---
tags: PhD
---

graph LR

raw_text[Raw Text Data] --> preprocessing[Preprocessing]
preprocessing --> model_training[Model Training]
model_training --> output_generation[Output Generation]
output_generation --> final_model[Trained LLM]

# PhD Structure

## Top-level outline

graph LR

A[PhD Dissertation] --> B[Ch 1: Introduction]
A --> C[Ch 2: Literature Review]
A --> D[Ch 3: Research Methodology]
A --> E[Ch 4: Results]
A --> F[Ch 5: Case Studies]
A --> G[Ch 6: Conclusion]
A --> H[Ch 7: Appendices]

### Chapter 1: Introduction

graph LR

A[Introduction] --> B[Background and Motivation]
A --> C[Research Problem]
A --> D[Research Approach: Questions and Objectives]
A --> E[Scope and Delimitations]
A --> F[Proposed Methodology]
A --> G[Dissertation Structure]
A --> H[Conclusion]

B --> BA[The importance of scientific communication]
B --> BB[Challenges in processing and understanding large volumes of data]
B --> BC[Role of MDS in mitigating these challenges]

C --> CA[Defining the problem]
C --> CB[Challenges and limitations of existing methods]
C --> CC[Significance of addressing this research problem]

D --> DA[Developing efficient MDS framework]
D --> DB[Investigating the use of advanced NLP techniques]
D --> DC[Evaluating framework performance]
D --> DD[Potential applications/implications of the research]
D --> DE[Research Questions and Objectives]

E --> EA[Justifying focus on scientific papers]
E --> EB[Limitations of research context/data/methodologies]
E --> EC[Assumptions and biases]

F --> FA[Introduction to main proposed methodolgies]
F --> FB[Rationale for these methodologies]
F --> FC[Ethical considerations and risks]

G --> GA[Overview of chapters and contents]
G --> GB[Navigation guidance]

H --> HA[Recap of problem/objectives/significance]
H --> HB[Expected contributions and impact/ Contrib to knowledge]

### Chapter 2: Literature Review

graph LR

A[Literature Review] --> B[General Approaches to MDS]
A --> C[Transformer-based Approaches to MDS]
A --> D[Hybrid Approaches to MDS]
A --> E[LLM approaches to MDS]

B --> BA[Extractive vs. Abstractive Summarisation]
B --> BB[Single-Document Summarisation vs. MDS]
B --> BC[Evaluation Metrics for MDS]
B --> BD[Choosing Datasets for MDS]

C --> CA[Introduction to Transformers]
C --> CB[BART-Bidirectional and Auto-Regressive Transformers]
C --> CC[BERT-Bidirectional Encoder Representations from Transformers]
C --> CD[GPT-Generative Pre-trained Transformer]
C --> CE[XLNet-eXtreme Multi-task Learning via Adversarial Training of a Transformer Network]

D --> DA[Introduction to Hybrid Approaches]
D --> DB[Combinatorial Approaches]
D --> DC[Iterative Approaches]
D --> DD[Reinforcement Learning-based Approaches]
D --> DE[Transformer-based Hybrid Approaches]

E --> EA[Introduction to LLMs]
E --> EB[Pre-training Strategies for LLMs]
E --> EC[Fine-tuning Strategies for LLMs]
E --> ED[Ethical and Societal Implications of LLMs]

### Updated section headings

1. **Introduction to Literature Review**
- Brief overview of the chapter
- Explanation of the relevance of the reviewed literature to your research topic

2. **Background on Document Summarization**
- Overview and history of document summarisation
- The importance of document summarisation in various fields
- Introduction to single document vs. multi-document summarisation

3. **Extractive vs. Abstractive Summarisation**
- Detailed description and comparison of extractive and abstractive summarisation techniques
- Advantages and disadvantages of each approach
- Explanation of why abstractive summarisation is more relevant to your research

4. **Multi-Document Summarisation (MDS)**
- Introduction to MDS and its complexities
- Review of key methodologies and techniques used in MDS

5. **Challenges in MDS**
- Discussion on the challenges in MDS
- Explanation of pre-trained language models like BERT, RoBERTa, GPT-3/4, and T5

6. **Pre-trained Language Models in Summarisation**
- Review of studies utilising these models for summarisation tasks
- Discussion on the effectiveness of these models in the context of MDS

7. **Domain-Specific Summarisation: Focusing on Scientific Papers**
- Overview of the specific challenges in summarising scientific papers
- Review of previous work on summarising scientific literature
- Discussion on the gaps and potential areas for improvement in current methodologies

8. **Application of Advanced NLP Techniques in MDS**
- Overview of advanced NLP techniques, such as transfer learning, attention mechanisms, graph-based methods, and domain adaptation
- Review of studies applying these techniques in MDS
- Discussion on the potential benefits and drawbacks of these techniques in the context of your research

9. **Choosing and Preparing Data for MDS**
- Importance of dataset selection in MDS
- Techniques for data preprocessing and formatting
- Discussion on the quality and characteristics of suitable datasets

10. **Choice of Evaluation Metrics**
- Overview of the common evaluation metrics for summarisation tasks (e.g., ROUGE, BLEU)
- Discussion on their strengths and limitations
- Explanation of why specific metrics are chosen for this research

11. **Summary of Literature Review**
- Recap of the major findings from the literature review
- Identification of the gaps in the literature that your research aims to fill

digraph attention 
    rankdir=LR;
    node [shape=record];

    input [label="Input Embeddings"];
    scaleddotproduct [label="Scaled Dot-Product Attention"];
    multihead [label="Multi-Head Attention"];
    ffnn [label="Feed-Forward Neural Networks"];
    output [label="Output Embeddings"];

    input -> scaleddotproduct [label="Q, K, V"];
    scaleddotproduct -> multihead [label="Multiple Attention Outputs"];
    multihead -> ffnn [label="Concatenated Outputs"];
    ffnn -> output [label="Final Output"];