Scribd
Upload
29 views8 pages

Team Repurpost - Prototype (October 28)

The prototype provides the status of project deliverables for an AI assistant being developed by Team Repurpost. Several deliverables are completed, including diagrams, wireframes, and initial data cleaning. Development of the machine learning model and API is in progress. Risks around data size, model accuracy, and API deployment are being addressed.

Uploaded by

Shenba
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
29 views8 pages

Team Repurpost - Prototype (October 28)

The prototype provides the status of project deliverables for an AI assistant being developed by Team Repurpost. Several deliverables are completed, including diagrams, wireframes, and initial data cleaning. Development of the machine learning model and API is in progress. Risks around data size, model accuracy, and API deployment are being addressed.

Uploaded by

Shenba
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 8

PROJECT REPURPOST

Prototype

Abstract
The prototype document will provide status of the deliverables and snapshots of work-in-
progress deliverables and report any potential risks.

Yashwanth Balan Arumugam


Pragya Avinash Mishra
Shruti Sham Kotwal
Atharva Shantanu Kulkarni
Saju Chacko Rajan
Table of Contents
Project Deliverables .............................................................................................................................................. 2
Artifacts – Work Completed ................................................................................................................................. 2
Solution deployment and workflow diagram ................................................................................................... 2
User Flow Diagram ............................................................................................................................................ 3
Wireframe for end state goal for tag suggestions ............................................................................................ 3
Artifacts – Work in Progress ................................................................................................................................. 5
Working ML Model for tag suggestions ............................................................................................................ 5
GitLab repository of model code ...................................................................................................................... 5
Risks – Emerging & Retired ................................................................................................................................... 6
Works Cited........................................................................................................................................................... 7
Repurpost Prototype Document
Project Deliverables
Following are the main project deliverables and its statuses:

Si# Deliverable Status & Details


1 Solution deployment and workflow diagram Completed
2 User Flow diagram Completed
3 Wireframe for end state goal for tag suggestions Completed
4 Working ML Model for tag suggestions In Progress
5 API endpoint to invoke ML model Not Started
6 GitLab repository of model code In Progress
7 Provide Web API usage guide and documentation Not Started

Artifacts – Work Completed


Solution deployment and workflow diagram
The solution deployment and workflow diagram display the major deployment components that
are required for Repurpost to use the NLP model. For the purpose of this project, the focus is
more on creating the web API NLP model is the main business logic that supplies the tag
suggestions to the Repurpost platform UI.
User Flow Diagram
The project team has performed some research on the Repurpost platform and has been able to
identify the user flow that the user will go through for getting the content tag suggestions while
they focus on content creation.

Wireframe for end state goal for tag suggestions


The end state goal for Repurpost is documented in these wireframes. Each wireframe screen is
focused on using the existing
Repurpost platform
architecture and style and
utilizing the available screen real
estate on existing screen to
avoid any major screen changes
for users already using the
platform.

Screen 1
The landing page of the
Repurpost Dashboard. The user
has 4 different textual content
creation options.
Screen 2
The 4 different textual
content creation screens
have most the elements
that are common to each
other. The main
commonality is the
content title and context
text sections which will be
the inputs for model
suggestions. The tags
section is accessible
through a button on the
lower section of the page.

Screen 3
The tag screen will be
shown when user clicks
the ‘Tag’ button. The
content title and text on
the content screen will be
provided as inputs to the
web API and tag
suggestions will be
displayed on the screen
for user to select, if the
user chooses to after
checking the related
content.
Screen 4
The user has an option to
choose any or all of the
suggested tags after
reviewing related
content details based on
the title and text of the
content that user
provided on the main
content creation page.

Artifacts – Work in Progress


Working ML Model for tag suggestions
As per the project plan, the dataset identification and cleanup steps are in progress. The python
notebook used for dataset load and cleanup steps is provided in the embedded document
provided below. The dataset load and join steps have been completed. The cleanup steps are in
progress and details are as follows

1. Remove English stop words – Complete


2. Remove HTML tags – Complete
3. Convert to lowercase – Complete
4. Remove punctuations and special characters – In Progress
5. Lemmatize words – Pending

StackOverFlow_Ra
w_Dataset_Cleanup - Jupyter Notebook.pdf

GitLab repository of model code


The team plans to utilize GitLab for storing the project code and dataset that will be used for this
project. Currently, only the datasets have been checked-in to the repository. The python
notebook and documentation will be added as soon as we have the artifacts ready for check-in.
The GitLab repository is accessible from Repurpost Auto Tag Suggestions.

Risks – Emerging & Retired


Following are the main risks that are identified in the project along with the plan to address each
of them:

Identified Risks Details and Risk Mitigation Plan Status


Dataset Size Processing and preparing the 1.75 GB dataset with 1.25 million rows Retired
is causing slowness in cleaning up the dataset.
As a mitigation, using parallel processing packages like Pandarallel
and Dask is allowing us to speed up function application on each
row.

The main limitations is the number of CPU cores that are available
on the machines.

Model Accuracy Unlike traditional machine learning model accuracy, accuracy, Emerging
precision, recall and f1 score is not the right metrics for evaluating
the performance of the trained models.

Since the problem statement requires a multi-label classification


model the models are going to be evaluated using the Hamming
loss factor and Jaccard similarity index.

Web API Other than using Flask for creating an API in local machines, the Emerging
Deployment team has not explored any other cloud API deployment yet.

Options are being explored and render.com is an option that is


being considered to deploy the NLP model as a public API for demo
and testing purposes.
Works Cited
Artifacts Reference Documents
Solution deployment and workflow diagram
Repurpose high
level solution diagram.pdf

User Flow diagram


Repurpost user
flows.pdf

Wireframe for end state goal for tag


suggestions
Repurpost
proposed wireframe.pdf

Working ML Model for tag suggestions


StackOverFlow_Ra
w_Dataset_Cleanup - Jupyter Notebook.pdf

Updated Project Plan (28th October 2022)


Latest Project Plan -
Team Repurpost.docx

You might also like