I designed two free gadgets for you who are writing papers.

This article was originally published in “Vista Newsletter (opens in a new tab)” No.153

Hello, hello!

If you are writing a thesis, this newsletter is for you.

The longest period of writing a thesis was the months spent writing chapter by chapter. There are two things that get stuck most often during this period: First, I don’t know if I can finish writing before the oral exam at the current speed; second, I write the literature review over and over again, always turning into abstracts one after another.

This week, I launched two free tools on the researcher base. Paper Backward Scheduling (opens in a new tab), work backwards from the date of the oral exam to calculate how many words you should write this week; Abstract String Detection (opens in a new tab), help you check the literature review and find out which paragraphs are just paraphrasing other people’s research.

There is no need to register or leave an email for either of them, just open it and use it, and the content you fill in will only stay in your browser.

In this issue of the e-newsletter, I want to talk to you about how I came up with these two tools, how to use them, and some small methods for writing papers that I compiled during the process of making the tools. Even if you don’t plan to use tools for the time being, these methods can be used directly.

Let’s start with a checkpoint map.

Last week, I posted a Thesis Blockage Map (opens in a new tab) in the researcher base, dividing the 21 topics that graduate students most often get stuck on from topic setting to submission into seven stages. Each card will write down the possible reasons why you are stuck, the first step you can take today, and what you should pay attention to when using AI.

As of today, these 21 cards have been clicked a total of 138 times. The number of times is not many, but the ranking is very interesting: the most clicked is “I forgot which article this sentence comes from” (20 times), followed by “the title is too big” (18 times), “downloaded too many PDFs to read” (17 times), and the fourth place is “literature review written as an abstract and skewered” (12 times).

Among the top four, three fall in the section of reading literature and writing literature review.

Seeing this result, I went back and checked the six original tools on the researcher’s base, and found that most of them were “fill in the blanks”: fill in a few boxes, get a card, use it once and end it. The most time-consuming months for graduate students are when I am writing chapter by chapter, and there is almost nothing I can catch.

Before I started, there were a few design principles that I insisted on, and I later used these two tools.

First, I only give suggestions and don’t make changes for you. It lists each issue and it is up to the author to decide whether to adopt it or not. The role of the tool is to “find it out”, leaving the judgment to the person who writes it.

Second, write the limits honestly. Each inspection is accompanied by “Inspection Scope and Limitations”, which clearly states what it does not care about. A tool that dares to write what you can’t do makes people feel more at ease.

Third, the data does not leave the user’s computer. The reason is very simple. The draft of the paper is a very personal thing. If it can be processed in the browser, it should not be sent to the server.

These three rules happen to be common to all tools in the researcher base.

Tool 1: Paper Backward Scheduling

URL: researcher.tw/tools/thesis-planner (opens in a new tab)

If you can’t finish the paper, it’s rarely because you’re lazy one day, but mostly because you don’t know how far behind you are. When I found out, there was only one month left before the oral exam.

When many people arrange their schedule, they will start from today and work backward: they will write the second chapter this month and the third chapter next month. The problem is that by arranging it this way, it is easy to miss three periods of time that do not belong to you.

The first paragraph is the time it takes for the instructor to read the full text. After you hand it in, the teacher has to read it and you have to correct it. During this time, you cannot continue writing.

The second paragraph is the time to send it to the oral examination committee. Most departments require that the paper be sent to the committee several weeks before the oral examination, during which time the paper is already being finalized.

The third paragraph is the time to wait for information. The questionnaires haven’t been collected yet, the interviews haven’t been completed, and Chapter 4 just can’t be written.

Therefore, the tool I designed deliberately works backwards from the date of the oral exam. Usage is very simple, only four steps:

1. Fill in the expected date of the oral examination, and then fill in how many weeks it will take for the teacher to review it, and how many weeks it will take for it to be sent to the oral examination committee and for the oral examination. Please fill in these two numbers according to the department’s regulations and the teacher’s habits.
2. List the target number of words and the number of words written for each chapter. The default is five chapters. The word count is just a common starting point. Please adjust it according to your academic discipline.
3. If you are still collecting information, fill in the number of weeks it will take and choose which chapter to write until the information is available.
4. The tool will calculate the number of days you can actually write before handing it over to your advisor, allocate time according to how many words are left in each chapter, calculate the completion date of each chapter, and average how many words you need to write per week.

My favorite is the extra fifth step: come back once a week to keep track of your word count. After recording it more than twice, the tool no longer just calculates “how much you should theoretically write per week”, but uses your recent actual speed to calculate the completion date, directly telling you how many weeks ahead or behind you.

Everyone can arrange an ideal schedule, but only by comparing your own speed can you know whether you can catch up or not?

The scheduled schedule can be downloaded as a calendar file and imported into Google Calendar or Apple Calendar. In addition to the completion date and deadlines of each chapter, there is also a weekly reminder in the file, asking you to come back and record your progress. You can also print it as a one-page PDF and bring it to discuss the schedule with your advisor.

Small reminder: word count is not quality, and writing fast does not mean writing well. This tool only saves time. Whether each chapter is well written or not depends on reading it with the teacher. If you want to be more productive every time you sit down to write, you can also refer to “The secret of completing the first draft of the paper on the weekend: the five-step process of “writing backwards and preparing materials” learned from Prof. David Stuckler”.

If you calculate that there are only about six weeks left until the oral exam, you can start preparing for the oral exam. My “Thesis Oral Examination Preparation Package (opens in a new tab)” starts three weeks before the oral exam. Let AI play the role of the oral exam committee member. It will ask you one thing at a time, find out what you know but can’t explain clearly, and then practice until you can speak it. AI only asks questions and does not write answers for you.

If you have to report progress in the laboratory every week or every other week, “Graduate Academic Briefing Package (opens in a new tab)” comes with three sets of templates for book discussion, progress report, and oral examination briefing. It also teaches you to let AI act as a strange reader before going on stage and find out page by page where the discussion will get stuck. There is also a combined purchase version (opens in a new tab) of the two copies. The two copies of the oral interview presentation template are shared and there will be no duplicate pricing.

Tool 2: Literature abstract skewering detection

URL: researcher.tw/tools/summary-chain (opens in a new tab)

I think you must have seen this literature review:

Lin (2019) found that ⋯⋯. Chen (2020) pointed out that ⋯⋯. In addition, Zhang (2021) believes that… Lee and Park (2022) show ⋯⋯.

One report after another, each sentence speaks for others. This is often what the instructor is referring to when he writes “lack of integration” on the side of a manuscript.

The most troublesome thing about this kind of writing is that it is difficult for you to detect the crux of the problem. Each sentence is correct when viewed individually, and the citation format is also correct. The problem lies in the lack of your judgment in the entire paragraph: where these studies are consistent, where they conflict, and what remains unanswered when combined.

Usage is very simple:

1. Paste the literature review you have written and copy it directly from Word. It supports two citation formats: author year (APA) and [1] number.
2. Click “Start Analysis” and the tool will mark sentences one by one in three colors: sentences that paraphrase a single study, sentences that compare or synthesize multiple studies, and statements that you have not cited.
3. It will point out two types of paragraphs that should be reviewed: paragraphs that are paraphrased in several consecutive sentences, and paragraphs where the first sentence is someone else’s research and lacks a topic sentence.
4. Each marked paragraph will be accompanied by a rewriting direction.

Of course, the tool won’t rewrite it for you. You have to write the sentences yourself, that is your literature review, and you will gain something.

Its judgment is based on sentence patterns: if a sentence contains a quote and starts with a certain author, it is considered a paraphrase; if words such as “compared to”, “overall”, “these studies”, and “yet” appear for comparison or induction, or if two or more documents are quoted simultaneously in one sentence, it is considered a synthesis. So it will misjudge. The mark is just to ask you to look back, not to score. Some disciplines are accustomed to introducing each chapter one by one, so please refer to the requirements of your supervisor.

I will give you four sentence patterns to change the paraphrase into a comprehensive one.

How to change it after being marked? Look at those sentences together, first determine their relationship, then choose a sentence pattern, write a sentence in your own words to bring them out, and put the original quotation in brackets.

The first is to find consensus: write the consensus first, then write the exceptions.

Most studies point out that ⋯⋯ (A, year; B, year), but ⋯⋯

The second is to discover conflicts: write out the conflicts and put forward your judgment on the reasons.

Different from the discovery of A (year), B (year) ⋯⋯, the difference between the two may come from ⋯⋯

The third method is different: group according to methods or samples and compare the results.

Research based on questionnaires mostly found…; studies that directly analyzed…

The fourth is to point out the gaps: Taken together, there is something that has not been answered.

However, these studies all focus on ⋯⋯, and ⋯⋯ is still lacking in discussion. This is exactly what this study aims to answer.

In addition, the first sentence of each paragraph should preferably be your argument rather than an author’s discovery. The reader must first know what this paragraph is trying to prove, and then the following documents will have their place.

Let AI ask you questions instead of writing them for you

The summary skewer detection results page has a “Bring your AI” section. It will connect your marked paragraph to a prompt word and let you copy it to your own ChatGPT, Claude or Gemini. The researcher base will not call any AI, it is up to you to decide where to post it.

The point of this prompt is to limit what the AI can do:

• Do not rewrite for me or produce complete paragraphs.
• Do not include references that I have not cited, and do not make up any authors, years, or research results.
• Respond only based on the paragraphs I provide, and ask me directly if anything is not covered in the paragraphs.

Then ask it to do three things. First, point out which sentences are simply paraphrasing. Second, find out which studies can be compared together and throw them to you by asking questions. Third, ask you a question in each paragraph to help you come up with the topic sentence of this paragraph. The last sentence is “Please ask me questions one paragraph at a time, and wait for my answer before moving on to the next paragraph.”

I really like this usage. AI is good at discovering that “these two documents seem to be comparable together.” But why the two documents have different results and what significance they have for your research can only be answered by you who have read the original texts. Let AI be the one who asks the question and you the one who answers. What is written is yours. If AI is used in your dissertation, remember to disclose it in accordance with the relevant regulations of your school department. You can also refer to “How graduate students should use AI so as not to cross the line: A practical guide to maintaining academic integrity”.

If you find that almost the entire paragraph of your paragraph is paraphrased, the problem may not be with the writing, but with the way you read the literature: read one by one, take notes one by one, and naturally report one by one when writing. “Gemini Notebook Research Workflow Manual (opens in a new tab)” demonstrates how to put a whole batch of documents into the same notebook for cross-reference, and identify consensus, differences, and gaps before writing.

If you often ask AI to help find literature or organize opinions, but are worried that it will give you a paper that cannot be found, the 30 prompts in the “AI Academic Research Prompt Manual (opens in a new tab)” are each accompanied by “How AI may go wrong and how you should check.” Regarding the current situation and verification methods of fake documents, I also wrote How many references have you clicked on from AI? 〉Organized.

Two things to do after writing the paper

The above two tools deal with the process of writing a paper. After finishing writing, there are still two things that often give people headaches.

One is a revealing statement. AI is used, but the statement only contains one sentence: “This research was assisted by AI.” This situation is becoming more and more common. The free tool “AI Usage Disclosure + Ethical Self-examination (opens in a new tab)” on the researcher’s base can help you sort out a draft first (for usage, please refer to “It is not shameful to use AI, it is risky to hide it”); if you want to check the original regulations of the journal or school where you are submitting, “AI Usage Disclosure Statement Template Pack (opens in a new tab)” compiles a comparison table of regulations for international publishers, Taiwanese journals and school units, as well as Chinese and English templates for various usage scenarios.

The other thing is the review comments after submission. I received more than twenty comments mixed into one letter, and it took several days just to sort out the format, reply to them one by one, and mark the positions for modification. “Reply Packet for Submission and Review Opinions (opens in a new tab)” takes you step by step from reading the decision letter, tracking modifications item by item, to writing a reply letter, and switching submissions after rejection. The complete introductions of the six manuals are all placed on the Manual Page of the Researcher Base (opens in a new tab). You can first check which one is suitable for where you are stuck now. If you want someone to practice with you, you can also check out AI Empowered Academic Writing Workshop.

Finally, please do me a favor

These two tools have just been launched, and there must be many areas that are incomplete. If after using it you find something wrong with the calculation, something you can’t understand, or if the summary skewering detection marks a certain sentence incorrectly, please leave a message and let me know. It’s best to include that sentence. Examples of misjudgments are very precious to me, and I rely on them to correct or enhance the next version of the rules.

If you have friends or classmates who are writing papers, you are welcome to share these two tools with them, or invite them to subscribe to the “Vista Newsletter (opens in a new tab)”.

See you next time!

Vista