How to write better Survey Paper ?
A Researcher begins his research journey by first writing a survey paper in the domain of his research. Writing a survey paper helps a researcher in i) understanding his domain of research thoroughly ii) Identifying the existing research gaps, iii)Understanding the various parameters and their role in solving the problem and iv) Infrastructure and Data set requirements for research. In fact a survey paper is also a service to the scientific community. You are doing research for young research scholars. Instead of reading vast amount of papers to understand what a scientific topic is about, a researcher just need to read your paper and can start his research at the earliest with a clear direction in mind.
What is expected in a survey paper? A survey paper is a research paper which lists and analyses the latest research works in a particular research domain of interest. The survey paper derives some conclusions from the work carried out so far and provides new avenues for the future research. A good survey paper provides a concise but broad review of a domain that is accessible to a wide range of readers who are naive and willing to carry out research in the domain presented. This introduces two primary challenges for writing such a survey paper.
The first challenge is to pick representative papers from within the research area and summarize them concisely. There can be vast amount of papers available and survey paper has limited space to capture the key work in the field. The author needs to go through abstracts and conclusions for a relatively large number of papers and select a subset that covers selected topic area for detailed reading and presentation in survey. Identifying the papers having higher citations and which are published in conferences and journals of high reputation will have to be given higher priority for selection.
The second challenge is to make the reader comfortable in reading and comprehending the analysis done for the various papers. The author has to go through each paper considered for survey at least two-three times before deriving any conclusion.
A survey paper should
- Pick at least 10-20 papers on a specific topic from the collected paper list.
- The papers selected should be a mix of papers including the base paper in the selected domain to the most recently published paper.
- Should have its own analysis on the significance of the approach and the results presented in each paper
- Give a a critical assessment of the work that has been done.
- Include a discussion on future research directions
- Give precise details of the experimental setup used for carrying out research in each paper
- Compare only those works which have common experimental platform or data set . Otherwise you have to recreate a common platform or use common data set and test the methodologies used in various platforms.
A typical structure of a survey paper includes
Title
The primary function of a title is to provide a clear summary of the paper’s content. So keep the title brief and clear. Use active verbs instead of complex noun-based phrases, and avoid unnecessary details. Moreover, a good title for a research paper is typically around 10 to 12 words long. A lengthy title may seem unfocused and take the readers’ attention away from an important point. A good research paper title should contain key words used in the manuscript and should define the nature of the study. Think about terms people would use to search for your study and include them in your title. Do not use abbreviations in the title. Usually a Title for survey paper starts with ” A Survey on …..”, “Recent trends in ….”, “Advances in…..”. Some survey papers end with “……. : A Survey”. For example ” A Survey on leaf image analysis for bacterial disease detection” or ” Advances in leaf image analysis for bacterial disease detection” or “Leaf image analysis for bacterial disease detection: A Survey”.
Abstract
Abstract is a summary of a research paper describing the problem investigated, the methods applied, the main results and conclusions. Abstracts are a good way to sum up the key contents of a paper, from the research that it uses to the ideas that you want to share with the reader. It is a single paragraph containing minimum 200 words up to 300 words. An abstract offers a preview, highlights key points, and helps the audience decide whether to view the entire work.
Example: Extraction of meaningful leaf disease features by applying image processing techniques is a problem that has been studied by the image processing community for decades. Image processing research for leaf disease identification has matured significantly throughout the years, and many advances image processing techniques continue to be made, allowing new techniques to be applied to new and more demanding pathological problems. In this paper, we review recent advances in diseased part extraction of leaf images affected by pathogens , focusing primarily on three important Soft computing techniques namely : Neural networks, Fuzzy logic and Genetic algorithms. Throughout, we present tables that summarize and draw distinctions among key ideas and approaches. Where available, we provide comparative analyses, and we make suggestions for analyses yet to be done.
Key Terms
The purpose of keywords in a research paper is to help other researchers find your paper when they are conducting a search on the topic. Keywords define the field, subfield, topic, research issue, etc. that are covered by the article. Most electronic search engines, databases, or journal websites use keywords to decide whether and when to display your paper to interested readers. Keywords make your paper searchable and ensure that you get more citations. Thus, it is important to include the most relevant keywords that will help other authors find your paper.
For Example for the abstract written in the previous section the keywords can be
Keywords: Plant pathology, bacterial blight, diseased part extraction, Image processing, Soft Computing,
Introduction
A good introduction in a survey paper explains how the research problem has been solved by various researchers and creates ‘leads’ to make the reader want to delve further into research domain. Introduce the terminology of the field, describe what the various terms mean. The introduction does not have a strict word limit, unlike the abstract, but it should be as concise as possible. The introduction works upon the principle of introducing the topic of the paper and setting it into a broad context, gradually narrowing down to a research problem. The first task of the introduction is to set the scene, giving your paper a context and seeing how it fits in with previous research in the field. The first paragraphs of your introduction, can be based around a historical narrative, from the very first research in the field to the current day. The entire introduction should logically end at the research question. The reader, by the end of the introduction, should know exactly what research issue you are trying to survey with your paper.
Survey
The survey has to be based on specific theme of research which will help the reader in focusing his/her research on specific concepts. Some possible themes can be:
1. Complexity of the problem: There can be various types of solutions for a given problem domain and the author has to organize them in the increasing level of complexity or scale. For example in Image Processing scene analysis is one of the core problems. It can start from simple gray scale image scene consisting of one or two objects of same shape with constant background to a complex image scene consisting of objects of various shapes with varying background color.
2. Static vs. Dynamic : Many fields can be organized by static techniques, dynamic techniques, and even hybrid. For example Static or Dynamic Routing in Computer Networking.
3. Segregating the Design Space : Many systems are made up of components, so maybe for a computer network paper, the author could divide the problems into physical layer, application layer, session layer, transport layer, data link layer and physical layer.
4. Major approaches in a specific domain : For example, i) In Software Testing : Black box or White box testing ii) In Networking: Wired or Wireless Networking and iii) In Image processing Spatial or Temporal based Image Processing etc.
5. History of Development : Some research domains like Cloud Computing, Big Data Management, Mobile Technology, Television technology etc are linear in nature. Such developments can be explained in chronological order.
One can find various options for a selected domain of research and it is this organization that is the challenging part in writing a survey paper.
Following points are to be elaborated for each paper which is surveyed as a part of writing the survey paper.
– What are you going to tell about the paper under consideration.
– Research direction of the paper
– Methods , mathematical modeling and approach or algorithms used to solve the problem : eg. Fuzzy logic, Gaussian process , neural network etc.
– Whether the paper considers theoretical issues of the concept or solves any application using the concept?
– Is the paper considers the continuation of another work? is it an improvement on another work?
– How validation of work is done i)through theoretical proofs? ii) simulation? iii)hardware test bed?or iv)real life deployment?
– How the work is compared with other methods? and under what circumstances the method under consideration performs better?
-On what parameters the paper under consideration stands itself apart from other papers like i)higher performance ? ii) higher robustness? iii) lower computational complexity?
Author of each survey paper must be acknowledged by citing the paper referred. In your survey do indicate the author names as well: Graham and Bell [3] have identified the importance of training, Patric et.al. [3] developed a simple methodology etc. There are two reasons for this. One, it is the gratitude towards the authors, whose work you are referring. Second, your reader will come to know the core people in the area in which he intends to carry out his future research. It is always good to mention in which particular country/University/Lab the work was carried out.
Conclusion
The conclusion must answer the queries presented by your survey goals and objectives. The conclusion must be written in an interesting yet academic manner. No emotions should be attached to your conclusions but a commentary as a third person is required. Being the final portion of your survey report, the conclusion serves as the researcher’s final say on the subject of the survey. The conclusion must be a synthesis of the survey results with i) interpretation of each result ii) the proposal of a course of action based on result and the iii) solution to the issues that emerged from the survey. The tone of conclusion should match that of the results and the rest of the data collection process. The conclusion should be able to wrap up the entire survey from the formulation of survey goals up to the satisfaction of such objectives.
Example : After roughly two decades of research on leaf image analysis for pathological issues, many elements of pathological issues associated with leaf are well understood. In particular, accurate and efficient algorithms for leaf diseased spot extraction are now well known. As a result, during the past few years, we have seen the focus turn from the fundamentals of disease spot extraction to to more difficult problems such as, type of the leaf disease and the stage of the leaf disease. Few algorithms in this context are available. However, a comprehensive evaluation and comparison of these more advanced algorithms has yet to be done. One of our goals in this review was to consolidate existing quantitative results and to carry out comparative analyses . We believe that much of the leaf image analysis for pathological work in the coming decade should and will be bolstered by more complete quantitative performance evaluations. The recent article by Wimar [10] is a promising first step. Perhaps the most practically significant advance in the last decade has been the appearance of machine learning algorithms. However current implementation of Machine learning algorithms are still relatively simplistic. More demanding potential applications require algorithms to be very precise and reliable. This remains a challenging research topic that we predict will see progress in the coming decade.
Very much informative sir.
Thank you sir