Diss Presentation

wow

Personalised AI News Digest for Investors Leveraging NLP, Big Data and AWS

Chirag Galaiya | u2209513

Supervisor: Dr. Nadeem Qazi

Overview

Why is this a computing problem?

Literature Review Findings

Objectives

Aim

Proposed Solution

Planning & Methodology

References

Evaluation

Project Aims

What problems does this project aim to solve?

Information Overload

Filtering Noise

Decision Support

Time Constraints

Simplify the large amounts of news articles available across a range of news sources relating to certain stocks by generating a personalised daily digest, reducing the overwhelming volume of information for users.

Mitigate the struggle of searching through irrelevant or repetitive information by incorporating event and entity detection, streamlining the identification of key events impacting the selected stocks.

Provide users with useful insights and relevant news, allowing them to make more informed investment decisions within the dynamic and fast-paced stock market landscape.

Cater for the average personal investor juggling busy schedules with concise news updates, eliminating the need to spend hours each week reading up on articles.

Project Objectives

To research relevant AI-based methodologies which can be applied on news articles and relating data to extract useful insights, along with the required tools such as python libraries or cloud services, which can carry out these analytical functions.

To investigate the most appropriate tools and services to create the architecture of the application, including the cloud services, frontend, backend, data solutions, and any additional resources required.

To determine security measures to enhance security throughout the software application, such as user authentication services/protocols, data encryption measures, and more.

To develop the required scripts to run on the relevant cloud services to automate the data collection, processing, analysis, and storage in real-time, and configure the relevant services within the architecture to produce a working application

To evaluate the developed application, and the effectiveness of the insights provided by the functionality of the app.

Generative AI

Algorithmic & Data Processing

Why is this a Computing problem?

Cloud Computing

This project contains various computing elements which are required to generate content, extract insights, develop the software, and more.

Full-stack Development

Application Security

Literature Review Findings

3.Event and entity detection

1.LLMs

GPT-3 is a preferred LLM when it comes to news summarisations over fine-tuned models

Practical applications through means of python libraries, with abilities to examine vast amounts of data

4.Cloud solution

2.News consumption

Research shows a shift towards personalised news feeds, and that younger generations consume news from various sources and outlets

Research suggests that the AWS platform would be the best suited to such an application offering reliability, affordability and a large range of services suited to this project

Proposed Solution

The proposed solution is to develop a cloud-based backend along with a mobile app to provide the following functionality:

• Provide the user with an interface to search and select multiple stocks
• A script gets triggered daily, which fetches the latest news articles relating to each stock in the user's selection of stocks and stores this data
• This data is then processed to extract useful insights
• The news articles are then passed onto a Generative AI model to summarise the latest events and useful information present across all those articles
• The app would then be updated with this personalised news digest everyday for the user to read and stay up-to-date with the news relating to their selection of stocks
• User authentication and data encryption
• User-friendly and modern GUI
• Chatbot functionality within the app to assist the user with questions relating to the latest news associated with their stock selection.

The evaluation method I will use is survey to find user satisfaction and the effectiveness of the insights provided

Evaluation

Risks

Benefits

• Direct user feedback
• Quantitative and qualitative data
• Structured evaluation
• Post-implementation insights
• Scalability
• Cost effectiveness

• Response bias
• Limited depth
• Low response rates / survey fatigue
• Question design challenges
• Non-representative samples

Progress / Deadlines

• Submit Final Proposal
• Submit Ethical Form

• Submit Literature Review and Project Plan

Dec

May

Oct

• Submit Presentation Slides
• Initial VIVA

• Choose Supervisor
• Submit Initial Proposal

Jan

Nov

• Submit Project Report
• Final VIVA

Project Plan

References

Ofcom. (2022). News Consumption in the UK 2022 Report. Available at: https://www.ofcom.org.uk/__data/assets/pdf_file/0027/241947/News-Consumption-in-the-UK-2022-report.pdf (Accessed: 11/12/2023).

Goyal, T., Li, J.J., Durrett, G. (2022). News Summarization and Evaluation in the Era of GPT-3. arXiv preprint arXiv:2209.12356. Available at: https://arxiv.org/pdf/2209.12356.pdf (Accessed: 11/12/2023).

Balta, A. (2020). Personalised News in the World of Internet: Exploring solutions for a tailored news experience. Available at: https://urn.kb.se/resolve?urn=urn:nbn:se:umu:diva-168397 (Accessed: 11/12/2023).

Vinzens, G.A., 2015. Newspaper 2.0 (Master's thesis, Center of Economic Research at ETH Zurich). Available at: https://doi.org/10.3929/ethz-a-010475954 (Accessed: 11/12/2023).

Weng, J. and Lee, B.-S. (2021) “Event Detection in Twitter”, Proceedings of the International AAAI Conference on Web and Social Media, 5(1), pp. 401-408. doi: 10.1609/icwsm.v5i1.14102.

Liu, S. and Healey, C.G., 2023. Abstractive Summarization of Large Document Collections Using GPT. arXiv preprint arXiv:2310.05690. Available at: https://arxiv.org/pdf/2310.05690.pdf (Accessed: 11/12/2023).