when the user says something like 'hello'

Create intent for sector related questions

Train model to accept queries regarding particular sectors or subsectors from FTSE 100.

Handle case when question asks for top performers (both risers and fallers)

Define entities to train model

Create a json file with the FTSE 100 Company codes and their possible synonyms (names, common abbreviations etc.)

Fix scrapper typo

Add tags to news stories

Reference:

http://materializecss.com/chips.html

Implement the My Interests page as a news feed

We could use some sort of frequency analysis to rank the entities based on their number of searches and then display some useful news or company/sector data in the My Interests page - similar to how a news feed would work. Could also include the entities from user preferences.

Would act like the left sidebar from the original mock-up but can contain more and better spread-out information.

Asking about percentage difference doesn't return the two related attributes in the card and the text is black

Display message timestamps

Could have a centered divider with the date and also display the time under individual messages as a toggle feature while clicking on them.

Fix get_top5() method from scraper

The method should return a list of tuples ideally or a JSON (it would make it easier to work with instead of having a string with values separated by tabs/new lines).

When questions about 'percentage change', the percentage that shows up is black (should be red/green)

Asking 'How are banks performing?' returns information about just Lloyds.

Add voice input & output

Fix slow scrolling while loading history

Resize images related to news articles

While it is difficult to maintain the same size for news cards (due to different image sizes and summary lengths), they should be pretty similar and sometimes images can be too big which is why they need shrinking.

Key point: maintain aspect ration!

Fix empty query submission through AJAX

Implement different colour schemes

Could add another field to user preferences with 2-3 different colour schemes that would change the chat bubbles, suggestion chips and sidebar.

Fix bug where asking about the revenue for a particular year returns the table with all years

Fix bugs identified in scraper/test/test_scraper.py

Run the tests to understand the bugs, most are to do with not returning None when it should be returned, or failing to return when the input has been identified as None and instead trying to iterate over NoneTypes...

Ajaxify web app

Use Ajax when sending queries through POST requests and displaying their corresponding replies. Add Ajax to settings page to give feedback to user when changes have been saved as well as have autocomplete field for the companies and sectors that are being tracked.

Create choices for the UserPreferences model

I.e for colour_scheme we might have "DEFAULT", "DARK", "LIGHT" etc..., for voice_or_text it could be changed from boolean to charField and then the choices could be "Voice" and "Text".

https://docs.djangoproject.com/en/2.0/ref/models/fields/#choices

Define contexts in the Dialogflow agent

Make the bot more conversational by adding contexts:
https://dialogflow.com/docs/contexts

Fix problem where no response is displayed if an unknown company is asked about.

Timestamps missing from ajax inserted questions

Scrape financial news from LSE and potentially FT

Financial data successfully scraped from London Stock Exchange.

Display top risers and fallers

• Define new intent to detect such queries.
• Create a card table visualisation for the feature mentioned above.

Add value as synonym for entities

If we don't include it, the value of the entity won't be returned.

Create landing page for chatbot

Modify Top Risers/Fallers to only display FTSE 100 companies

Scrape the time of the collected data

We are displaying the time and date when the stock price had a particular value in our cards but we are not scraping it.

http://www.londonstockexchange.com/exchange/prices-and-markets/stocks/summary/company-summary/GB00B5ZN1N88GBGBXSET1.html

Under the graph it displays the timing of the most recent data:
As at 17-Feb-2018 22:59:35 - All data delayed by at least 15 minutes.

Add unit tests for financial news data

Fix "What about revenue" throwing errors in console

Fix Dialogflow contexts for time periods and what is being done if there is no time period specified.

Add help modal trigger on all the pages

Create UI for user preferences

-Voice or Text
-Colour Scheme

connect to UserPreferences form/model

Create user preferences form

Define the form associated with the model from #17

Asking 'How is the mining sector performing?' yields infinite spinner and console errors.

Traceback (most recent call last):
File "/usr/local/lib/python3.5/dist-packages/django/core/handlers/exception.py", line 35, in inner
response = get_response(request)
File "/usr/local/lib/python3.5/dist-packages/django/core/handlers/base.py", line 128, in _get_response
response = self.process_exception_by_middleware(e, request)
File "/usr/local/lib/python3.5/dist-packages/django/core/handlers/base.py", line 126, in _get_response
response = wrapped_callback(request, *callback_args, **callback_kwargs)
File "/media/mattleitzell/Windows/Users/Matt/Documents/CompSci/Term 2/CS261 - Software Engineering/Trader-ChatBot/webapp/bot/views.py", line 75, in chat
if response['type'] not in ('input.unknown', 'incomplete', 'simple.response'):
TypeError: string indices must be integers

Add help modal pop-up for the sidebar button

Add instructions detailing how the product should be used and the common queries it can reply to. Could potentially make use of feature discovery although all the UI elements are pretty straightforward.

References:

• http://materializecss.com/modals.html
• http://materializecss.com/feature-discovery.html