We are going to fix bad chatbots!
Created by Kaylee Chan, Emily Wan, Sam Weiss, Helia Sajjadian, and William Yao, along with the help of the University of Toronto's Technology Leadership Initiative professors and teaching assistants.
Diagnostic tool for chatbot transcripts that help retailers identify poor or strong questions and intents through an easily digestible and intuitive display.
We aim to:
- Reduce time spent on chatbot optimization
- Reduce bad chatbot intents
- Display an average of 30% of diagnostic problems in any transcript's flow
- Optimize the flow of at least three intents in x transcripts
Each transcript will be displayed with a question and the responses it leads to. Each response also has a percentage displayed to show the proportions of leading to a specific branch. Our helpful star tool⭐ will also let users tag which questions they think are important to add to a chatbot. They can also use the flag tool🚩 to tag questions that are not helpful and need to be rewritten or deleted.
Want to test out our tool but don't have any chatbot transcripts?
We have a series of example transcripts for different types of stores that can be found in the examples folder.
You can download any of the examples and follow the installation and instructions to run.
React, Axios, Node, Tailwind, MongoDB, Express, and more :D Licensed under the Apache-2.0 license.
🌐 You can access our frontend by going to our hosted site: https://bestflow.netlify.app/
Download and unzip our code in any folder of your choice, then open the unzipped folder in any JavaScript editor of your choice. In the project directory, you must open two terminals.
Add a .env file in the back-end folder, with:
DATABASE_URI=[Private key, please contact one of us if you want to test this on your local build, otherwise use our hosted version]
TESTDATA_NS=Transcripts
API_KEY=[Private key that we do not currently use. Same as DATABASE_URI, message us if you want to use one]
PORT=5000
In this terminal, run cd back-end to navigate to our backend folder, npm install to stay up-to-date,
and then run nodemon server where our backend will start listening on port 5000.
Add a .env file in the front-end folder, with:
REACT_APP_BASE_URL=http://localhost:5000
In this terminal, run cd front-end to navigate to our frontend folder, npm install to stay up-to-date,
and then run npm start where our react app will run in http://localhost:3000/ in your browser.
On your browser, you will be directed to the main page. Please upload a transcript which will be stored in our database. If a valid project id already exists in our database, it can be accessed by typing the id in the available input box. It is also possible to override the existing transcript with a new one. We have a template transcript available to be downloaded, which shows the minimum required fields for each transcript.
Once a transcript is uploaded, you can click on View Analysis which will take you to the analysis page. On the page, all the intents will be displayed along with previously starred and flagged intents. Our intent menu also allows for users to search by keywords or filter the questions by flags or stars to easily find their important questions.
To load a new transcript, go back to the main page by pressing on the logo in the menu and upload another transcript.
Open your command prompt and navigate to the folder you downloaded our repository in.
Navigate to our backend folder by running cd back-end and then run the tests using npm run test.
Then navigate to our frontend folder by running cd.., then cd front-end. You can run the tests using the same command npm run test.
See icon_bibliography.md in front-end/src/components/icons for more details.
Link: https://www.svgrepo.com/svg/127033/magnifying-glass
Link: https://www.iconfinder.com/icons/211646/left_chevron_icon
Link: https://www.iconfinder.com/icons/211647/right_chevron_icon
Link: https://www.iconfinder.com/icons/211820/skipbackward_icon
Link: https://www.iconfinder.com/icons/211822/skipforward_icon
Link: https://uxwing.com/upload-note-icon/