From Abstract to Conclusion: How I, a Ph.D. Student, Read and Digest Academic Papers

• How to Review a Paper
• 怎麼讀論文

How to Review a Paper

How exactly should one review a paper? This is likely a question many people have who are about to attend graduate school. No one teaches you in undergraduate, and certainly not in graduate school either. Of course, you could read it from beginning to end, but with the sheer volume of papers these days, we need to quickly determine whether a papers is relevant to our needs before we dive into a detailed, intensive reading. Here, I aim to share some methods for reading research papers that I’ve developed from referencing various literature and advice from professors.

We know that a thesis has a certain structure, which is why it should be read three times, and not just from the beginning straight through to the end. Each reading has a specific goal and builds upon the previous one:

• The first reading gives you a rough understanding of the thesis, mainly to be able to briefly outline what research questions the thesis addresses.
• The second reading allows you to grasp the content of the thesis and explain the arguments and supporting evidence to others.
• The third reading helps you to deeply understand the papers.

The Three Pass of Review Papers

Pass 1

The first pass of reading is a quick scan to understand the structure of the entire paper, then decide whether to proceed with more reading.

1. First, read the title and the Abstract; after reading the Abstract, skip directly to the Conclusion section. The Conclusion usually contains the same outcomes as the Abstract, but it will include the actual results and experimental data to support the conclusions.

[Swin Transformer]

Next, we can look at the key figures for the Method and Result to get an idea of the general structure of the paper and the experimental outcomes. In my first pass through, I will write a review, mainly divided into three parts: Concept, Conclusion, and Contribution.

1. Concept: What is the core of the entire thesis? What problem does it aim to solve? Or what is the research direction? Try to write the concept into an abstract paragraph.
2. Conclusion: What is the conclusion? What methods have been proposed? Which problems have been addressed? Or what are the limitations of this study? Organize in bullet points.
3. Contribution: Finally, I will summarize in one paragraph or sentence the main contribution of this thesis — what is it and why does it matter? If throughout the paper there is a constant reference to the work of other researchers, then it is necessary to articulate their findings in a way that immediately connects with our own.

Pass 1 Review

Pass 2

After the first pass, if you think it is relevant to your own research, or you want to gain a deeper understanding of the theories or research methods in this paper, during the second reading you should focus on grasping the entire research flowchart. You don’t need to work through the proofs or such details step by step on your own.

After the second pass, you should be able to summarize the main arguments and supporting evidence of the paper and explain it to others.

In this part, I will first break down each section into individual paragraphs. For example, the Introduction might have five paragraphs, and you need to understand the main idea of each one. During the reading process, I will highlight what I believe to be important parts in yellow, and if I come across anything questionable, I will underline it in red. Finally, I will summarize the key points of each section. By trying to connect the questions that arose during the reading to the points you deemed important, you can usually find the answers. This can help achieve an effective connection between the concepts and the problems.

Pass 2 Review

Pass 3

The third close reading primarily involves empathetic thinking, where one reads as if taking the position of the author to consider if anything could be done better. For instance, how the same experimental results could be presented differently, or how one might approach the arguments and conclusions, and so on. This comparison between the original paper and one’s own thoughts allows an understanding of how the same points may be presented differently by different people, including oneself. Through this, one can learn from the ideas of other researchers, which can lead to a more comprehensive set of arguments when writing one’s own paper. By the end of this reading, one should be able to recount the structure of the entire paper from memory, particularly being able to accurately identify any potential issues with the hypotheses, arguments, experiments, or analyses.

Finally, I will also jot down any thoughts this paper stimulates regarding future work. This includes considering how one might extend the current work or, if other researchers have already published related work, using their papers as a springboard for further ideas.

Paper Organization

After reading a paper, it is always necessary to organize the material, and of course, everyone can use their preferred software to do so, such as Zotero, Endnote, Notion, Obsidian, Heptabase, etc., so it can be used later on. Here, I will mainly divide it into four sections.

Paper Organization

1. Paper Title
2. Classification: Classify the domain or topic of this paper, which will make it easier to find related papers in the future.
3. Read: Whether the paper has been thoroughly read three times.
4. Review: As mentioned earlier, regardless of whether you have read it three times, you can write a post-reading review, including the main argument, conclusions, contributions, and reflections. This can quickly help you recall the paper when you look back on it in the future.

I hope this article can help students who have just entered research, and you can also give me some suggestions for me to continuously adjust my methods of reading literature.

reference

1. How to read a research paper.
2. How to Read a Paper
3. 做一個好研究，寫一篇好論文 | 研究生新訓講堂

怎麼讀論文

到底要怎麼讀論文?相信應該多數要讀研究所的人都會有的問題，大學沒人教，研究所更不會有人教你，當然從頭讀到尾也是可以，但是現在論文這麼多，我們需要快速知道這篇論文到底是不我需要的，然後再進行詳細的精讀，這裡就是來介紹一下參考一些文獻以及老們後我自己使用的閱讀論文方法。

我們知道論文有一定的結構，分為三次閱讀這篇論文，而不是從頭開始一直讀到結尾。每次閱讀都有特定的目標並在之前的閱讀基礎上進一步深入：

• 第一次閱讀讓你對這篇論文有一個大致的了解，主要是能概說這篇論問題出了什麼研究。
• 第二次閱讀讓你掌握論文的內容，可以像其他人解釋這篇論文的論點，有什麼支持證據。
• 第三次閱讀幫助你深入理解這篇論文。

論文三步驟

Pass 1

第一遍閱讀是快速掃描以了解整篇論文的架構，再決定是否要進行更多的閱讀

1. 首先閱讀標題、摘要Abstract，讀完摘要直接跳到Conclusion的部分，Conclusion通常跟Abstract裡的結論是一樣的，但是會有實際的結論、實驗數據來支持結論。

[Swin Transformer]

接著可以看一下Method跟Result的關鍵圖表，知道大概論文的架構跟實驗結果，第一次Pass through我會做一篇review，主要分為三段Concept、Conclusion、Contribution

1. Concept : 整篇論文的核心是什麼？想要解決的問題？或是研究方向？試著用將concept寫成一段abstract
2. Conclusion：結論是什麼？提出了什麼方法？解決了哪些問題？或是此研究有哪些不足？用列點的方式整理
3. Contribution ：最後我會總結一段話來說明論文主要貢獻是什麼？那為什麼要這樣做？如果在撰寫論文時總是會提及其他研究者的研究，這時候就需要以一段或一句話來述說別人的研究成果，這時候就派就可以馬上做連結。

Pass 1 Review

Pass 2

在經過第一次閱讀如果認為這和自己的研究相關，或是想更深入了解這篇論文的理論或是研究方法，在第二次閱讀時就要了解整個研究的流程圖，證明之類的細節可以不用到自己去一步一步推導。

第二次閱讀後應該要能夠總結論文的主要論點和支持證據，並向其他人解釋，基本上有讀過Pass 2就可以向老師報論文了。

這裡我會先把每一個Section拆解成一個一個段落，例如Introduction可能有五個段落，分別要讀懂每一段想表達的重點，在閱讀的過程中我會將我認爲是重要的部分畫黃線，讀到覺得有疑問的地方會畫紅線，最後把每一個Section的重點做一次整理，可以試試把閱讀時有問題的部分連結到認為的重點，通常都會找到解答，可以達到概念問題連結的效果。

Pass 2 Review

Pass 3

第三次精讀主要會以換位思考來閱讀，可以把自己當作作者的立場上能否做得更好

例如同樣的實驗結果要如何呈現，或是論點、結論等等。這種對比原本論文與自己的想法能了解同一個論點其他人和自己呈現上的不同，可以學習其他研究者的想法，之後再寫自己的論文就會有更全面的論點。在這次閱讀結束時，應該能够憑記憶重述整個論文的結構，特別是能應該能夠精確指出假設、論點、實驗或分析可能存在的問題。

最後我也會記下本篇論文對未來工作的想法。未來可以延伸，或是已經有其他研究者有提出相關的論文就可以以他的論文去做發想。

論文整理

讀完論文總是需要整理，當然大家可以用自己習慣的軟體整理Zotero、Endnote、Notion、Obsidian、Heptabase etc.，以便以後可以使用，這裡我主要會分為四格

Paper Organization

1. Paper 題目
2. Topic 分類 : 可以分類這篇論文是屬於哪一個領域、主題，之後想找相關領域的paper會比較方便
3. Read ：是否有完成3次pass
4. Review：前面有提到不管是否有讀到三次，都可以寫成一篇讀後Review，包括主要論點、結論、貢獻、心得，未來再回頭看的時候可以快速幫助自己回想這篇論文。

希望這篇能夠幫助剛進入研究的同學們，也可以給我一些建議，讓我持續調整自己閱讀文獻的方法

reference

1. How to read a research paper.
2. How to Read a Paper
3. 做一個好研究，寫一篇好論文 | 研究生新訓講堂

授權遵循CC BY-NC-ND 4.0 協議，轉載內容請標明出處，同步發表到個人網站 Y Kuan方格子