Step 1: The first draft usually has a high repetition rate (unless you are a great god who writes word for word). You can use Wanfang and papertest to test it and then modify it sentence by sentence. This system is sentence-by-sentence detection, which means that any sentence you copy will be detected. This detection algorithm is strict, and the analysis from the point of view of program is simple. So everything sold online is very cheap. After measuring 30 thousand words, it feels good and cheap. (Note: 1 does not include your senior thesis of the last graduate student, so be sure to pay attention to the revision. Personally, if the school uses Wanfang to do the test, there is no need to take the VIP test and so on.
Make a copy of the electronic version of the paper first, and keep one copy. According to the test results, in a backup test paper, the duplicate parts can be deleted first, and those that cannot be deleted can be modified within 15 words. It is best to add and subtract characters, and changing the order is of little use. Some references have been deleted. If they cannot be deleted, you can modify them first. English documents can be changed from 15 characters to one word. Hand in the revised ones and recheck the system. The saved original paper is slightly modified and submitted to the paper version. That system is very troublesome. You can add many documents that you haven't seen or applied for. It can be seen that China people copy from each other, but in order to ensure the integrity and accuracy of the paper, don't change it at will. The paper version submitted must be considered, and it will not pass the test system after general inspection, so don't worry about the paper version.
Step 2: After modification, the repetition rate is greatly reduced. At this time, you can use HowNet to check. HowNet duplicate checking system is a piece-by-piece detection, which is more intelligent. After testing, it is basically to make some local changes. Finally, I checked with HowNet, and it was 4% online, but it was simply changed, and it was 1.5% at school.
Note: remember, the most taboo is to change the sentences of the paper illogical and illogical in order to check the duplicate, so that you can't escape the teacher. Haha, come on!
Detailed description of hownet system calculation standard:
1. After reading the introduction of this system, I have a question. This system is good for text copy recognition, but what about other contents, such as data and charts? Isn't it useless to detect it?
Among all kinds of academic misconduct, plagiarism is the most common and serious. At present, the detection system has reached a high level. The detection of plagiarism and tampering of charts, formulas and data is currently under development and has made great progress. You are welcome to continue to pay attention to the progress of this detection system and put forward more critical and constructive opinions and suggestions.
2. According to this system, less than 39% is displayed in yellow, so does it mean that it is within the tolerable limit? Recently, I read the news that the national social science fund project of a teacher in Shanghai University was cancelled because two papers he published were plagiarized, accounting for 25% and 30% respectively. Please specify what the warning line is.
Percentage only describes the proportion of overlapping words in the detected documents, and does not refer to the plagiarism of the documents. It can only be said that the greater the percentage, the more overlapping words, and the greater the possibility of plagiarism. Whether it is plagiarism or not and the severity of plagiarism need to be decided by experts after review.
3. How to prevent the academic misconduct detection system of dissertations from becoming a platform for personal revenge?
This is something we are seriously considering. At present, this detection system is only used by users at the institutional level. We have established a strict management process. At the same time, technically, we have also taken various measures to prevent malicious acts as much as possible, including a series of strict identity authentication and login.
4. The minimum detection unit is one sentence, so you can't detect one or two words in each sentence?
We also deal with sentences accordingly, and have an algorithm of sentence similarity. It is not the same sentence that is judged to be the same. Sentences have sentence-level similarity algorithms, and paragraphs have paragraph-level similarity algorithms. Calculating whether a document or paragraph is similar to other documents is based on this.
5. If the original word is taken from relevant books, but the word has been copied from relevant documents in the database, that is to say, the previous article also took the same word from relevant books, but the words marked in my paper are from relevant books, is this academic plagiarism?
The detection system can't draw a conclusion, whether it is plagiarism or not, and finally there is manual review. So if it is the situation you describe, experts will make corresponding judgments. Our system only provides all kinds of clues and basis, so that people can quickly grasp the information of test literature.
6. The authority of HowNet detection system?
The detection system of academic misconduct documents has not reached a conclusion, that is, the detection system does not characterize the detection documents, but only shows the similarities between the detection documents and other published documents and lists objective facts, and whether such detection documents belong to academic misconduct needs the final examination and confirmation of experts.