Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Application of Discriminative Training Algorithm Based on the Improved Gaussian Mixture Model in English Translation Evaluation

Application of Discriminative Training Algorithm Based on the Improved Gaussian Mixture Model in... Hindawi International Transactions on Electrical Energy Systems Volume 2022, Article ID 8021240, 11 pages https://doi.org/10.1155/2022/8021240 Research Article Application of Discriminative Training Algorithm Based on the Improved Gaussian Mixture Model in English Translation Evaluation Shengnan Wang Surrey International Institute, Dongbei University of Finance and Economics, Dalian 116025, Liaoning, China Correspondence should be addressed to Shengnan Wang; 1764100180@e.gzhu.edu.cn Received 30 June 2022; Revised 19 July 2022; Accepted 2 August 2022; Published 31 August 2022 Academic Editor: Raghavan Dhanasekaran Copyright © 2022 Shengnan Wang. ,is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. English translation, also a kind of language conversion, refers to the activities of expressing each other in English to express another language or another language to express English. ,is paper aimed to evaluate and research methods of English translation based on the improved Gaussian mixture model (GMM). ,e paper proposed a simple evaluation and analysis of English translation using a discriminative training algorithm. In the experimental part, on the one hand, taking students, teachers, and translators as case study objects, through a questionnaire survey, it can be known that the number of students who scored more than 80 points in the English translation test is increasing every year, from 11.30% in the first year to 21.13% in the fourth year, an increase of nearly 10%. On the other hand, it can be seen from the speech translation experiment that the correct translation rate was 46.18% by using intelligent technology to distinguish translation. ,e experimental results showed that the discriminative training algorithm under the improved GMM algorithm is effective in the research of English translation evaluation. GMM discriminative training algorithm to conduct evalua- 1. Introduction tion research in English translation. In the experimental part, the model was explored in depth, and the research results Due to the continuous progress of science and technology, show that the method is effective in the experiment. the English translation evaluation system based on artificial intelligence has been widely used in English teaching and testing. ,e computer gives the learners the objective scores 2. Related Work of English translation, so as to know whether their own English translation is standard. English translation assess- In recent years, with the continuous development of the ment is a wide-ranging research topic, including the learning economy, frequent trade exchanges between enterprises in mode of translation and the selection of teaching materials various countries, and a large population flow, the market for translation. ,is paper used the GMM to study English demand for English translation has continued to rise. translation assessment, which is helpful to understand the ,rough Pan’s research, using data from the Translation current social situation. It can help students to remove Corpus, a corpus of learners developed for the study of learning obstacles and help teachers to establish a correct lexical cohesion, through case studies, the results showed the translation teaching concept. complexity of a learner’s language and its relationship to ,is paper mainly studied the evaluation of English various factors such as learner, text, and task [1]. ,rough translation by mathematical models, which has a huge pos- Li’s research, some advantages of Hu’s ecological translation itive effect on the field of English translation in the future. ,e theory were explained, the translation was used to minimize innovation of this paper is mainly based on the improved translation problems, and corpus linguistics method was 2 International Transactions on Electrical Energy Systems used, which excels in quantitative and qualitative analysis English translation evaluation refers to the test and score [2]. Zhao et al. aimed to explore how narrative space can be after language translation. It is a specific expression of transferred from one language to another. Studies have translation ability by numbers or other standards that people shown that selective appropriation is the most commonly can generally understand, and it has scoring standards [12]. used framing strategy [3]. Si and Wang aimed to apply Translation assessment is required in many cases, such as grammatical metaphors from systematic functional lin- translation majors and employees engaged in translation guistics to translation studies. In order to achieve the work, both need to assess English translation ability. Figure 2 purpose of translation, an accurate and appropriate method shows scenes related to English translation evaluation. must be selected [4]. ,rough Tan Z’s and Ke research, the purpose was to analyze the formula of English-Chinese 3.2. Gaussian Mixture Model (GMM). Gaussian mixture translation thinking mode and explore the possible ad- model (GMM) aims to use Gaussian probability density vantages in translation practice [5]. However, due to the function to quantify things accurately [13]. ,e features of extensive use of language, English can be translated into each pixel in the image are represented by K Gaussian many languages, and relevant mathematical models are used models. After a new frame of image is obtained, the Gaussian for statistical analysis, and the above studies are all in-depth mixture model is updated, and each pixel in the current analysis of this. image is matched with the Gaussian mixture model. Figure 3 With the development of science and technology, is a GMM diagram. Gaussian mixture model is applied in many fields, and many GMM can be regarded as a continuous hidden Markov scholars have studied it. ,rough Yang J et al.’s research, model (CHMM) with a state of 1. It can better display the based on GMM, an efficient GNH approximation method probability distribution of multi-category observation data was proposed and used as a preconditioner for the mismatch in the sample space using GMM [14]. ,e probability density gradient to speed up its convergence [6]. ,rough the re- function of a K-order GMM is obtained by the weighted search of Gao G et al., a new method was proposed to in- summation of K Gaussian probability density functions: crease the complexity by adding a large number of Gaussian components and then check the accuracy of the GMM P(M|λ) � 􏽘 w b (M). (1) approximation of PDF [7]. ,rough Xu Y et al.’s research, i i i�1 the Gaussian mixture model was used for state estimation, and the main problem was that the number of Gaussian In (1), M represents a D-dimensional random vector; components increases exponentially [8]. After Han J et al.’s b (M), i � 1, 2, ..., K, is the sub-distribution; and w is the i i analysis, in order to obtain the effective user’s actual mixing weight. Each sub-distribution is a D-dimensional emotional state and promote a harmonious human-machine joint Gaussian probability distribution, which can be interactive experience, combined with emotional space and expressed as follows: personality theory, an incremental emotion mapping model 1 1 −1 based on Gaussian mixture model was proposed [9]. b (M) � exp􏼚− M − η􏼁 􏽘 M − η􏼁 􏼛. 􏼌 􏼌 i i i i D/2􏼌 􏼌1/2 􏼌 􏼌 2 (2π) 􏼌􏽐 􏼌 ,rough the research of Sagratella S, a new algorithm was proposed to calculate the approximate equilibrium of (2) generalized latent games with mixed integer variables. ,e Among them, η is the mean vector, 􏽐 is the covariance behavior of approximate equilibrium relative to exact i t matrix, (M − η ) is the transpose of the vector (M − η ), equilibrium was analyzed, and finally, the effectiveness of the i i −1 | 􏽐 | is the determinant, and 􏽐 is the inverse of the matrix method was demonstrated through numerous numerical 􏽐 . ,e mean vector η is the expected value of the ei- experiments [10]. However, the above scholars have studied t i genvector M, and the covariance matrix 􏽐 represents the the original GMM and have not conducted a thorough study cross-correlation and variance of the eigenvector elements. of the improved GMM. Let 3. Method of English Translation Assessment K 􏽘 w � 1. (3) Based on Discriminative Training Algorithm i�1 3.1. English Translation Assessment. English translation re- ,is ensures that the mixing density can represent a true fers to a way of language conversion, which includes Chi- probability density function [15]. ,erefore, the parameter nese-English translation, Japanese-English translation, set of the complete model θ contains the GMM mean η English-Korean translation, and translation between English covariance 􏽐 , and weight w ; namely, t i and other languages [11]. English translation also includes some translation skills, such as two basic methods of literal θ � 􏽮w , η , 􏽘 􏽯, i � 1, 2, ..., K. (4) i i t translation and meaning, as well as other techniques such as provincial translation, positive translation, and reverse GMM can be explained; it is a functional expression of translation. English translation can be used in many occa- probability density function. As a linear combination of sions, such as warning signs and learning courseware. ,e Gaussian probability density functions, GMM can ap- specific situation is shown in Figure 1, which is an illus- proximate any density function as long as there are a suf- tration related to English translation. ficient number of mixed components [16]. Language International Transactions on Electrical Energy Systems 3 Figure 1: Diagrams related to English translation. Figure 2: Scenarios related to English translation assessment. translation usually has a smooth probability density func- parameter estimates of the GMM in iterations, increasing the tion, so a finite Gaussian density function is sufficient to matching probability of the model estimate θ and the observed form a smooth approximation to the density function of feature vector at each iteration, and at each iteration, there is k+1 k English translation features [17]. P(|M|θ )> P(|M|θ ), where k is the number of iterations. ,e method used to estimate the GMM parameters is the Figure 4 is a step diagram of the expectation maximi- maximum likelihood estimation method; that is, with respect zation algorithm. to θ, the conditional probability P(M|θ) is maximized [18]. It is assumed that feature vectors are independent of each ,e expectation maximization (EM) algorithm improves the other; there are 4 International Transactions on Electrical Energy Systems 300 w b M i t t P i|M , θ􏼁 � . (11) 􏽐 w b M􏼁 j�1 j j t ,en, we change the symbols in (8), (9), (10), and (11), and the iterative algorithm expression of θ can be obtained as follows: T−1 k+1 k w � 􏽘 P i||M , θ , (12) 􏼐 􏼑 i t t�0 T−1 k P i|M , θ M 􏽐 􏼐 􏼑 t�0 t t k+1 η � , (13) T−1 k 􏽐 P􏼐i|M , θ 􏼑 t�0 t T−1 k T k+1 􏽐 P􏼐i|M , θ 􏼑 M − η􏼁 M − η􏼁 t�0 t t i t i 􏽘 � T−1 k 􏽐 P􏼐i|M , θ 􏼑 0 t t�0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 (14) T−1 k T 􏽐 P􏼐i|M , θ 􏼑M M t t t t�0 k+1 k+1 Figure 3: Gaussian mixture model fitted image histogram. � − η 􏼐η 􏼑 . t t T−1 k 􏽐 P􏼐i|M , θ 􏼑 t�0 t In (14), the superscript T in the covariance matrix represents the transpose of the matrix. T−1 Among them, P(|M|θ) � 􏽙 P(|M|θ). (5) t�0 k k w b M i i t P􏼐i|M , θ 􏼑 � . (15) It is assumed that P(M |θ) is differentiable to θ; when K k k 􏽐 w b M􏼁 j�1 i i P(M |θ) takes the maximum value, it satisfies the following: Among them is the mixture component of the ith d[InP(|M|θ)] (6) g � � 0. Gaussian probability density function in the kth iteration. dθ In a given sample of translated speech, the purpose of ,rough (5) and (6), the estimated iterative algorithm is identifying the language is to determine which language the obtained. speech belongs to [19]. ,e block diagram of the language Take η as an example: i recognition system based on GMM is shown in Figure 5. By Bayesian theory, maximizing the verification prob- d[InP(|M|θ)] g � ability can be expressed as follows: dη P M|θ􏼁 P θ􏼁 i i T−1 (16) P θ |M􏼁 � . d w P M , i|θ i 􏽨􏽐 􏼁 􏽩 i�1 i t − 1 P(M) � 􏽘 P(|M|θ) (7) dη t�0 Among them, T−1 −1 T−1 − 1 � 􏽘 P M |θ􏼁 w P M , i|θ􏼁 􏽘 M − η􏼁 � 0. P M|θ􏼁 � 􏽙 P M|θ􏼁 . (17) t i t t i i t�0 t t�0 ,en, we get the following: Its logarithmic form is as follows: T−1 T−1 􏽐 P i|M , θ􏼁 M t t t�0 η � . (8) InP X|θ􏼁 � 􏽘 InP X |θ􏼁 . (18) T−1 i t i 􏽐 P i|M , θ􏼁 t�0 t�0 Similarly, we can also get the following: Because of the prior probability of θ , T−1 (19) P θ􏼁 � , 1≤ i≤ . w � P i|M , θ , (9) 􏽘 􏼁 i i t t�0 For a certain eigenvector M, P(M) is a certain constant, T−1 equal for all languages. ,erefore, the maximum value of the 􏽐 p i|M , θ􏼁 M − η􏼁 M − η􏼁 t�0 t t t t t � . (10) T−1 posterior probability can be obtained by calculating P(M|θ ) 􏽐 p i|M , θ􏼁 t�0 t so that the recognized language in the phonetic database can be expressed as follows: In (10), International Transactions on Electrical Energy Systems 5 Start Initialization parameters Expectation step Computational Parameter Changes distribution Re-estimate Maximization parameters step Figure 4: Expectation maximization algorithm steps. Feature Input Language Extraction Front-end Language Recognizer Processing Language Model Most LikelyLanguage Figure 5: Language recognition system based on GMM. X N r rx i � arg max P M|θ . (20) G (Γ) � 􏽘 􏽘 􏽘 P w (n)|Q ς w (Q) � t . 􏼁 􏼁 GR θ rx rx rx rx ∗ r�1 x�1 n−1 ,e i here represents the recognized language. (21) Among them, Γ represents the parameter set of the 3.3. Discriminative Training Algorithm. Discriminative hidden Markov model; t is the text of the xth word in the rx training is proposed to improve language recognition rate. rth training sentence; and the posterior probability is as Different from the maximum likelihood algorithm that follows: emphasizes the training data, the language model based on L Q |w (n) 􏼁 P w (n)|w (Q)􏼁 the discriminative training algorithm emphasizes the opti- θ rx rx rx rx P w (n)|Q � . θ rx rx mal discrimination of the training data [20]. After several 􏽐 L Q |v􏼁 P v|w (Q)􏼁 v θ rx rx consecutive recognitions, similar languages can be distin- (22) guished by using this algorithm, and the language model based on the discriminative training algorithm is more Among them, Q is the speech output vector of the xth rx conducive to improving the recognition rate of test data than word in the rth training sentence; v is the possible wrong the traditional maximum likelihood algorithm. Figure 6 is a translation given w (Q); and L (Q |v) represents the rx θ rx scene graph related to the discriminative training algorithm. likelihood probability function of Q given v. rx For a corpus with N training sentences, label each Similarly, “misaccepted” means that although the sentence with r � 1, 2, ..., R. Its objective function is defined identification is wrong, it is wrongly regarded as a correct as its mathematical expectation: judgment; its objective function is as follows: Feature Vector 6 International Transactions on Electrical Energy Systems Figure 6: Scenario diagram related to the discriminative training algorithm. R r 4. Experiments in English Translation (23) G (θ) � 􏽘 􏽘 P w (Q)|Q 􏼁 ς w (Q)≠ t 􏼁. GΓ θ rx rx rx rx Assessment Based on Gaussian r�1 x�1 Mixture Model Corresponding to the correct diagnosis under the true 4.1. Evaluation Plan for English Translation. By collecting the rejection category, the mathematical expectation of the existing research results in the field of English translation, wrong diagnosis is as follows: the importance of language translation in various fields can X N r rx be further understood. ,is paper randomly selected 500 G (θ) � 􏽘 􏽘 􏽘 P w (n)|Q DE θ rx rx students majoring in English translation, 200 teachers (24) r�1 x�1 n�1 majoring in English translation, and 300 white-collar · ς w (Q)≠ t , w (n)≠ t 􏼁. rx rx rx rx workers who work in English translation as the research objects. A total of 1,000 copies of the “English Translation In order to minimize the error, consider the new ob- Assessment Questionnaire” were distributed, and 979 copies jective function G(θ) obtained by summing the three error were recovered. ,ere were 972 valid questionnaires, and the objective functions. When the scaling factor k � 1, the effective recovery rate was 97.2%. derivation result is as follows: Among the valid questionnaires collected, there are 478 students, accounting for about 49.18% of the total number; G(θ) � G (θ) + G (θ) + G (θ) GR GΓ DE 196 teachers, accounting for about 20.16% of the total X N r rx number; and 298 white-collar workers majoring in English � 􏽘 􏽘 􏽘 P w (n)|Q 􏼁 ς w (Q) � t 􏼁 (25) θ rx rx rx rx translation, accounting for about 30.66% of the total. r�1 x�1 n�1 In this questionnaire, a total of 4 questions about English translation were raised, namely, as to whether translation � 􏽘 􏽘 P s|Q 􏼁 Raw Word Error s, t 􏼁 . r�1 S θ r r software is used in daily study or work, opinions on using Among them, Q is the observation vector; t is the translation software, self-evaluation of English translation, and r r translated text of a given text sentence; s is the competing translation proficiency test situation in daily work or study. text identified according to Q ; and the function Raw Word Error(s, t ) is the number of mismatches between 4.2. Evaluation Results of English Translation s and t translations. L Q |s􏼁 P s|t 􏼁 4.2.1. Questionnaire θ r r P s|Q 􏼁 � . (26) θ r 􏽐 L Q |h􏼁 P h|t 􏼁 h θ r r (1) Whether Translation Software is Used in Daily Study or In (26), h is any possible wrong translation given t . Work. Translation software is created to facilitate people’s r International Transactions on Electrical Energy Systems 7 Table 1: Frequency of using translation software in daily life. daily work and study. Even people with high achievements in English translation will have unskilled words, sentences, Corporate white- Students Teachers etc. and will also use translation software [21]. Table 1 shows collar workers the frequency of using translation software in the daily life of Used frequently 386 6 81 the three types of people. Occasionally used 58 111 159 It can be seen from Table 1 that in daily work and life, 386 Hardly ever used 34 79 58 students frequently use translation software, accounting for about 80.75%; 58 students use the software occasionally, accounting for about 12.13%; and 34 students hardly use Table 2: Opinions of three categories of people on translation software. translation software, accounting for about 7.12%. It can be seen from this set of data that most of the students’ Corporate white- Students Teachers translation foundation is still relatively weak, they are still collar workers studying, and relying on translation software is still relatively Very convenient 405 56 146 high. From this part of the data of teachers, it can be found Having pros and cons 40 93 150 that there are 79 teachers who hardly use translation soft- Very bad 33 47 2 ware, accounting for about 40.31%; 111 who use it occa- sionally, accounting for about 56.63%; only 6 who use the software regularly, accounting for about 3.06%. From this that students' use of translation software may make them dependent on it, so that they will not memorize words. part, it can be seen that the teachers’ job is to contact the students, and their English foundation is solid, so they are Learning English is limited to learning knowledge from translation software, which is bad for both teachers and less dependent on translation software. However, there are students. It can be seen from the data of enterprise white- still unfamiliar words or sentences in daily work, and collar workers that 146 people, which account for about translation software is also needed, so the number of people 48.99%, think this kind of software is very convenient, who use it occasionally accounts for more percentage. because they encounter too many problems in their daily Looking at corporate white-collar workers, the English work. ,e use of translation software not only helps them translation ability of this kind of staff is often exercised in their daily work, so their translation ability is strong; solve problems but also saves them a lot of time. And 150 people think that translation software is a double-edged therefore, 58 workers hardly use translation software, ac- counting for about 19.46%. However, due to too many sword, their reasons are the same as those of the students, and they account for about 50.34%. unexpected situations at work, 81 use the software frequently and 159 use it occasionally, accounting for 27.18% and (3) Self-Evaluation of English Translation. As the saying goes, 53.36%, respectively. the person who knows oneself best is oneself. Although the evaluation of one’s English translation ability includes (2) Opinions on the Use of Translation Software. Translation subjective factors, it also has certain reference value [23]. software has advantages and disadvantages in work and Table 3 shows the self-evaluation of the English translation study. ,e advantage is that it facilitates work and study, ability of the three categories of people. facilitates word search, and saves time. ,e disadvantage is that frequent use will rely on it, resulting in lazy emotions From the data in Table 3, 117 students in this group of data, accounting for about 24.48%, think their English and reluctance to learn new words [22]. Table 2 shows three types of respondents’ opinions on translation software. translation ability is relatively good; 238 students think their English translation ability is average, and the proportion is It can be seen from Table 2 that 405 students think that about 49.79%; and 123 students think their translation level the translation software is very convenient and greatly saves still needs to be improved, and the proportion of this group time. In daily learning, there is no need to look up dictio- is about 25.73%. ,e biggest reference for students’ self- naries or do direct online search. ,ose students account for evaluation is their usual assessment scores. Better grades are about 84.73%. 40 students maintained a neutral attitude, considered excellent, average grades are considered ordi- thinking that translation software has both an advantage and disadvantage. ,e advantage is that it is convenient to query nary, and poor grades are considered very poor. It can be seen from the data of teachers that 97, accounting for about vocabulary and quickly solve translation problems. ,e disadvantage is that this software may replace the English 49.49%, think their translation ability is excellent; 96 think that their translation ability is average, and the proportion is translation major, which is a bad phenomenon for people with this major. ,ose students account for 8.37%. And 33 about 48.98%; and some teachers think that their translation ability still needs to be improved. ,is happens mainly think that the software is very bad. Many words have too because teachers evaluate themselves from their own many translation results on the software, and it is difficult to teaching ability. Among the white-collar workers in the third distinguish which translation is the answer they need. ,ose part, 156, accounting for about 52.35%, think their trans- students account for about 6.90%. From the data of teachers, lation ability is relatively good; 140, accounting for about it can be seen that 56 teachers, accounting for about 28.57%, 46.98%, think that their English translation level is average; think that translation software is very convenient. 93 people think that the software has an advantage and disadvantage; of course, some people think that their translation level is very poor, and those account for about 0.67%. ,e main the advantage is that it saves time, and the disadvantage is 8 International Transactions on Electrical Energy Systems Table 3: Self-evaluation of English translation ability of three 227 227 categories of people. 209 210 Students Teachers Corporate white-collar workers Excellent 117 97 156 150 Ordinary 238 96 140 100 90 Very poor 123 3 2 reason for this kind of situation is that the daily work of the company is both difficult and easy. According to this part of Freshman Sophomore Junior year Senior year the data, it can be concluded that the ability to subjectively Years judge oneself cannot be limited to only one condition. It is Below 60 points 70-80 recommended that many aspects are considered, so that the 60-70 80 points or more possibility of subjective judgment is smaller. Figure 7: Students’ final test scores for the past four years. (4) Translation Proficiency Test in Daily Work and Study. Whether students or staff, in order to objectively understand their English translation level, they need to be tested. Stu- dents’ tests are mainly daily tests, as shown in Figure 7, which compares and analyzes the final exams of students of the same major in the past four years. ,e teachers' test is 100 mainly about the passing rate of English proficiency, as 89 shown in Figure 8, which is based on the passing rate after 80 graduation. ,e test of corporate white-collar workers is mainly about the accuracy of translation at work and the time to translate a paragraph. ,e specific situation is shown in Figure 9. Figure 7 shows the distribution map of the number of students in each grade of final grades in the past four years. 1 1 It can be seen from Figure 7 that in the past four years, 2018 2019 2020 2021 the number of students in low grades has decreased year by Years year, and the number of students in high grades has in- pass the test creased year by year. Among them, the number of people Excellent grades Failed grades below 60 was decreasing year by year, from 38 in the Failed grades freshman year to 3 in the senior year, and the population Figure 8: ,e number of teachers who passed the English below 60 in the English translation subject test was de- translation proficiency test. creasing. ,e number of students with scores above 80 was increasing every year, from 54 in the first year, accounting for 11.30% of the total, to 101 in the fourth year, accounting for 21.13% of the total, an increase of nearly 10%. ,is showed that with the deepening of learning content, the level 70.00 66.44% of translation of students is gradually rising. 62.42% 59.40% Figure 8 shows the number of teachers who have passed 60.00 54.03% the English translation proficiency test in the past four years. 50.00 It can be seen from Figure 8 that in the past four years, teachers have participated in the English translation profi- 40.00 30.87% ciency test with a high pass rate, but there are also people 28.18% 26.85% 30.00 who fail. ,e number of teachers who pass and perform well 19.46% 19.12% also increased every year. From 2018 to 2021, the pass rate 20.00 14.10% was above 98%. ,e lowest year was 2019, and the pass rate 9.73% 9.40% 10.00 was also 98.47%. Among them, the number of teachers who have passed with excellent grades has increased year by year. 0.00 2018 2019 2020 2021 In 2018, there were 92 teachers, accounting for about Years 46.94%. Looking at 2021, we find that the number was 107, accounting for more than 50%. ,is showed that with the pass the test Excellent grades increase of working years, the professional level of these Failed grades teachers also maintains an upward trend. Figure 9 shows a survey of corporate white-collar Figure 9: Enterprise white-collar workers’ survey of their trans- workers’ translation capabilities in the past four years. lation capabilities in the past four years. Proportion (%) Quantity Quantity International Transactions on Electrical Energy Systems 9 Table 4: Confusion matrix of the number of 1-frame speech tests. As can be seen from Figure 9, the test requirements of enterprises for English translators have increased year by Recognition result Sample size year. It can be seen from the data that the failure rate in the Japanese Chinese German French four years from 2018 to 2021 has almost doubled from 14.1% Japanese 98 27 51 7 21 to 28.18%. Although there were small changes in the middle, Chinese 96 28 38 15 15 the overall failure rate was on the rise. Of course, during this German 90 10 29 36 15 period, there are still many workers who are constantly French 88 29 19 31 9 improving their strength and professional level. In the work test, their scores are also improving year by year. Table 5: Confusion matrix for number of 2-second speech tests. Recognition result 4.2.2. Language Recognition Experiment Based on Gaussian Japanese Chinese German French Mixture Model. English translation includes not only text translation, but also voice translation. ,e first part of the Japanese 0.5578 0.0938 0.2263 0.1221 Chinese 0.080 0.6612 0.229 0.0298 questionnaire survey was mainly based on text translation, German 0.0779 0.2335 0.5321 0.1565 and then experiments were conducted on voice translation. French 0.09 0.05 0.2111 0.6489 In this experiment, four languages were selected for translation results experiment, which are Japanese, Chinese, German, and French. ,e total number of experimental the sample size is small, and the results are not accurate; it is samples is 400 8-second speeches and 400 2-second recommended that the sample size is increased. Second, in speeches. After processing through feature parameters, the the speech translation experiment, the sample data is large, dimension of each feature vector is 16, and the mixing but the selected sample time period is short, and the sample degree of GMM is 32. Experiments were conducted with selection source is single too. By increasing the sample speech with a length of one frame and a speech with a length sources, the experimental data can be more complete and the of 2 seconds (188 frames). ,e details are shown in Table 4. experiments can be comparable. ,e application of the It can be seen from Table 4 and Table 5 that only the first discriminative training algorithm based on the improved frame of each 2-second sample was taken to test the Gaussian mixture model in the English page-turning pro- translation effect and then the sample was tested. ,e rec- fession is a field that experts and scholars from all walks of ognition results obtained are shown in Table 5, and the life have been keen in recent years, and many issues still need correct translation rate was only 46.18%. ,is first-order further research and discussion. polynomial contains the information between frames, but only one frame, with 2 seconds of speech segment, was used 5. Discussion for sample testing, and its translation recognition result was not high. With the development of economic globalization, the de- To sum up, through the questionnaire survey and GMM mand for English translation professionals is also increasing, experiment, the English translation ability of the three and the requirements for this profession will also increase. groups of people and the language translation technology of To understand the basic level of this profession, it also needs the GMM can be learned. In this regard, there are still some to be tested [24]. ,is paper firstly integrated the achieve- small problems in the experiment that need to be improved, ments and ideas of English translation professionals in the and the evaluation of the English test needs to be carefully major and used the discriminative training algorithm of analyzed. In terms of written translation and oral transla- Gaussian mixture model to discuss the problems of 1,000 tion, the results are diametrically opposite. English translation professionals. ,en, it used the GMM to study the related problems of English translation in speech translation. Finally, through a series of analyses, two op- 4.3. Evaluation Results of English Translation by Improved posite results were obtained. In the process of written Gaussian Mixture Model. According to the analysis of this questionnaires, it was found that the English translation paper, the application of the discriminative training algo- rithm of GMM in English translation is relatively perfect. ability of all kinds of people was relatively strong, but in the voice translation, their translation ability was still weak. With the advancement of science and technology, intelligent translation technology has gradually been recognized and ,is paper is devoted to researching the related application of the discriminative training algorithm based on GMM and used by the public. ,e use of GMM’s discriminative applying it in the field of English translation evaluation. ,is is training algorithm to collect the opinions of English not only an expansion and extension of the application scope of translation professionals on the profession greatly facilitates the algorithm, but also a new attempt to evaluate English the research of this experiment. However, because the re- translation. ,rough the above case studies, opposite experi- search in this field is not comprehensive enough, there are mental results are obtained in the written translation experi- some small problems, which need to be corrected. ,e details are as follows: First, in the questionnaire survey, three ment and the speech translation experiment, which both show the effectiveness of the discriminative training algorithm in this groups of people were selected for experimental investiga- tion; the results are better, but in this part of the experiment, paper under the Gaussian mixture model. 10 International Transactions on Electrical Energy Systems wavenumber filter and its applications in least-squares seismic 6. Conclusions imaging,” IEEE Transactions on Geoscience and Remote Sensing, vol. 60, no. 99, pp. 1–13, 2022. In this paper, the discriminative training algorithm of GMM [7] G. Gao, H. Jiang, J. C. Vink, C. Chen, Y. El Khamra, and was used as the main analysis method; 1,000 randomly J. J. Ita, “Gaussian mixture model fitting method for uncer- selected people of three types were selected as the research tainty quantification by conditioning to production data,” object; and the related problems of students, teachers, and Computational Geosciences, vol. 24, no. 2, pp. 663–681, 2019. translators in English translation were selected. ,rough the [8] Y. Xu, Y. Fang, W. Peng, and Y. Wu, “An efficient Gaussian case study, the following conclusions were drawn: ,ere are sum filter based on prune-cluster-merge scheme,” IEEE Ac- subjective evaluation and objective evaluation in English cess, vol. 7, pp. 150992–151005, 2019. translation professional test; the subjective test needs to [9] J. Han, L. Xie, Z. Wang, and F. Renl, “Incrementally emotion comprehensively consider many aspects and cannot deter- mapping based on GMM,” Harbin Gongye Daxue Xuebao/ mine one’s translation ability only by one factor, and the Journal of Harbin Institute of Technology, vol. 50, no. 8, same is true for the objective evaluation. ,e translation pp. 168–173, 2018. includes voice translation and written translation. It can be [10] S. Sagratella, “Algorithms for generalized potential games seen from the experiment that the written translation survey with mixed-integer variables,” Computational Optimization and Applications, vol. 68, no. 3, pp. 689–717, 2017. results are better. All kinds of people are good at written [11] L. Zhou, Y. Zheng, L. Huo, Y. Ye, X. Wang, and G. Song, translation, while voice translation is the use of intelligent “Fracture behaviors of HVFA-SCC mixed with seawater and technology to memorize translation, and the result has sea-sand under three-point bending,” Advances in Structural certain defects. From one angle of experiment, the reliability Engineering, vol. 25, no. 4, pp. 716–735, 2022. of the results is low, so it needs to be analyzed from multiple [12] M. V. Farahani, “Review of feng (2020): form, meaning and angles and multiple aspects to better highlight the efficient function in collocation: a corpus study on commercial Chi- application of GMM’s discriminative training algorithm in nese-to-English translation,” International Journal of Corpus English translation evaluation. Linguistics, vol. 27, no. 2, pp. 254–258, 2022. [13] S. R. Sweger, S. Pribitzer, and S. Stoll, “Bayesian probabilistic Data Availability analysis of DEER spectroscopy data using parametric distance distribution models,” =e Journal of Physical Chemistry A, ,e data that support the findings of this study are available vol. 124, no. 30, pp. 6193–6202, 2020. from the author upon reasonable request. [14] Z. Liu, S. Li, and K. Zhao, “Extended multi-interval Legendre- Gauss-Radau pseudospectral method for mixed-integer op- timal control problem in engineering,” International Journal Conflicts of Interest of Systems Science, vol. 52, no. 5, pp. 928–951, 2020. [15] S. Rupesh Kumar and B. Bharathi, “Generative and dis- ,e author declares no potential conflicts of interest with criminative modelling of linear energy sub-bands for spoof respect to the research, authorship, and/or publication of detection in speaker verification systems,” Circuits, Systems, this article. and Signal Processing, vol. 41, no. 7, pp. 3811–3831, 2022. [16] Y. Chen, Y. Chen, and B. Yu, “Speed distribution prediction of References freight vehicles on mountainous freeway using deep learning methods,” Journal of Advanced Transportation, vol. 2020, [1] J. Pan, T. M. Wong, and H. Wang, “Navigating learner data in no. 3, pp. 1–14, 2020. translator and interpreter training:insights from the Chinese/ [17] G. D. Barriga, G. M. Cordeiro, D. K. Dey, V. G. Cancho, English translation and interpreting learner corpus F. Louzada, and A. K. Suzuki, “,e Marshall-Olkin gener- (CETILC),” Babel, vol. 68, no. 2, pp. 236–266, 2022. alized gamma distribution,” Communications for Statistical [2] Q. J. Li, R. R. Wu, and Y. Ng, “Developing culturally effective Applications and Methods, vol. 25, no. 3, pp. 245–261, 2018. strategies for Chinese to English geotourism translation by [18] C. Li, G. Casella, and M. Ghosh, “Estimation of regression corpus-based interdisciplinary translation analysis,” Geo- vectors in linear mixed models with Dirichlet process random heritage, vol. 14, no. 1, pp. 1–24, 2022. effects,” Communications in Statistics - =eory and Methods, [3] M. Zhao, L. H. Ang, S. Md Rashid, and F. H. C. Toh, vol. 47, no. 16, pp. 3935–3954, 2018. “Translating narrative space in children’s FictionBronze and [19] Y. Zhang and X. Chen, “,e deformed gesture tracking al- SunflowerFrom Chinese to English,” Sage Open, vol. 12, no. 1, gorithm based on feature space segmentation modeling,” Jiqiren/Robot, vol. 40, no. 4, pp. 401–412, 2018. [4] X. Si and J. Wang, “Grammatical metaphor in English-Chi- [20] I. J. David, O. E. Asiribo, and H. G. Dikko, “Nonlinear split- nese translation,” International Journal of Translation Inter- plot design model in parameters estimation using estimated pretation and Applied Linguistics, vol. 3, no. 1, pp. 15–29, 2021. [5] Z. Tan and X. Ke, “On EST sentence translation based on generalized least square-maximum likelihood estimation,” ComTech Computer Mathematics and Engineering Applica- English and Chinese structures: natural linear expansion vs. Reversed linear expansion,” =eory and Practice in Language tions, vol. 9, no. 2, pp. 65–71, 2018. [21] E. M. M. Ortega, G. M. Cordeiro, A. K. Suzuki, and Studies, vol. 11, no. 4, pp. 396–402, 2021. [6] J. Yang, J. Huang, Z. Li, H. Zhu, G. A. McMechan, and X. Luo, T. G. Ramires, “A new extended Birnbaum-Saunders model “Approximating the gauss-Newton hessian using a space- with cure fraction: classical and Bayesian approach,” International Transactions on Electrical Energy Systems 11 Communications for Statistical Applications and Methods, vol. 24, no. 4, pp. 397–419, 2017. [22] G. Xie, Y. Wu, H. Gong, C. Li, and W. Zhang, “Data pro- cessing in residual stress measurements using contour method,” China Mechanical Engineering, vol. 28, no. 15, pp. 1778–1783, 2017. [23] L. Ko, “Chinese certificate translation in the Australian context:A purpose-oriented practice,” Babel, vol. 68, no. 1, pp. 24–54, 2022. [24] C. Huang, “Analysis of common errors in financial English- Chinese translation,” OALib, vol. 08, no. 01, pp. 1–7, 2021. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png International Transactions on Electrical Energy Systems Hindawi Publishing Corporation

Application of Discriminative Training Algorithm Based on the Improved Gaussian Mixture Model in English Translation Evaluation

Loading next page...
 
/lp/hindawi-publishing-corporation/application-of-discriminative-training-algorithm-based-on-the-improved-0gobn0z1EL
Publisher
Hindawi Publishing Corporation
Copyright
Copyright © 2022 Shengnan Wang. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
eISSN
2050-7038
DOI
10.1155/2022/8021240
Publisher site
See Article on Publisher Site

Abstract

Hindawi International Transactions on Electrical Energy Systems Volume 2022, Article ID 8021240, 11 pages https://doi.org/10.1155/2022/8021240 Research Article Application of Discriminative Training Algorithm Based on the Improved Gaussian Mixture Model in English Translation Evaluation Shengnan Wang Surrey International Institute, Dongbei University of Finance and Economics, Dalian 116025, Liaoning, China Correspondence should be addressed to Shengnan Wang; 1764100180@e.gzhu.edu.cn Received 30 June 2022; Revised 19 July 2022; Accepted 2 August 2022; Published 31 August 2022 Academic Editor: Raghavan Dhanasekaran Copyright © 2022 Shengnan Wang. ,is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. English translation, also a kind of language conversion, refers to the activities of expressing each other in English to express another language or another language to express English. ,is paper aimed to evaluate and research methods of English translation based on the improved Gaussian mixture model (GMM). ,e paper proposed a simple evaluation and analysis of English translation using a discriminative training algorithm. In the experimental part, on the one hand, taking students, teachers, and translators as case study objects, through a questionnaire survey, it can be known that the number of students who scored more than 80 points in the English translation test is increasing every year, from 11.30% in the first year to 21.13% in the fourth year, an increase of nearly 10%. On the other hand, it can be seen from the speech translation experiment that the correct translation rate was 46.18% by using intelligent technology to distinguish translation. ,e experimental results showed that the discriminative training algorithm under the improved GMM algorithm is effective in the research of English translation evaluation. GMM discriminative training algorithm to conduct evalua- 1. Introduction tion research in English translation. In the experimental part, the model was explored in depth, and the research results Due to the continuous progress of science and technology, show that the method is effective in the experiment. the English translation evaluation system based on artificial intelligence has been widely used in English teaching and testing. ,e computer gives the learners the objective scores 2. Related Work of English translation, so as to know whether their own English translation is standard. English translation assess- In recent years, with the continuous development of the ment is a wide-ranging research topic, including the learning economy, frequent trade exchanges between enterprises in mode of translation and the selection of teaching materials various countries, and a large population flow, the market for translation. ,is paper used the GMM to study English demand for English translation has continued to rise. translation assessment, which is helpful to understand the ,rough Pan’s research, using data from the Translation current social situation. It can help students to remove Corpus, a corpus of learners developed for the study of learning obstacles and help teachers to establish a correct lexical cohesion, through case studies, the results showed the translation teaching concept. complexity of a learner’s language and its relationship to ,is paper mainly studied the evaluation of English various factors such as learner, text, and task [1]. ,rough translation by mathematical models, which has a huge pos- Li’s research, some advantages of Hu’s ecological translation itive effect on the field of English translation in the future. ,e theory were explained, the translation was used to minimize innovation of this paper is mainly based on the improved translation problems, and corpus linguistics method was 2 International Transactions on Electrical Energy Systems used, which excels in quantitative and qualitative analysis English translation evaluation refers to the test and score [2]. Zhao et al. aimed to explore how narrative space can be after language translation. It is a specific expression of transferred from one language to another. Studies have translation ability by numbers or other standards that people shown that selective appropriation is the most commonly can generally understand, and it has scoring standards [12]. used framing strategy [3]. Si and Wang aimed to apply Translation assessment is required in many cases, such as grammatical metaphors from systematic functional lin- translation majors and employees engaged in translation guistics to translation studies. In order to achieve the work, both need to assess English translation ability. Figure 2 purpose of translation, an accurate and appropriate method shows scenes related to English translation evaluation. must be selected [4]. ,rough Tan Z’s and Ke research, the purpose was to analyze the formula of English-Chinese 3.2. Gaussian Mixture Model (GMM). Gaussian mixture translation thinking mode and explore the possible ad- model (GMM) aims to use Gaussian probability density vantages in translation practice [5]. However, due to the function to quantify things accurately [13]. ,e features of extensive use of language, English can be translated into each pixel in the image are represented by K Gaussian many languages, and relevant mathematical models are used models. After a new frame of image is obtained, the Gaussian for statistical analysis, and the above studies are all in-depth mixture model is updated, and each pixel in the current analysis of this. image is matched with the Gaussian mixture model. Figure 3 With the development of science and technology, is a GMM diagram. Gaussian mixture model is applied in many fields, and many GMM can be regarded as a continuous hidden Markov scholars have studied it. ,rough Yang J et al.’s research, model (CHMM) with a state of 1. It can better display the based on GMM, an efficient GNH approximation method probability distribution of multi-category observation data was proposed and used as a preconditioner for the mismatch in the sample space using GMM [14]. ,e probability density gradient to speed up its convergence [6]. ,rough the re- function of a K-order GMM is obtained by the weighted search of Gao G et al., a new method was proposed to in- summation of K Gaussian probability density functions: crease the complexity by adding a large number of Gaussian components and then check the accuracy of the GMM P(M|λ) � 􏽘 w b (M). (1) approximation of PDF [7]. ,rough Xu Y et al.’s research, i i i�1 the Gaussian mixture model was used for state estimation, and the main problem was that the number of Gaussian In (1), M represents a D-dimensional random vector; components increases exponentially [8]. After Han J et al.’s b (M), i � 1, 2, ..., K, is the sub-distribution; and w is the i i analysis, in order to obtain the effective user’s actual mixing weight. Each sub-distribution is a D-dimensional emotional state and promote a harmonious human-machine joint Gaussian probability distribution, which can be interactive experience, combined with emotional space and expressed as follows: personality theory, an incremental emotion mapping model 1 1 −1 based on Gaussian mixture model was proposed [9]. b (M) � exp􏼚− M − η􏼁 􏽘 M − η􏼁 􏼛. 􏼌 􏼌 i i i i D/2􏼌 􏼌1/2 􏼌 􏼌 2 (2π) 􏼌􏽐 􏼌 ,rough the research of Sagratella S, a new algorithm was proposed to calculate the approximate equilibrium of (2) generalized latent games with mixed integer variables. ,e Among them, η is the mean vector, 􏽐 is the covariance behavior of approximate equilibrium relative to exact i t matrix, (M − η ) is the transpose of the vector (M − η ), equilibrium was analyzed, and finally, the effectiveness of the i i −1 | 􏽐 | is the determinant, and 􏽐 is the inverse of the matrix method was demonstrated through numerous numerical 􏽐 . ,e mean vector η is the expected value of the ei- experiments [10]. However, the above scholars have studied t i genvector M, and the covariance matrix 􏽐 represents the the original GMM and have not conducted a thorough study cross-correlation and variance of the eigenvector elements. of the improved GMM. Let 3. Method of English Translation Assessment K 􏽘 w � 1. (3) Based on Discriminative Training Algorithm i�1 3.1. English Translation Assessment. English translation re- ,is ensures that the mixing density can represent a true fers to a way of language conversion, which includes Chi- probability density function [15]. ,erefore, the parameter nese-English translation, Japanese-English translation, set of the complete model θ contains the GMM mean η English-Korean translation, and translation between English covariance 􏽐 , and weight w ; namely, t i and other languages [11]. English translation also includes some translation skills, such as two basic methods of literal θ � 􏽮w , η , 􏽘 􏽯, i � 1, 2, ..., K. (4) i i t translation and meaning, as well as other techniques such as provincial translation, positive translation, and reverse GMM can be explained; it is a functional expression of translation. English translation can be used in many occa- probability density function. As a linear combination of sions, such as warning signs and learning courseware. ,e Gaussian probability density functions, GMM can ap- specific situation is shown in Figure 1, which is an illus- proximate any density function as long as there are a suf- tration related to English translation. ficient number of mixed components [16]. Language International Transactions on Electrical Energy Systems 3 Figure 1: Diagrams related to English translation. Figure 2: Scenarios related to English translation assessment. translation usually has a smooth probability density func- parameter estimates of the GMM in iterations, increasing the tion, so a finite Gaussian density function is sufficient to matching probability of the model estimate θ and the observed form a smooth approximation to the density function of feature vector at each iteration, and at each iteration, there is k+1 k English translation features [17]. P(|M|θ )> P(|M|θ ), where k is the number of iterations. ,e method used to estimate the GMM parameters is the Figure 4 is a step diagram of the expectation maximi- maximum likelihood estimation method; that is, with respect zation algorithm. to θ, the conditional probability P(M|θ) is maximized [18]. It is assumed that feature vectors are independent of each ,e expectation maximization (EM) algorithm improves the other; there are 4 International Transactions on Electrical Energy Systems 300 w b M i t t P i|M , θ􏼁 � . (11) 􏽐 w b M􏼁 j�1 j j t ,en, we change the symbols in (8), (9), (10), and (11), and the iterative algorithm expression of θ can be obtained as follows: T−1 k+1 k w � 􏽘 P i||M , θ , (12) 􏼐 􏼑 i t t�0 T−1 k P i|M , θ M 􏽐 􏼐 􏼑 t�0 t t k+1 η � , (13) T−1 k 􏽐 P􏼐i|M , θ 􏼑 t�0 t T−1 k T k+1 􏽐 P􏼐i|M , θ 􏼑 M − η􏼁 M − η􏼁 t�0 t t i t i 􏽘 � T−1 k 􏽐 P􏼐i|M , θ 􏼑 0 t t�0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 (14) T−1 k T 􏽐 P􏼐i|M , θ 􏼑M M t t t t�0 k+1 k+1 Figure 3: Gaussian mixture model fitted image histogram. � − η 􏼐η 􏼑 . t t T−1 k 􏽐 P􏼐i|M , θ 􏼑 t�0 t In (14), the superscript T in the covariance matrix represents the transpose of the matrix. T−1 Among them, P(|M|θ) � 􏽙 P(|M|θ). (5) t�0 k k w b M i i t P􏼐i|M , θ 􏼑 � . (15) It is assumed that P(M |θ) is differentiable to θ; when K k k 􏽐 w b M􏼁 j�1 i i P(M |θ) takes the maximum value, it satisfies the following: Among them is the mixture component of the ith d[InP(|M|θ)] (6) g � � 0. Gaussian probability density function in the kth iteration. dθ In a given sample of translated speech, the purpose of ,rough (5) and (6), the estimated iterative algorithm is identifying the language is to determine which language the obtained. speech belongs to [19]. ,e block diagram of the language Take η as an example: i recognition system based on GMM is shown in Figure 5. By Bayesian theory, maximizing the verification prob- d[InP(|M|θ)] g � ability can be expressed as follows: dη P M|θ􏼁 P θ􏼁 i i T−1 (16) P θ |M􏼁 � . d w P M , i|θ i 􏽨􏽐 􏼁 􏽩 i�1 i t − 1 P(M) � 􏽘 P(|M|θ) (7) dη t�0 Among them, T−1 −1 T−1 − 1 � 􏽘 P M |θ􏼁 w P M , i|θ􏼁 􏽘 M − η􏼁 � 0. P M|θ􏼁 � 􏽙 P M|θ􏼁 . (17) t i t t i i t�0 t t�0 ,en, we get the following: Its logarithmic form is as follows: T−1 T−1 􏽐 P i|M , θ􏼁 M t t t�0 η � . (8) InP X|θ􏼁 � 􏽘 InP X |θ􏼁 . (18) T−1 i t i 􏽐 P i|M , θ􏼁 t�0 t�0 Similarly, we can also get the following: Because of the prior probability of θ , T−1 (19) P θ􏼁 � , 1≤ i≤ . w � P i|M , θ , (9) 􏽘 􏼁 i i t t�0 For a certain eigenvector M, P(M) is a certain constant, T−1 equal for all languages. ,erefore, the maximum value of the 􏽐 p i|M , θ􏼁 M − η􏼁 M − η􏼁 t�0 t t t t t � . (10) T−1 posterior probability can be obtained by calculating P(M|θ ) 􏽐 p i|M , θ􏼁 t�0 t so that the recognized language in the phonetic database can be expressed as follows: In (10), International Transactions on Electrical Energy Systems 5 Start Initialization parameters Expectation step Computational Parameter Changes distribution Re-estimate Maximization parameters step Figure 4: Expectation maximization algorithm steps. Feature Input Language Extraction Front-end Language Recognizer Processing Language Model Most LikelyLanguage Figure 5: Language recognition system based on GMM. X N r rx i � arg max P M|θ . (20) G (Γ) � 􏽘 􏽘 􏽘 P w (n)|Q ς w (Q) � t . 􏼁 􏼁 GR θ rx rx rx rx ∗ r�1 x�1 n−1 ,e i here represents the recognized language. (21) Among them, Γ represents the parameter set of the 3.3. Discriminative Training Algorithm. Discriminative hidden Markov model; t is the text of the xth word in the rx training is proposed to improve language recognition rate. rth training sentence; and the posterior probability is as Different from the maximum likelihood algorithm that follows: emphasizes the training data, the language model based on L Q |w (n) 􏼁 P w (n)|w (Q)􏼁 the discriminative training algorithm emphasizes the opti- θ rx rx rx rx P w (n)|Q � . θ rx rx mal discrimination of the training data [20]. After several 􏽐 L Q |v􏼁 P v|w (Q)􏼁 v θ rx rx consecutive recognitions, similar languages can be distin- (22) guished by using this algorithm, and the language model based on the discriminative training algorithm is more Among them, Q is the speech output vector of the xth rx conducive to improving the recognition rate of test data than word in the rth training sentence; v is the possible wrong the traditional maximum likelihood algorithm. Figure 6 is a translation given w (Q); and L (Q |v) represents the rx θ rx scene graph related to the discriminative training algorithm. likelihood probability function of Q given v. rx For a corpus with N training sentences, label each Similarly, “misaccepted” means that although the sentence with r � 1, 2, ..., R. Its objective function is defined identification is wrong, it is wrongly regarded as a correct as its mathematical expectation: judgment; its objective function is as follows: Feature Vector 6 International Transactions on Electrical Energy Systems Figure 6: Scenario diagram related to the discriminative training algorithm. R r 4. Experiments in English Translation (23) G (θ) � 􏽘 􏽘 P w (Q)|Q 􏼁 ς w (Q)≠ t 􏼁. GΓ θ rx rx rx rx Assessment Based on Gaussian r�1 x�1 Mixture Model Corresponding to the correct diagnosis under the true 4.1. Evaluation Plan for English Translation. By collecting the rejection category, the mathematical expectation of the existing research results in the field of English translation, wrong diagnosis is as follows: the importance of language translation in various fields can X N r rx be further understood. ,is paper randomly selected 500 G (θ) � 􏽘 􏽘 􏽘 P w (n)|Q DE θ rx rx students majoring in English translation, 200 teachers (24) r�1 x�1 n�1 majoring in English translation, and 300 white-collar · ς w (Q)≠ t , w (n)≠ t 􏼁. rx rx rx rx workers who work in English translation as the research objects. A total of 1,000 copies of the “English Translation In order to minimize the error, consider the new ob- Assessment Questionnaire” were distributed, and 979 copies jective function G(θ) obtained by summing the three error were recovered. ,ere were 972 valid questionnaires, and the objective functions. When the scaling factor k � 1, the effective recovery rate was 97.2%. derivation result is as follows: Among the valid questionnaires collected, there are 478 students, accounting for about 49.18% of the total number; G(θ) � G (θ) + G (θ) + G (θ) GR GΓ DE 196 teachers, accounting for about 20.16% of the total X N r rx number; and 298 white-collar workers majoring in English � 􏽘 􏽘 􏽘 P w (n)|Q 􏼁 ς w (Q) � t 􏼁 (25) θ rx rx rx rx translation, accounting for about 30.66% of the total. r�1 x�1 n�1 In this questionnaire, a total of 4 questions about English translation were raised, namely, as to whether translation � 􏽘 􏽘 P s|Q 􏼁 Raw Word Error s, t 􏼁 . r�1 S θ r r software is used in daily study or work, opinions on using Among them, Q is the observation vector; t is the translation software, self-evaluation of English translation, and r r translated text of a given text sentence; s is the competing translation proficiency test situation in daily work or study. text identified according to Q ; and the function Raw Word Error(s, t ) is the number of mismatches between 4.2. Evaluation Results of English Translation s and t translations. L Q |s􏼁 P s|t 􏼁 4.2.1. Questionnaire θ r r P s|Q 􏼁 � . (26) θ r 􏽐 L Q |h􏼁 P h|t 􏼁 h θ r r (1) Whether Translation Software is Used in Daily Study or In (26), h is any possible wrong translation given t . Work. Translation software is created to facilitate people’s r International Transactions on Electrical Energy Systems 7 Table 1: Frequency of using translation software in daily life. daily work and study. Even people with high achievements in English translation will have unskilled words, sentences, Corporate white- Students Teachers etc. and will also use translation software [21]. Table 1 shows collar workers the frequency of using translation software in the daily life of Used frequently 386 6 81 the three types of people. Occasionally used 58 111 159 It can be seen from Table 1 that in daily work and life, 386 Hardly ever used 34 79 58 students frequently use translation software, accounting for about 80.75%; 58 students use the software occasionally, accounting for about 12.13%; and 34 students hardly use Table 2: Opinions of three categories of people on translation software. translation software, accounting for about 7.12%. It can be seen from this set of data that most of the students’ Corporate white- Students Teachers translation foundation is still relatively weak, they are still collar workers studying, and relying on translation software is still relatively Very convenient 405 56 146 high. From this part of the data of teachers, it can be found Having pros and cons 40 93 150 that there are 79 teachers who hardly use translation soft- Very bad 33 47 2 ware, accounting for about 40.31%; 111 who use it occa- sionally, accounting for about 56.63%; only 6 who use the software regularly, accounting for about 3.06%. From this that students' use of translation software may make them dependent on it, so that they will not memorize words. part, it can be seen that the teachers’ job is to contact the students, and their English foundation is solid, so they are Learning English is limited to learning knowledge from translation software, which is bad for both teachers and less dependent on translation software. However, there are students. It can be seen from the data of enterprise white- still unfamiliar words or sentences in daily work, and collar workers that 146 people, which account for about translation software is also needed, so the number of people 48.99%, think this kind of software is very convenient, who use it occasionally accounts for more percentage. because they encounter too many problems in their daily Looking at corporate white-collar workers, the English work. ,e use of translation software not only helps them translation ability of this kind of staff is often exercised in their daily work, so their translation ability is strong; solve problems but also saves them a lot of time. And 150 people think that translation software is a double-edged therefore, 58 workers hardly use translation software, ac- counting for about 19.46%. However, due to too many sword, their reasons are the same as those of the students, and they account for about 50.34%. unexpected situations at work, 81 use the software frequently and 159 use it occasionally, accounting for 27.18% and (3) Self-Evaluation of English Translation. As the saying goes, 53.36%, respectively. the person who knows oneself best is oneself. Although the evaluation of one’s English translation ability includes (2) Opinions on the Use of Translation Software. Translation subjective factors, it also has certain reference value [23]. software has advantages and disadvantages in work and Table 3 shows the self-evaluation of the English translation study. ,e advantage is that it facilitates work and study, ability of the three categories of people. facilitates word search, and saves time. ,e disadvantage is that frequent use will rely on it, resulting in lazy emotions From the data in Table 3, 117 students in this group of data, accounting for about 24.48%, think their English and reluctance to learn new words [22]. Table 2 shows three types of respondents’ opinions on translation software. translation ability is relatively good; 238 students think their English translation ability is average, and the proportion is It can be seen from Table 2 that 405 students think that about 49.79%; and 123 students think their translation level the translation software is very convenient and greatly saves still needs to be improved, and the proportion of this group time. In daily learning, there is no need to look up dictio- is about 25.73%. ,e biggest reference for students’ self- naries or do direct online search. ,ose students account for evaluation is their usual assessment scores. Better grades are about 84.73%. 40 students maintained a neutral attitude, considered excellent, average grades are considered ordi- thinking that translation software has both an advantage and disadvantage. ,e advantage is that it is convenient to query nary, and poor grades are considered very poor. It can be seen from the data of teachers that 97, accounting for about vocabulary and quickly solve translation problems. ,e disadvantage is that this software may replace the English 49.49%, think their translation ability is excellent; 96 think that their translation ability is average, and the proportion is translation major, which is a bad phenomenon for people with this major. ,ose students account for 8.37%. And 33 about 48.98%; and some teachers think that their translation ability still needs to be improved. ,is happens mainly think that the software is very bad. Many words have too because teachers evaluate themselves from their own many translation results on the software, and it is difficult to teaching ability. Among the white-collar workers in the third distinguish which translation is the answer they need. ,ose part, 156, accounting for about 52.35%, think their trans- students account for about 6.90%. From the data of teachers, lation ability is relatively good; 140, accounting for about it can be seen that 56 teachers, accounting for about 28.57%, 46.98%, think that their English translation level is average; think that translation software is very convenient. 93 people think that the software has an advantage and disadvantage; of course, some people think that their translation level is very poor, and those account for about 0.67%. ,e main the advantage is that it saves time, and the disadvantage is 8 International Transactions on Electrical Energy Systems Table 3: Self-evaluation of English translation ability of three 227 227 categories of people. 209 210 Students Teachers Corporate white-collar workers Excellent 117 97 156 150 Ordinary 238 96 140 100 90 Very poor 123 3 2 reason for this kind of situation is that the daily work of the company is both difficult and easy. According to this part of Freshman Sophomore Junior year Senior year the data, it can be concluded that the ability to subjectively Years judge oneself cannot be limited to only one condition. It is Below 60 points 70-80 recommended that many aspects are considered, so that the 60-70 80 points or more possibility of subjective judgment is smaller. Figure 7: Students’ final test scores for the past four years. (4) Translation Proficiency Test in Daily Work and Study. Whether students or staff, in order to objectively understand their English translation level, they need to be tested. Stu- dents’ tests are mainly daily tests, as shown in Figure 7, which compares and analyzes the final exams of students of the same major in the past four years. ,e teachers' test is 100 mainly about the passing rate of English proficiency, as 89 shown in Figure 8, which is based on the passing rate after 80 graduation. ,e test of corporate white-collar workers is mainly about the accuracy of translation at work and the time to translate a paragraph. ,e specific situation is shown in Figure 9. Figure 7 shows the distribution map of the number of students in each grade of final grades in the past four years. 1 1 It can be seen from Figure 7 that in the past four years, 2018 2019 2020 2021 the number of students in low grades has decreased year by Years year, and the number of students in high grades has in- pass the test creased year by year. Among them, the number of people Excellent grades Failed grades below 60 was decreasing year by year, from 38 in the Failed grades freshman year to 3 in the senior year, and the population Figure 8: ,e number of teachers who passed the English below 60 in the English translation subject test was de- translation proficiency test. creasing. ,e number of students with scores above 80 was increasing every year, from 54 in the first year, accounting for 11.30% of the total, to 101 in the fourth year, accounting for 21.13% of the total, an increase of nearly 10%. ,is showed that with the deepening of learning content, the level 70.00 66.44% of translation of students is gradually rising. 62.42% 59.40% Figure 8 shows the number of teachers who have passed 60.00 54.03% the English translation proficiency test in the past four years. 50.00 It can be seen from Figure 8 that in the past four years, teachers have participated in the English translation profi- 40.00 30.87% ciency test with a high pass rate, but there are also people 28.18% 26.85% 30.00 who fail. ,e number of teachers who pass and perform well 19.46% 19.12% also increased every year. From 2018 to 2021, the pass rate 20.00 14.10% was above 98%. ,e lowest year was 2019, and the pass rate 9.73% 9.40% 10.00 was also 98.47%. Among them, the number of teachers who have passed with excellent grades has increased year by year. 0.00 2018 2019 2020 2021 In 2018, there were 92 teachers, accounting for about Years 46.94%. Looking at 2021, we find that the number was 107, accounting for more than 50%. ,is showed that with the pass the test Excellent grades increase of working years, the professional level of these Failed grades teachers also maintains an upward trend. Figure 9 shows a survey of corporate white-collar Figure 9: Enterprise white-collar workers’ survey of their trans- workers’ translation capabilities in the past four years. lation capabilities in the past four years. Proportion (%) Quantity Quantity International Transactions on Electrical Energy Systems 9 Table 4: Confusion matrix of the number of 1-frame speech tests. As can be seen from Figure 9, the test requirements of enterprises for English translators have increased year by Recognition result Sample size year. It can be seen from the data that the failure rate in the Japanese Chinese German French four years from 2018 to 2021 has almost doubled from 14.1% Japanese 98 27 51 7 21 to 28.18%. Although there were small changes in the middle, Chinese 96 28 38 15 15 the overall failure rate was on the rise. Of course, during this German 90 10 29 36 15 period, there are still many workers who are constantly French 88 29 19 31 9 improving their strength and professional level. In the work test, their scores are also improving year by year. Table 5: Confusion matrix for number of 2-second speech tests. Recognition result 4.2.2. Language Recognition Experiment Based on Gaussian Japanese Chinese German French Mixture Model. English translation includes not only text translation, but also voice translation. ,e first part of the Japanese 0.5578 0.0938 0.2263 0.1221 Chinese 0.080 0.6612 0.229 0.0298 questionnaire survey was mainly based on text translation, German 0.0779 0.2335 0.5321 0.1565 and then experiments were conducted on voice translation. French 0.09 0.05 0.2111 0.6489 In this experiment, four languages were selected for translation results experiment, which are Japanese, Chinese, German, and French. ,e total number of experimental the sample size is small, and the results are not accurate; it is samples is 400 8-second speeches and 400 2-second recommended that the sample size is increased. Second, in speeches. After processing through feature parameters, the the speech translation experiment, the sample data is large, dimension of each feature vector is 16, and the mixing but the selected sample time period is short, and the sample degree of GMM is 32. Experiments were conducted with selection source is single too. By increasing the sample speech with a length of one frame and a speech with a length sources, the experimental data can be more complete and the of 2 seconds (188 frames). ,e details are shown in Table 4. experiments can be comparable. ,e application of the It can be seen from Table 4 and Table 5 that only the first discriminative training algorithm based on the improved frame of each 2-second sample was taken to test the Gaussian mixture model in the English page-turning pro- translation effect and then the sample was tested. ,e rec- fession is a field that experts and scholars from all walks of ognition results obtained are shown in Table 5, and the life have been keen in recent years, and many issues still need correct translation rate was only 46.18%. ,is first-order further research and discussion. polynomial contains the information between frames, but only one frame, with 2 seconds of speech segment, was used 5. Discussion for sample testing, and its translation recognition result was not high. With the development of economic globalization, the de- To sum up, through the questionnaire survey and GMM mand for English translation professionals is also increasing, experiment, the English translation ability of the three and the requirements for this profession will also increase. groups of people and the language translation technology of To understand the basic level of this profession, it also needs the GMM can be learned. In this regard, there are still some to be tested [24]. ,is paper firstly integrated the achieve- small problems in the experiment that need to be improved, ments and ideas of English translation professionals in the and the evaluation of the English test needs to be carefully major and used the discriminative training algorithm of analyzed. In terms of written translation and oral transla- Gaussian mixture model to discuss the problems of 1,000 tion, the results are diametrically opposite. English translation professionals. ,en, it used the GMM to study the related problems of English translation in speech translation. Finally, through a series of analyses, two op- 4.3. Evaluation Results of English Translation by Improved posite results were obtained. In the process of written Gaussian Mixture Model. According to the analysis of this questionnaires, it was found that the English translation paper, the application of the discriminative training algo- rithm of GMM in English translation is relatively perfect. ability of all kinds of people was relatively strong, but in the voice translation, their translation ability was still weak. With the advancement of science and technology, intelligent translation technology has gradually been recognized and ,is paper is devoted to researching the related application of the discriminative training algorithm based on GMM and used by the public. ,e use of GMM’s discriminative applying it in the field of English translation evaluation. ,is is training algorithm to collect the opinions of English not only an expansion and extension of the application scope of translation professionals on the profession greatly facilitates the algorithm, but also a new attempt to evaluate English the research of this experiment. However, because the re- translation. ,rough the above case studies, opposite experi- search in this field is not comprehensive enough, there are mental results are obtained in the written translation experi- some small problems, which need to be corrected. ,e details are as follows: First, in the questionnaire survey, three ment and the speech translation experiment, which both show the effectiveness of the discriminative training algorithm in this groups of people were selected for experimental investiga- tion; the results are better, but in this part of the experiment, paper under the Gaussian mixture model. 10 International Transactions on Electrical Energy Systems wavenumber filter and its applications in least-squares seismic 6. Conclusions imaging,” IEEE Transactions on Geoscience and Remote Sensing, vol. 60, no. 99, pp. 1–13, 2022. In this paper, the discriminative training algorithm of GMM [7] G. Gao, H. Jiang, J. C. Vink, C. Chen, Y. El Khamra, and was used as the main analysis method; 1,000 randomly J. J. Ita, “Gaussian mixture model fitting method for uncer- selected people of three types were selected as the research tainty quantification by conditioning to production data,” object; and the related problems of students, teachers, and Computational Geosciences, vol. 24, no. 2, pp. 663–681, 2019. translators in English translation were selected. ,rough the [8] Y. Xu, Y. Fang, W. Peng, and Y. Wu, “An efficient Gaussian case study, the following conclusions were drawn: ,ere are sum filter based on prune-cluster-merge scheme,” IEEE Ac- subjective evaluation and objective evaluation in English cess, vol. 7, pp. 150992–151005, 2019. translation professional test; the subjective test needs to [9] J. Han, L. Xie, Z. Wang, and F. Renl, “Incrementally emotion comprehensively consider many aspects and cannot deter- mapping based on GMM,” Harbin Gongye Daxue Xuebao/ mine one’s translation ability only by one factor, and the Journal of Harbin Institute of Technology, vol. 50, no. 8, same is true for the objective evaluation. ,e translation pp. 168–173, 2018. includes voice translation and written translation. It can be [10] S. Sagratella, “Algorithms for generalized potential games seen from the experiment that the written translation survey with mixed-integer variables,” Computational Optimization and Applications, vol. 68, no. 3, pp. 689–717, 2017. results are better. All kinds of people are good at written [11] L. Zhou, Y. Zheng, L. Huo, Y. Ye, X. Wang, and G. Song, translation, while voice translation is the use of intelligent “Fracture behaviors of HVFA-SCC mixed with seawater and technology to memorize translation, and the result has sea-sand under three-point bending,” Advances in Structural certain defects. From one angle of experiment, the reliability Engineering, vol. 25, no. 4, pp. 716–735, 2022. of the results is low, so it needs to be analyzed from multiple [12] M. V. Farahani, “Review of feng (2020): form, meaning and angles and multiple aspects to better highlight the efficient function in collocation: a corpus study on commercial Chi- application of GMM’s discriminative training algorithm in nese-to-English translation,” International Journal of Corpus English translation evaluation. Linguistics, vol. 27, no. 2, pp. 254–258, 2022. [13] S. R. Sweger, S. Pribitzer, and S. Stoll, “Bayesian probabilistic Data Availability analysis of DEER spectroscopy data using parametric distance distribution models,” =e Journal of Physical Chemistry A, ,e data that support the findings of this study are available vol. 124, no. 30, pp. 6193–6202, 2020. from the author upon reasonable request. [14] Z. Liu, S. Li, and K. Zhao, “Extended multi-interval Legendre- Gauss-Radau pseudospectral method for mixed-integer op- timal control problem in engineering,” International Journal Conflicts of Interest of Systems Science, vol. 52, no. 5, pp. 928–951, 2020. [15] S. Rupesh Kumar and B. Bharathi, “Generative and dis- ,e author declares no potential conflicts of interest with criminative modelling of linear energy sub-bands for spoof respect to the research, authorship, and/or publication of detection in speaker verification systems,” Circuits, Systems, this article. and Signal Processing, vol. 41, no. 7, pp. 3811–3831, 2022. [16] Y. Chen, Y. Chen, and B. Yu, “Speed distribution prediction of References freight vehicles on mountainous freeway using deep learning methods,” Journal of Advanced Transportation, vol. 2020, [1] J. Pan, T. M. Wong, and H. Wang, “Navigating learner data in no. 3, pp. 1–14, 2020. translator and interpreter training:insights from the Chinese/ [17] G. D. Barriga, G. M. Cordeiro, D. K. Dey, V. G. Cancho, English translation and interpreting learner corpus F. Louzada, and A. K. Suzuki, “,e Marshall-Olkin gener- (CETILC),” Babel, vol. 68, no. 2, pp. 236–266, 2022. alized gamma distribution,” Communications for Statistical [2] Q. J. Li, R. R. Wu, and Y. Ng, “Developing culturally effective Applications and Methods, vol. 25, no. 3, pp. 245–261, 2018. strategies for Chinese to English geotourism translation by [18] C. Li, G. Casella, and M. Ghosh, “Estimation of regression corpus-based interdisciplinary translation analysis,” Geo- vectors in linear mixed models with Dirichlet process random heritage, vol. 14, no. 1, pp. 1–24, 2022. effects,” Communications in Statistics - =eory and Methods, [3] M. Zhao, L. H. Ang, S. Md Rashid, and F. H. C. Toh, vol. 47, no. 16, pp. 3935–3954, 2018. “Translating narrative space in children’s FictionBronze and [19] Y. Zhang and X. Chen, “,e deformed gesture tracking al- SunflowerFrom Chinese to English,” Sage Open, vol. 12, no. 1, gorithm based on feature space segmentation modeling,” Jiqiren/Robot, vol. 40, no. 4, pp. 401–412, 2018. [4] X. Si and J. Wang, “Grammatical metaphor in English-Chi- [20] I. J. David, O. E. Asiribo, and H. G. Dikko, “Nonlinear split- nese translation,” International Journal of Translation Inter- plot design model in parameters estimation using estimated pretation and Applied Linguistics, vol. 3, no. 1, pp. 15–29, 2021. [5] Z. Tan and X. Ke, “On EST sentence translation based on generalized least square-maximum likelihood estimation,” ComTech Computer Mathematics and Engineering Applica- English and Chinese structures: natural linear expansion vs. Reversed linear expansion,” =eory and Practice in Language tions, vol. 9, no. 2, pp. 65–71, 2018. [21] E. M. M. Ortega, G. M. Cordeiro, A. K. Suzuki, and Studies, vol. 11, no. 4, pp. 396–402, 2021. [6] J. Yang, J. Huang, Z. Li, H. Zhu, G. A. McMechan, and X. Luo, T. G. Ramires, “A new extended Birnbaum-Saunders model “Approximating the gauss-Newton hessian using a space- with cure fraction: classical and Bayesian approach,” International Transactions on Electrical Energy Systems 11 Communications for Statistical Applications and Methods, vol. 24, no. 4, pp. 397–419, 2017. [22] G. Xie, Y. Wu, H. Gong, C. Li, and W. Zhang, “Data pro- cessing in residual stress measurements using contour method,” China Mechanical Engineering, vol. 28, no. 15, pp. 1778–1783, 2017. [23] L. Ko, “Chinese certificate translation in the Australian context:A purpose-oriented practice,” Babel, vol. 68, no. 1, pp. 24–54, 2022. [24] C. Huang, “Analysis of common errors in financial English- Chinese translation,” OALib, vol. 08, no. 01, pp. 1–7, 2021.

Journal

International Transactions on Electrical Energy SystemsHindawi Publishing Corporation

Published: Aug 31, 2022

References