Retracted: The Relationship between Machine Translation and Human Translation under the Influence of Artificial Intelligence Machine Translation

,

Tis article has been retracted by Hindawi following an investigation undertaken by the publisher [1].Tis investigation has uncovered evidence of one or more of the following indicators of systematic manipulation of the publication process: (1) Discrepancies in scope (2) Discrepancies in the description of the research reported (3) Discrepancies between the availability of data and the research described (4) Inappropriate citations (5) Incoherent, meaningless and/or irrelevant content included in the article (6) Peer-review manipulation Te presence of these indicators undermines our confdence in the integrity of the article's content and we cannot, therefore, vouch for its reliability.Please note that this notice is intended solely to alert readers that the content of this article is unreliable.We have not investigated whether authors were aware of or involved in the systematic manipulation of the publication process.
Wiley and Hindawi regrets that the usual quality checks did not identify these issues before publication and have since put additional measures in place to safeguard research integrity.
We wish to credit our own Research Integrity and Research Publishing teams and anonymous and named external researchers and research integrity experts for contributing to this investigation.
Te corresponding author, as the representative of all authors, has been given the opportunity to register their agreement or disagreement to this retraction.We have kept a record of any response received.

Introduction
Machine translation refers to the mutual transformation between two languages with the help of computers.Its fundamental algorithm is to integrate the corresponding relationship between the two languages into the word database in advance.When the language to be translated is input, the translated sentences will be split according to the word structure, and then the split required translation words will be transformed into the words corresponding to the translation language according to the corresponding relationship in the word database, Finally, it is integrated into the translated sentence meaning according to the part of speech structure of the target language [1].Human translation means that the translator first masters and masters the two languages to be translated and then translates the language object to be translated into the translated language content after understanding the meaning.'e two methods have their own advantages and disadvantages.In terms of accuracy, the accuracy of human translation is usually greater than that of machine translation [2].However, in terms of the portability of translation, the convenience of machine translation is far more than that of human translation.'is is because machine translation can get rid of the limitation of translators.It carries out machine translation according to the predetermined correspondence of database words, and there is no need for the object who puts forward the translation demand to master the translated language [3].As far as the shortcomings of the two are concerned, the biggest disadvantage of human translation is that the quality of translation results greatly depends on the translator's translation skills.At the same time, because the translator digests the language to be translated first and then transforms it into another language he is familiar with, the meaning tendency of the translated content is easily affected by the translator [4].In the age when the computer has not been invented, the communication between different languages mainly depends on translators to translate works

R E T R A C T E D
such as characters so that people can understand the local customs and national culture outside different languages.Yan Fu, a famous translator in early China, once proposed that translation should follow the principle of "faithfulness, expressiveness, and elegance."Among them, "letter" mainly refers to the accuracy of translation.'e translated content should correspond to the meaning of the original text and should not violate the meaning of the original text [5]."Da" refers to the fluency of translation.Due to the different semantic structures and speaking methods between different languages, if the translation is carried out word by word, the content as a whole may not be smooth or reversed, making people unable to understand the meaning of the original text.In the process of translation, translators should not stick to the format of the original text, reintegrate according to the translated language, and make the translated content smooth and clear without violating the meaning of the original text [6]."Elegance" refers to the fit between the translation content and the language.In the process of translation, for example, the sentence patterns and artistic conception of the two languages are different when the ancient text is translated into modern vernacular.If the ancient text is simply translated into modern vernacular, it will lose a bit of charm, so the translator should keep the original text as quaint as possible in the process of translation.In China's early translation works, translators followed the above principles for manual translation [7].However, it seems unrealistic to require machine translation according to this standard.Machine translation mainly depends on computer language, and its underlying logic is binary language that can only distinguish 01.'erefore, computer translation cannot think about the smoothness and elegance of translation like human beings.'erefore, with the popularization of artificial intelligence machine translation, the problems of machine translation accuracy are well solved.In this context, we should also reconsider the relationship between machine translation and human translation.'erefore, this paper focuses on the principle of artificial intelligence machine translation and its impact on the relationship between machine translation and human translation.

Related Work
According to the time when the artificial intelligence algorithm appears, machine translation is mainly divided into two stages-traditional machine translation and neural machine translation, that is, what we call artificial intelligence machine translation.'e origin of machine translation comes from the pursuit of hegemony between the United States and the Soviet Union.During the cold war, the United States had a demand for translation in order to learn more about the relevant information of the Soviet Union [8].IBM was the first company to use computers to complete translation tasks.'ey used computers to translate Russian into English for the first time.'is is the first time that human beings use computer to translate.In fact, the function of the translation system is very simple, which is very similar to today's common language manual.Nevertheless, IBM has also opened the competition for the development of machine translation among countries.However, in the 40 years since the rise of computers, machine translation has not made greater progress and remains at the stage of traditional machine translation [9].In 1966, the American automatic processing Advisory Committee even declared that machine translation was useless, saying that it was not necessary to continue its research.Until the 1990s, with the further development of computer network technology, people began to carry out mathematical modeling and analysis through computer technology.At this time, it is found that a large number of parallel and research can be carried out on a language through statistical knowledge, and the statistical analysis results can be imported into the database to establish the best mathematical model, and finally machine translation can be carried out through the seen mathematical model [10].'is discovery once again made machine translation a hot research field and then achieved a lot of results.For example, Li explored the shortcomings of traditional machine translation in one translation according to the word comparison relationship, proposed to use the mathematical model to make entropy statistics on the translated content, and differentiated training on the statistical results according to the maximum entropy.'e best training result is the translation priority and the highest content [11].'is method greatly improves the performance of statistics-based machine translation mechanization and greatly improves the translation results of words with multiple meanings.And this method is far more innovative than other studies in the same period.'en, in 2006, Google created Google translation platform, which is one of the largest translation platforms on the Internet.'e underlying machine translation algorithm is still statistical machine translation, which means that statistical machine translation method has become the mainstream method in practical application.In the following time, most translation platforms adopt statistical machine translation method, but the biggest problem of this method is that the translation results mainly rely on the statistical results obtained after the computer analyzes a large number of parallel sentences [12].However, insufficient attention is paid to the sentence meaning and word meaning of the translated object itself.In the process of translation, grammar and other problems often occur.'erefore, statistical machine translation algorithms are still under continuous research, but due to the limitations of the development of the times, these shortcomings of statistical machine translation have not been well solved.Until 2013, nal developed a decoder architecture suitable for machine translation based on neural network algorithm under artificial intelligence.'e architecture uses the cyclic neural network for autonomous learning and converts the learning results into a continuous vector.'rough the continuous training of the vector, the translation results of the target language are finally obtained [13].'is research is the first time to combine artificial intelligence with machine translation.Later, with the development of artificial intelligence neural network, it is found that artificial intelligence has the ability of autonomous learning, which can carry out autonomous learning on the existing 2 Mobile Information Systems

R E T R A C T E D
results of statistical machine translation and obtain the most appropriate translation results again.And experiments have proved that the performance of artificial intelligence machine translation model is much better than that of statistical machine translation model when the learning corpus is sufficient.'erefore, more and more researchers began to explore machine translation under artificial intelligence [14].In 2015, Red combined with the principle of psychological attention mechanism and introduced it into the artificial intelligence algorithm, aiming at the problem that artificial intelligence could not focus on the fixed continuous vector in the training process of translating long sentences, resulting in a large loss of source language information in the training process, which significantly improved the accuracy of artificial intelligence translation model [15].

Method
'e accuracy of human translation can approach 100%, but it also depends on the level of translators, the level of expression of the original text, the industry, the time of submission, and other factors.'e accuracy of machine translation depends on the language, industry, the quality of the original text, the training corpus, the training model, and other factors.From the perspective of translation fluency, human translation stresses "faithfulness, expressiveness, and elegance," but it will not be fully reflected in actual commercial translation.'e development of mathematical logic has experienced continuous improvement.From the initial assumption of universal language to the establishment of logical typology.In essence, the establishment of mathematical logic is to symbolize the human thinking process, so a set of highly formal symbolic languages is constructed.To further explore the relationship between human translation and machine translation, we must first analyze the two translation methods and explore the relevance of the results from the method.'e methods and meanings of human translation have been described in detail above, so this section only focuses on the traditional machine translation and artificial intelligence machine translation methods.In this part, we first introduce the most commonly used method based on editing distance in traditional machine translation methods as an example.On this basis, we further introduce the optimization direction of artificial intelligence machine translation with transformer model and compare their translation results with human translation results to find the similarities and differences.'e transformer algorithm model based on artificial intelligence machine translation is shown in Figure 1.Among the traditional machine translation methods, the machine translation method based on similarity calculation is one of the earliest and its translation methods.It is the first time that people use computers to translate between two languages.First, multilingual machine translation adopts a unique model framework, which can reduce some deployment or training costs.Unified training of a model will bring about some knowledge sharing.Some rich languages can transfer some knowledge to some low-resource languages, which can improve the translation effect of low-resource language pairs.At the same time, due to multilanguage mixing, some low-resource language pairs can see some inputs that were not seen before, which can improve the generalization ability of low-resource languages to the model.It has a certain creativity in model architecture and algorithm ideas.'erefore, this paper takes this as an example to explore the tradition and its translation algorithms.Computer translation algorithm based on similarity mainly comes from the field of speech recognition.It looks for the best translation result by converting the editing distance between the two words to be translated.'e editing distance here refers to finding the best translation result of the content to be translated with the help of operations, such as insertion, replacement, and deletion between two words.Among them, different operation modes represent different translation costs.'erefore, we introduce the concept of word error rate Bleu.Generally, the algorithm definition expression of word error rate is shown in the following formulas.
However, because the algorithm is converted from the field of speech recognition, the way of machine recognition is still quite different from that of speech recognition in the actual process of machine translation.For some sounds or word positions, the laws followed are not completely consistent, so there may be errors in the position of words or the meaning of translated words in actual translation.On this basis, some scholars have proposed an improved algorithm based on the position of words.'e improved algorithm is less sensitive to the position of words in the translation process and abandons the past way of translating a paragraph of translated content according to a specific word order for the first time in the translation process but adopts an independent and continuous translation method.'at is, when translating a paragraph of content, the translation order of each word does not need to follow its order in the whole translation content.In this way, the degree of freedom of word translation is higher, the whole translation result is more flexible, and the accuracy rate has increased.'e specific improved algorithm expression is shown in the following formula.

R E T R A C T E D
complete consistency, partial consistency, and consistency.In addition, for some translation errors, we divide them into semantic errors, word order errors, part of speech errors.and so on.'e measurement of accuracy is mainly based on the conclusion of the evaluators.For example, if the evaluators think that the translation result is inaccurate, the accuracy of the translation result is completely consistent.If some evaluators think that there is an error in a certain segment of the translation content, but other translators think that the segment is used alone, but there is an error when combined with other segments, we call it partial consistency error.On the other hand, the translator thinks

R E T R A C T E D
that there is a semantic error in the translation; that is, if there is a semantic error in the translation, he thinks that there is no semantic error in the translation.'e artificial intelligence machine translation model is mainly based on the neural network algorithm, and its underlying logic is mainly completed by relying on the endto-end algorithm principle.Compared with the traditional neural network, which only analyzes the best answer through statistical information, artificial intelligence machine translation can use the self-learning function of neural network to continuously iterate and update in the model training and finally find the best result.However, the algorithm as a whole is more complex than machine translation and has a certain premise of use.Generally, we believe that there is a certain space for the probability change when the set X composed of the input language to be and the set Y composed of the target language to be translated are closed sets.See formulas ( 5) to (7) for specific set expression and probability evaluation expression.
After determining the applicable premise of artificial intelligence machine translation, the long-term and shortterm memory network architecture is mainly used in the specific training of the model.'e long-term and short-term memory network architecture usually takes the input layer as the coding segment and the output layer as the decoding end.In this way, we can train each translation result as many times as possible, effectively avoid the complex calculation process caused by artificial intelligence algorithm, and effectively improve the calculation efficiency.'e calculation formula of long-term and short-term memory network is shown in the following formula.
However, in the process of practical application, we found that the value trained according to the formula cannot transfer all the training information to the next neuron when it outputs the final value through the sigmoid activation function, which means that the current training information cannot be transmitted evenly, which will eventually lead to inaccurate results.'erefore, we decided to choose tanh function with more convergent derivative as the activation function, which will produce more new information in transmission than sigmoid.
After different training iterations according to this algorithm, the error value and efficiency change of artificial intelligence machine translation results are shown in Figure 3.It can be seen that with the increase of training times, each training iteration increases the accuracy of the model, and its error value will decrease.'at is, there is a training growth period during the operation of artificial intelligence machine translation.With the iteration of the model, the translation results will be better.
Finally, we evaluate the results of traditional machine translation and artificial intelligence machine translation.In order to better reflect the relationship between robot translation and artificial translation, here we use the artificial way to evaluate the results of traditional machine translation and artificial machine translation so as to pave the way for the follow-up exploration of the relationship between the two and artificial translation.We take the correlation coefficient as the evaluation method.
After determining the evaluation method, we evaluate the translation of 37 languages randomly selected from the commonly used English-Chinese human translation corpus, and take the recall rate and accuracy rate as the data indicators for the actual test.
Finally, we will sort out the evaluation results of traditional machine translation and artificial intelligence machine translation algorithms from the perspective of algorithm and model, as shown in Figure 4. Small probability events caused by random events should be avoided, which will affect the accuracy of the evaluation results.Small-probability events, that is, events with very small probability of occurrence (usually P ≤ or), have important applications in statistics.'at is, such events can occur in theory, but the probability of occurrence is small, and the probability of occurrence in this test is almost zero.For example, winning the lottery is a typical small-probability event.'ere may be a grand prize in each issue (the probability is very low), but for a lottery winner, there is almost no possibility that he will win the grand prize by buying a bet (the probability that a small probability event will occur in a test).In fact, this is an important theoretical basis for the application of small-probability events in statistics-the principle of small probability.'at is, the probability of a small-probability event occurring in a test is very small.If it does happen, statistics doubts its authenticity.'e conclusion of statistics based on the principle of Mobile Information Systems

R E T R A C T E D
small probability is very correct, but there is also the risk of making mistakes.We selected four groups of samples from the English-Chinese human translation corpus and tested them on different evaluation dimensions.'e results show that compared with the traditional robot translation, the machine based on artificial intelligence performs better in the accuracy of translation answers.In addition, it performs better in part of speech and word order.

Result Analysis and Discussion
'rough the above exploration, we have a preliminary understanding of the meaning of human translation and machine translation, as well as the underlying algorithm logic of traditional machine translation and artificial intelligence machine translation.In order to further explore the relationship between human translation and machine translation, we also compare and evaluate the differences, advantages, and disadvantages between traditional machine translation and artificial intelligence machine translation.'e advantages of traditional human translation include the following: the translator can interpret the context and convey the same meaning, rather than direct literal translation.'e translator can understand the creative use of language, such as puns, metaphors, slogans, and so on.Machine translation features short processing time and faster processing speed than human translation.One tool can complete the translation of multiple languages and master many more languages than human translation.But the accuracy is not guaranteed.After mastering the relevant basic knowledge, we will discuss the relationship between human translation and machine translation in this chapter and talk about the changes of the relationship between them in combination with the background of artificial intelligence machine translation.'e relationship between machine translation and human translation began with the birth of machine translation.It has been 63 years since the birth of machine translation in 1949.'e development of machine translation is inseparable from the development of computer industry.'erefore, if you want to further explore the relationship between other human translation, you should analyze it in combination with the trend background of computer development.When computers were not widely used, machine translation was just a written theory.Warren Weaver, as a consulting theory researcher, first proposed the concept of machine translation in 1949, but it was not well applied.At this time, the relationship between machine translation and human translation began to appear, but the relationship is weak.'e main translation method in the translation industry is still human translation.Until 1954, IBM and Georgetown University in the United States first integrated the translation function into the computer and invented the world's first IBM 701.Limited by the development of computers at that time, the translation machine was almost as large as the computer, with complex and clumsy operation.'e specific model is shown in Figure 5.
However, in the following decades, computer translation was still used, only affected by the rigid algorithm, multiple semantics, and word order errors.At this time, the traditional computer translation was not widely used by the people.'erefore, the translation work was still more manual translation.After that, with the development of computer artificial intelligence algorithms, machine translation has also ushered in great changes.With the complexity and accuracy of algorithms and the wide promotion of computer applications, artificial intelligence machine translation is being used more and more in the translation industry.See Figure 6 for details.
It can be seen that machine translation is applied in all walks of life, among which the health care industry accounts for the highest proportion.'is is because with the development of globalization and the development of national economy, more and more foreign products are imported

R E T R A C T E D
into China.'e translation of products is relatively fixed, so brands usually choose machine translation with high efficiency and low cost.On the whole, at the present stage, with the development of artificial intelligence translation, the relationship between artificial translation and machine translation has gradually changed from human-oriented to machine-oriented to human-oriented.In order to further determine the reliability of this relationship, we compared the accuracy of human translation and machine translation based on artificial intelligence.'e results show that there are few inconsistencies between artificial translation results and artificial intelligence machine translation results, and the inconsistencies are mainly maintained at 15-20%.It shows that although the translation accuracy and overall situation of machine translation based on artificial intelligence are not much different from that of human translation, the fundamental algorithm of machine translation is still a program for judging right and wrong through computer code, which cannot simulate the degree of "faithfulness, expressiveness and elegance" of human translation in combination with social background and humanistic culture, but for some mechanical operations such as business translation, scenes with low requirements such as common vocabulary in daily tourism can still meet the needs well.'erefore, under the influence of artificial machine translation, the relationship between the two is that machine translation can replace human translation in some aspects, but it cannot replace human translation (Figure 7).

Conclusion
'is paper discusses the influence of the development of artificial intelligence machine translation on the relationship between human translation and machine translation.'e advantages and disadvantages of traditional human translation include that the translator can interpret the context and convey the same meaning instead of direct translation.'e translator can understand the creative use of language, such as puns, metaphors, slogans, and so on.Compared with human translation, machine translation has the characteristics of short processing time and fast processing speed.One tool can complete the translation of multiple languages and master many more languages than human translation.However, accuracy cannot be guaranteed.'erefore, this paper reexamines the impact of artificial intelligence on the relationship between human translation and machine translation.Although the translation accuracy and overall situation of machine translation based on artificial intelligence are similar to those of human translation, the basic algorithm of machine translation is still a program that judges right and wrong through computer code.It cannot be combined with social background and human culture to simulate human translation.
However, there are some limitations in this research.With the application of artificial intelligence technology in different fields, its potential risks have also caused ethical dilemmas at different levels.'is will cause severe unemployment problems for the wide application of artificial intelligence in the future.

BLEU − 4 � 4 i�1 precision i . ( 3 )
min 1, output − length reference − length    When measured based on n-gram parameters, the formula can be transformed into BLEU − N �  s∈ ReferenceSunnaries { }  gram n ∈S Count match gram n   s∈ ReferenceSunnaries { }  gram n ∈S Count gram n  .(4) 'e change of the accuracy of the improved algorithm is shown in Figure 2. From the perspective of error accuracy, we divide the translation results into three categories: Mobile Information Systems

Figure 2 :
Figure 2: 'e number and consistency of errors in improved corpus translation.

Figure 3 :
Figure 3: Error value and accuracy change of machine translation based on artificial intelligence algorithm.

Figure 4 :
Figure 4: Comparison of model evaluation between traditional and artificial intelligence machine translation algorithms.

Figure 5 :
Figure 5: Traditional machine translation system and practical operation mode.