How to write a good thesis——Ms. Min's experience in "Thesis Writing"
Article directory
When I first entered the graduate school, I thought that writing a thesis was not particularly difficult. As long as I mastered the direction and method, it was not difficult to write a thesis, but I didn't know where to start. After taking the "Thesis Writing" class taught by Mr. Min, I found that there are many norms for thesis writing, and some places also require skills. Through taking this course, I have a clearer understanding of the whole process and precautions of thesis writing. The specific summary is as follows:
1. Basic concepts of academic papers
What——is a staged summary of the research work, a scientific research essay, and a table with a clear structure
Why—meeting graduation requirements, title evaluation, knowledge inheritance and academic promotion
How - read the literature summary template by yourself or use the template summarized by others
When and Where - the sooner the better, anytime, anywhere
2. Topics
1. Read more literature, learn and understand the development and research in related fields, and choose the fields you are relatively interested in and good at for research and topic selection.
2. The scope of topic selection should be appropriate. The topic is too large to be developed within a relatively narrow range. Too small, the content cannot be expanded, and there are too few documents. Therefore, whether it is appropriate or not has a great impact on the success of the paper.
3. The topic selection should try to combine with the experimental object to raise key questions, combine it with reality, consider certain advancedness and appropriate difficulty, and have both theoretical analysis and experimental explanation.
4. If we don’t know how to choose a topic, then we can ask our tutors, brothers and sisters in the laboratory for advice, or find out slowly in the experiment.
3. Preparation
writing tool
LaTeX is recommended, where writing and formatting are properly separated. Allows you to define styles for individual elements, which will be styled consistently throughout the document; instead of editing formatting every time an element is added, just do all formatting edits once eventually. LaTeX includes a large number of packages to help solve complex problems in LaTeX; most of them can be found on CTAN (Comprehensive Tex Archive Network). Besides that, compared to word processing programs, it's easy to search for something like x 5 x^5x5 or \footnote such expressions.
Don't write from scratch, it is best to find the source file of the journal's accepted papers and find the Latex template file. Each journal will provide one or more format files. Latex source files only need to make a small change to get a completely different layout.
| IEEE Publisher | Elsevier Press | Springer Press | other publishers | |
|---|---|---|---|---|
| cls file - controls the overall format of the paper | IEEEtran.cls journal mode\documentclass[journal]{IEEEtran} title and abstract single-column, text double-column typesetting effect is very close to the official publication effect. Conference mode\documentclass[conference]{IEEEtran} is not as compact as journal mode. |
elsarticle.cls single column in reviewer mode: \documentclass[preprint]{elsarticle} |
journal mode svjour3.cls conference mode llncs.cls |
Use usepackage to import packages and try not to introduce special packages Use \newtheorem to add some automatic numbering items |
| bst file - controls the format of references | IEEEtran.bst title and abstract single-column, text double-column. |
Numbering mode: elsart-num-sort.bst, the format is number [1][2], etc. model5-names.bst, if it needs to be used as a part of the name when citing, such as subject and object, you need to use \citet, otherwise use \citep. |
splncs.bst |
Read a large number of high-quality papers
Read some good, high-quality papers. These papers can be selected from top journals and top conferences in the corresponding fields. For fields such as artificial intelligence, computer vision, and machine learning, there are CVPR, IEEE, AAAI, ICCV, etc. There are many benefits to reading top journal papers. It is good for you and the reviewers to be able to learn the writing norms and writing styles of top journal papers, so that your papers can avoid some low-level mistakes.
4. How to write an academic paper
topic
1. The title is the beginning of a paper, and it must attract attention. Reviewers will not be interested in reading a bad title.
2. It should be easy to understand. The terminology used in the paper should be commonly used in this field, so that people can understand it at a glance.
3. It should be easy to search. Popular terms can be found by others, and can also help increase the citation frequency of papers. For a scholar, the number of paper citations is more important than the number of papers.
4. The length is best controlled between 40-60 letters.
Summary
From the perspective of Chinese, "abstract" is to extract the important content of the paper; from the perspective of English, Abstract is to extract the important content. It usually includes three parts: review of existing work, description of this work, experiment Results. Some journals clearly require descriptions in terms of Background, contribution, and experiments. Teacher Min Fan summarized the following 10 points of the summary:
1. The problem and its importance.
2. Already have a job.
3. Existing work limitations.
4. This article works.
5. The first technique/step/aspect/advantage/contribution of the method in this paper.
6. The second technique/step/aspect/advantage/contribution of the method in this paper.
7. The third technique/step/aspect/advantage/contribution of the method in this paper.
8. Experimental setup.
9. Experimental results.
10. Boost.
Key words
Before the development of modern search engines (including word cloud), keywords were an important way to retrieve papers. Although people can perform full-text searches now, keywords are still very commonly used. Need to pay attention to the following points:
1. Keywords are often seen as a supplement to the abstract.
2. Generally, 3-5 keywords are required.
3. Keywords generally consist of 1-3 words.
4. Sort the keywords alphabetically.
introduction
The original intention of "Introduction" is to pave the way for the paper. However, in computer English papers, Introduction needs to tell a complete story. In recent years, some good Chinese magazines have gradually adopted this style.
1. It should give a detailed description of the whole paper.
2. The most stereotyped way, is to adopt the same rhythm as the abstract.
3. Each paragraph should have 5–10 sentences. In this way, the paragraph has 50-150 words . If two adjacent paragraphs are too short, they should be merged together. If a paragraph is too long, it should be split, or the text should be reduced directly. Too short paragraphs lead to fragmented feeling, too long paragraphs will It is tiring to read.
4. In the field of computer science (especially top conferences), it is popular to put "a picture of the beginning" in the introduction.
5. Some journals and conferences also like to let the authors list their contributions separately and put them in the experimental statement After the paragraph. This has the same meaning as the Highlights of the paper.
6. Journal papers require writing the organizational structure of the paper. The last paragraph of the introduction can start with The rest of the paper is organized as follows. No need to play tricks, A step-by-step presentation is sufficient.
literature review
Most references should be cited here.
1. Each paper should have a literature review.
2. Can be described in different places.
3. Documents need to be introduced in categories.
4. The citation of references is generally only an appendage.
5. References should not be used as the subject, object, etc. of a sentence.
6. A review based on the year as the main line is also not worth advocating.
7. Don't cite too many documents at once.
8. Full-sentence citations are discouraged.
9. Reasonably comment on the advantages and disadvantages of related work.
experiment
For machine learning papers, the experimental part occupies a large part of the paper, so it is crucial to write this part well. Teacher Min Fan also gave his experience accumulated over the years. details as follows:
The first is the data set, the more the better, you can’t make your own data set, you need to use the public and universal data set to prevent it from being called a toy data set by the reviewers. Secondly, the writing method adopts the method of self-questioning and self-answering. These questions are raised before the experiment, and they are answered one by one after the experimental results are listed. Finally, internal and external comparisons are made. Internal comparisons show the performance changes caused by parameter changes and the advantages of the main solution compared with its variants. The external comparison is to compare with the existing algorithms, which requires the help of graphs and tables to make readers understand at a glance.
in conclusion
1. The conclusion is generally not too long, 5 sentences is enough. If you want to discuss more internally, you should add a separate Disussions subsection before this section. 2. Avoid using the same
sentences as in the abstract. The abstract says we do What things have been mentioned, and what observations and conclusions we have obtained here should be mentioned. That is to say, the conclusion is more specific than the abstract. It can be specific to a certain algorithm, a certain property, a certain theorem, and a certain experimental result in the paper, so that It is naturally distinguished from the abstract.
3. If you want to discuss further work, you can list 3 to 5 items, which are not counted in the word count of the Conclusion. Readers are likely to be very concerned about this part, because they can continue to research in this way . As a research work, it is more important to open a door than to completely solve a certain problem. If this part is well written, there will be many citations. The number of citations is also more important than the number of papers published.
references
1. Each paper should have a literature review
2. Can be described in different places
3. Documents need to be introduced in categories
4. The citation of references is generally only an accessory
5. References should not be used as the subject, object, etc. of a sentence
6. A review based on the year is not worth advocating
7. Don’t cite too many references at once
8. Discourage full-sentence citations
9. Reasonably comment on the advantages and disadvantages of related work
References refer to citing published relevant documents for writing a thesis, which is an indispensable and important part of the thesis. References reflect the background and basis of the research work, provide readers with the source of relevant information, and the works have real and extensive scientific The basis shows that the author respects the seriousness of other people's research results, and also avoids the suspicion of plagiarism or plagiarism. Well-known references are not only an important part of the thesis, but also the last part of the thesis, which has an irreplaceable position.
picture
Use vector graphics as much as possible. In this way, there will be no distortion when zooming in. Except for low-level tools such as drawing, the graphics generated by visio, matlab, and python can all be saved as vector graphics. It is recommended not to use Excel to draw graphics
.
Legends can be quite long. Some journals require pictures to be self-enclosed, that is, just looking at the explanation of the legend is enough. So some legends can have three lines or longer. Color pictures are very expensive. For example, ESWA
requires A color picture costs $200. Some journals, such as KBS, are quite nice. You can choose black and white for the print version, and color for the electronic version. Therefore, you should ensure that your pictures can still express your meaning when printed in black and white. Generally speaking,
quote When using a picture, it should be consistent with the number of the picture. For example, use Fig. 1 to correspond to Fig. 1. But when it is used as the subject, it is recommended to use Figure 1. Some journals require a
Graphical abstract, that is, use a picture (including legend) to describe the paper The main idea is clearly introduced. It is worth spending a month drawing this picture.
Algorithm Pseudocode
Algorithm pseudocode is one of the core of the paper.
1. It is necessary to explain the input and output;
2. The method (function) name can be written or not, if it is called by other methods, it must be written;
3. It is necessary to write comments for the main steps;
4. The length is controlled at 15-30 lines;
5. Mathematical formulas or references to existing mathematical formulas can be used;
6. Unimportant steps can be omitted;
7. Generally, it is necessary to analyze time and space complexity, and write supporting properties and corresponding tables to to make it more standard.
5. Common problems and solutions
Pay Attention to Words in Essay Writing
Through the explanation of teacher Min Fan, I learned about the usage of some words. For example, the word and needs to be used with caution when writing, which is easy to cause confusion; for the word only, put only as far back as possible when writing. At the same time, I also learned that if there are two similar words and there is a tangle when using them, we can check the frequency of use of these two words on https://www.linggle.com and choose the most frequently used words.
Every field has sensitive words that grab the reader's attention and make or break a paper. For example: address and solve. In the field of machine learning, few problems can be really solved, so you need to use address the issue instead of solve the problem.
Writing Standards for Equations and Equations
It is recommended to use Latex for processing equations, rather than Word in Windows. Variables should be in italics, and sets, arrays, and vectors should be in bold. All equations should be numbered. If a mathematical expression is too long, it should be divided into multiple expressions to avoid exceeding the correct boundaries of the page
blank use
There should be no blanks before periods and commas, and there should be a blank separator between adjacent sentences. There should be blank space before and after the cited quotation, there should be blank space between adjacent quotations, and conceptual sentences should be concise. Excluding the abstract and introduction, it is best not to exceed 15 words per sentence. Take care to avoid repeated word usage in one sentence or two adjacent sentences, and the same meaning should not be repeated.
6. Review comments
1. If the manuscript is submitted to a journal, if the editor asks you to revise it, there will probably be nothing wrong with it.
2. You must have a correct attitude in your thinking. Editors and reviewers spend a lot of time voluntarily reviewing manuscripts not to embarrass the author, but to help improve the quality of the paper
. 1. The question should be answered directly, instead of talking about it.
4. The reply should not be too long. The key point is to make corresponding revisions in the text. Reviewers pay more attention to how the text of the paper is revised. After all, it is the content presented to readers
5. The modified parts of the text should be marked in blue.
6. For errors not found by the reviewers, it is best not to point them out to the reviewers.
7. Summary
After the study of the "Thesis Writing" course, I learned step by step what preparations need to be done before the thesis, how to choose the topic, how to determine the topic, the main content of the thesis framework, writing skills and problems that need attention, etc., in Some mistakes can be avoided in the subsequent thesis writing, and the process of writing the thesis will be clearer. Now it is only a study at the theoretical level, and true knowledge can only be obtained in practice. The writing of the thesis is not accomplished overnight, it needs a lot of revisions to improve, it also requires us to have enough patience, and we can also communicate with the tutor a lot, etc. These are all indispensable steps for a good thesis.
Thanks to Mr. Min Fan for leading us to understand thesis writing. Thesis writing is a long and cumbersome process that requires a lot of literature review, basic preparation, and experimental verification to get good results. There is a long way to go in the mountains of books, and there is a long way to go for scientific research. In the future thesis writing, we have to learn slowly, accumulate, be modest, and be serious in order to gain something.