Literature Synthesis

Literature synthesis turns multiple paper notes into a reusable research map. It is different from summarizing one paper: the goal is to compare claims, evidence, limitations, and open questions across related work.

A synthesis note can be viewed as:

$$\mathcal{S}=F(\{p_i{\}}_{i=1}^n,\mathcal{Q})$$

where $p_i$ are paper notes and $\mathcal{Q}$ is the research question being clarified.

Synthesis Axes

• Task: what problem is being solved?
• Input and output: what entities, modalities, or structures are modeled?
• Architecture: what inductive bias is used?
• Objective: what training signal is optimized?
• Data: what benchmark, split, and preprocessing are used?
• Evidence: which claims are supported by which experiments?
• Limitation: what remains uncertain or weak?
• Reusable concept: what should be promoted into a concept note?

Process

1. Start from a concrete research question.
2. Select a small set of papers with comparable tasks.
3. Extract claims with Claim extraction.
4. Map evidence with Evidence table.
5. Compare methods with Paper comparison matrix.
6. Convert stable ideas into Concepts.
7. Write a Korean post only when the synthesis has a clear reader path.

Checks

• Are papers compared on the same task and metric?
• Are missing values marked rather than guessed?
• Does the synthesis narrow broad claims?
• Does it expose a real research gap?
• Are copied abstracts avoided?

Related

• Papers
• Evidence table
• Limitation taxonomy
• Research question
• 글감 로드맵