Design of computer aided translation system for English communication language based on grey clustering evaluation

Abstract

In order to facilitate communication and communication, this paper studies the optimization of the current computer-aided translation system, and proposed a design method of English communication language computer-aided translation system based on grey clustering evaluation. By optimizing the hardware configuration and algorithm function keys of the system, the English translation mechanism of multilanguage interaction, the design idea of editing and modifying after English translation and knowledge database management are realized, and the system function framework was constructed, including the system transceiver unit, automatic translation unit, manual correction unit, task management unit and memory management unit, the performance of task management unit and memory management unit is analyzed. On this basis, the specific work flow of the design system mainly includes the English translation service flow integrating multilanguage interaction and the project-based multilanguage interaction English translation service flow design, which realizes the English translation online assistance under the multilanguage interaction environment. The experimental results show that the design system has the advantages of high online translation speed, pronunciation success rate and multilanguage translation success rate.

Keywords

Grey clustering evaluation English communication assisted translation

1. Introduction

Language translation is a common task in learning and scientific research. Manual translation has high requirements for translators and wastes a lot of time and energy, so people usually choose software translation [1]. However, at present, the translation quality of all translation software is far from that of manual translation, and most of the results can not be used directly. Although the algorithms of translation software are constantly improving, the quality of translation still needs to be improved [2]. Based on this, this paper designs a computer-aided translation system of English communication language based on grey clustering evaluation. Taking English Chinese translation as an example, combined with the idea of artificial intelligence, this paper constructs a translation system including system transceiver unit, automatic translation unit and manual error correction unit, optimizes the English translation service process integrated with multi language interaction, and carries out English translation under multi language interaction environment, so as to better improve the translation accuracy and ensure the operation effect of the system [3, 4].

2. Computer aided translation system for english communication language

2.1 Hardware configuration of computer aided translation system for English communication language

The translation system can use the communication computer as the platform, combine the web server with the database technology, transform the information needed in the auxiliary translation mode into data, and the professional management personnel can supervise and manage the system [5]. The translator can log in to the system and set up his own translation assistant studio to carry out the auxiliary translation activities. In the web-based assisted translation system, the establishment and composition of the assisted translation team and the assistance among translators will change with the demand [6]. In the system design, the assisted translation process should be taken as the data flow basis, and the translator management authority should be used to assist the management. Based on this, the hardware configuration of the language computer-aided translation system is optimized, as shown in Fig. 1.

Figure 1.

Structure of language computer aided translation system.

In this framework, the business logic layer is implemented by spring, which uses the control transformation IOC dependency injection Di mechanism as the core of the spring framework. IOC injects the implementation dependency of each module of business logic into the action module by reading the configuration file, so as to actually call the business logic. During runtime, Di container dynamically injects dependencies (such as construction parameters, construction objects or interfaces) into components. By using bean attributes to configure corresponding attributes in bean. XML file, spring framework uses bean’s reverse control to put bean into spring container to make bean become a component of business logic layer. For example, in the on-line auxiliary conversion module, only need to configure the bean properties, create a Dao object that inherits the hibernate Dao class provided by spring, and then copy the crud method to complete the loading of the business logic components and integrate them into the spring container management. Due to the complex hardware structure of the system, you need to add multiple circuits and use Pt pulse signal to measure the surge current of the arrester. Hardware circuit diagram as shown in Fig. 2.

Figure 2.

System hardware configuration structure.

In Fig. 2, in the online auxiliary transformation module, you only need to configure bean properties, create a DAO object that inherits the hibernatedao class provided by spring, and then copy the crud method to complete the loading of business logic components and integrate them into spring container management [7]. Due to the complexity of the hardware structure of the system, multiple circuits need to be added, and Pt pulse signal is used to measure the surge current of the arrester. MCU chip is selected as the hardware circuit structure of the system, which is produced by silicon company. It has the ability of real-time information collection, small size and strong anti-interference ability [8]. Because the oscillator is added in the chip, it has the programming ability. At the same time, it integrates 15 bit 150 ksps ADC and 2 UARTS, which greatly enhances the processing speed of the chip, up to 25 mips. The memory in the chip is 15 KB flash memory, which can store 1282 bytes of data. Two I/O ports are used to connect two external power supplies to ensure that the chip can stably output 2 V–5 V power supply voltage [9]. The monitoring system has 12 different timers in four different positions, and controls the system kernel with pipeline structure, which can meet the requirements of measurement and processing [10]. System framework is the basic structure of system design, which plays a guiding role in a system design. The framework of the system is based on B/S architecture. B/S architecture, full name of browser/server in Chinese, is a three-tier architecture, which is composed of data logic layer, business logic layer and display logic layer, as shown in Fig. 3.

Figure 3.

B/S three-layer architecture.

In Fig. 3, the main storage language information and a variety of translation programs. The communication module selects TL16C554 asynchronous communication module produced by TI company [11]. In order to ensure the expansion function of serial port, four serial ports are added. Each serial port can be converted to the serial port in MCU. All communication states, including normal state and error state, are recorded through registers to ensure the reliability of communication process [12, 13]. FIFO communication mode is extended to the common communication mode, which can receive and cache up to 16 bytes of data and improve the data transmission rate of the communication process [14]. The input level is TLT level. The structure of communication module is shown in Fig. 4.

Figure 4.

Communication module structure of English translation system.

The communication module in the figure has four TL16C554 asynchronous communication components, each communication component has a programmable baud rate generator, and the highest baud rate is 1.5 mbps [15]. The data state is controlled by data controller and register. The data control mode is full duplex control, and each receiving and sending is independent. At the same time, the design of interrupt control is strengthened, which can receive data even if the timeout interrupt. The integration and unified management of server resources through cloud operating system can save computer resources and ensure the safe and stable operation of the system [16]. The device used for language translation processing is a single chip microcomputer. It is composed of a series of hardware components such as arithmetic unit, controller, memory, input and output device, which is equivalent to a microcomputer. The single chip microcomputer in the system is atmega6490. The MCU achieves the throughput of 1 mips per megahertz and balances the power consumption and processing speed [17]. Therefore, high performance and low power consumption are its biggest advantages. The technical parameters of atmega6490 are shown in Table 1.

Table 1

Technical parameters of single chip microcomputer

Name	Parameter
Core processor	AVR
Running speed	8 Mhz
RAM capacity	4 K $\times$ 8
Program memory type	FLASH/64 KB (64K $\times$ 8)
Input/output number	68
Oscillator type	Internal
Data converter	SPI, UART/USART, USI A/D 8 $\times$ 10 b
Working voltage	1.8 V–5.5 V
Working temperature	$-$ 40 ${}^{\circ}$ C–85 ${}^{\circ}$ C

Using virtualization technology, while providing a large number of virtual resources to the data center, summarize the basic physical hardware resources among multiple systems. As an assistant translation system, VMware vSphere can manage large-scale infrastructure seamlessly and dynamically [18]. Through the services provided by the center server, the system manages and uses the resources in the virtual machine cluster, and manages the complex data center at the same time. Users can use the web browser to access the auxiliary translation system. Sphere web access is used to resolve the physical location of the virtual machine and redirect the web browser to the virtual machine [19]. The auxiliary translation system requires strong operation ability to support online translation activities.

2.2 Module optimization of computer aided translation software for English communication language

The auxiliary translation activities based on grey clustering evaluation only need a computer connected to the Internet to ensure the smooth development of translation work. Auxiliary can create a greater effect than the simple accumulation of a single workload, and realize the synergistic effect. The advantage of assistance is that it can make full use of the organization resources, shorten the working time, and focus on the tasks that are difficult for individuals to complete in a short time [20]. The construction of auxiliary platform is to make full use of organizational resources to achieve common goals. In the process of team assisted translation, creative ideas and translation methods often appear unexpectedly. In order to make the platform play a positive role and effectively promote the assistance and communication between translators, it is necessary to build a convenient and standardized network assisted translation environment. On the auxiliary translation platform, translators can create their own independent translation studios, and then invite translators to join their own translation studios according to the actual needs [21]. A translator can belong to multiple translation studios. In this way, translators jointly construct an assisted translation environment. In the process of assisted translation activities, translators can also participate in each other’s translation activities. The whole translation activity is a synchronous, dynamic and open process. The membership of the platform is shown in Fig. 5.

Figure 5.

Optimization of the function and structure of the assisted translation platform.

The knowledge base is divided into local data and network data, and some common templates are put in the local database to improve the search speed. However, the local database can only store a small part of the huge knowledge base, and a large amount of data needs to be obtained from the network. When translating, first search from the local knowledge base, and then search from the network [22, 23]. If the content to be translated is the same as the existing translation template, the results will be returned directly. If there is no same template, the search results will be ranked according to the possibility, and finally choose the translation method with the highest possibility as the formal translation method. Based on this, the translation process based on the knowledge base will be carried out Optimization, as shown in Fig. 6.

Figure 6.

Optimization of translation process based on knowledge base.

In computer programming, translation is widely used and important. According to the different amount of translation records processed by them, AI machine translation and external translation are two major parts of translation. Among them, the most widely used artificial intelligence machine translation, which stores the data to be arranged completely in the memory, and then translates according to their own translation rules, is suitable for the element sequence with a small amount of data. Translation can be divided into direct insertion translation, Hill translation, simple selection translation, heap translation, bubble translation, fast translation, merge translation and cardinal translation. These translation algorithms can translate data or records to achieve the effect of easy to find, but the actual data processing method of each translation algorithm is not the same. Direct insertion translation is a binary sequence to be translated, forming a pattern of ordered table and unordered table. The initial state is that no element is in the ordered table, and all elements are placed in another table. The data that can be inserted into the ordered table is the first element in the unordered table during each scan, and then the ordered table is reordered until the unordered table is empty complete.

2.3 Realization of computer aided translation of English communication language

In the process of designing the computer-aided translation system for English communication language, first of all, simply select the translation, select the number with the minimum (large) key value from the data, and put it in the first position, then select the maximum value from the data except the arranged number, and put this element in the back position of the previously selected element, and cycle until the full formation of the existing system Sequence. Bubble translation compares the key values of every two adjacent records in the data. When comparing each time, the condition that two elements need to be exchanged is that when the translation condition is opposite to the required translation condition, it will repeat until the sequence is in order. On this basis, the data to be arranged is divided into groups, and the elements in the same group are separated by multiple before translation. Each group forms an ordered sequence according to the principle of insertion translation. After each group is ordered, the second increment is taken as the basis of grouping and each group is inserted into translation. When the last increment is taken, all data are inserted into translation. The main algorithm steps of translating basic words of compound words into Chinese are as follows.

$\displaystyle J(t)=\frac{1}{T}\sum\limits_{t=1}^{T}{\sum\limits_{\log}}\left({% w_{t}}\right)$ (1)

The formula calculates the conditional probability of context words of a word, and the variable $T$ represents the number of training sentences, $w$ represents sort the increment. With the increase of training times, the probability model tends to be more accurate. Then the logarithm probability of the context words of the existing head words is maximized.

$\displaystyle\log p(o\mid c)=\log\frac{\exp\left({u_{o}^{T}v_{c}}\right)}{\sum% \limits_{w=1}^{W}{\exp}\left({u_{w}^{T}v_{c}}\right)}$ (2)

Where $V_{c}$ is the word vector of the head word. Each word has two vector representations ( $U_{O}$ ), one of which is when the word is used as the center word, and the other is when the word is used as an external word. These vectors are trained by the random gradient descent method. In the calculation of similarity, there are many ways to measure it, such as through the structural information and meaning of words. Generally speaking, the more words the two phrases contain, the more similar they are. In this paper, idce coefficient is introduced to judge the similarity. Similarity measure is to use quantitative method to classify things, that is, to describe the similarity between things quantitatively by similarity measure coefficient. The function of similarity measurement coefficient:

$\displaystyle\textit{sim}\left({\bar{f}_{i}^{\prime},\bar{f}_{i}^{\prime}}% \right)=\frac{\sum\limits_{j=1}^{J}{\delta_{c,w}}}{J(t)\log p(o\mid c)}$ (3)

Where:

$\displaystyle\delta_{ij}=\left\{{{\begin{array}[]{*{20}c}{1,ifi=j}\\ {0,ifi\neq j}\\ \end{array}}}\right.$ (4)

In the above formula, $c_{j}$ and $w_{j}$ represent words in different phrases, $i f i$ represents different phrases, and sim () represents similarity. The problem of insert translation itself is that as the amount of data increases, the number of comparisons cannot be determined and increases, which seriously affects the efficiency of runtime. At this time, we need binary search translation to improve it. Binary search, also known as half search, is a search algorithm to find a specific element in an ordered array. Its advantages are less comparison times and high retrieval efficiency. Binary search plays an important role in binary search translation. It can greatly shorten the number of keyword comparisons and improve the efficiency. Binary search is integrated into translation. According to the keyword value, the data to be arranged is divided into two categories: One is larger than the keyword value and the other is smaller than the keyword value. In the data group and the data group, divide them into two parts according to the first step of S. Until there are only 2–5 data left in the array, the translation algorithm is used to translate. Compared with the ordinary translation algorithm, the method of dividing the data into two parts first and then translating it effectively improves the efficiency and shortens the iteration time. Therefore, the application of binary retrieval in the insertion translation algorithm is very necessary and effective. The disadvantage of selective translation in operation is that the number of comparisons is too much, because it has a maximum value each time, so the operation efficiency is slow. Therefore, to solve this problem, in each scan, select two maximum values: the maximum value and the minimum value, slowly narrow the scope of comparison, complete the translation. In the past, only one maximum value was selected as the keyword value in each scan in the selection of translation, but after the improvement, two maximum values, i.e. the maximum value and the minimum value, were selected to exchange with the data respectively, which reduced the workload by half and the running efficiency by a lot. Through the improvement, two maximum values are taken out, and the running process is shortened by half. Compared with the ordinary selection translation algorithm, the efficiency is effectively improved, and the iteration time is also greatly improved. Therefore, the application of this improved method in the selection translation algorithm is very necessary and effective. Exchange translation is not stable enough, but it has a great advantage in running efficiency. However, with the increase of data volume, due to the complexity of the algorithm, its advantage gradually weakens. Therefore, we can use the stability of insert translation and the addition of two maximum and intermediate values to enhance the stability and efficiency. In exchange translation, it often refers to the use of a keyword value to segment the sequence for translation. But now we use three numbers as the benchmark to translate. In this way, we divide the data into three parts and translate them separately. Considering the stability, we integrate the insert translation into it to enhance the stability, so as to effectively improve the operation effect of the auxiliary translation system.

3. Experimental results

In order to ensure the effectiveness of the intelligent research of English language translation based on the grey clustering evaluation algorithm, the simulation experiment was carried out. In the process of the experiment, different English translations are taken as the experimental objects, and the translation accuracy and translation response time are simulated. The quantity and difficulty of English translation are simulated. In order to ensure the effectiveness of the experiment, the conventional English translation research method is used as the comparison object, and the results of the two simulation experiments are compared. In order to investigate the distribution of syntactic equivalent pairs, we put statistics on the number and length of syntactic equivalent pairs in the experiment. In the table, we give the statistical results of the number of syntactic equivalent pairs. Based on the above analysis process, combined with Figs 1 and 2, the setting is as follows: The first line is the average length of the extended reference translation. The second is the number of all syntactic nodes contained in the syntactic structure of the behavioral reference translation. The third is the number of syntactic equivalent pairs in the syntactic structure of behavioral reference translation. The fourth is the average number of syntactic nodes in each reference translation. Fifth, in the syntactic structure of the behavioral reference translation, the average number of nodes in the syntactic equivalent pairs contained in each reference translation.

Table 2
Node statistics of reference translation syntax tree

	LDC2006T04	LDC2008E43
Average length of reference translation	31.52	34.43
Total number of syntax tree nodes	211231	62569
Total number of syntactic equivalent pairs	21807	10073
Average number of nodes in syntax tree	57.46	62.82
Syntactic equivalence vs. average number of nodes	5.93	10.11

Based on the statistical results of Table 2, the similarity ratio of N groups of words between the translated version and the reference translation is calculated. Its values range from 0.0 to 1.0. If two sentences match perfectly, value is 1.0. On the contrary, if two sentences do not match, bleu is 0.0. Therefore, the closer to 1.0, the higher the translation quality and the more accurate the interaction.

Table 3

Comparison of operation accuracy of translation system

Case type number	Traditional system/%	The paper system/%
1#	64	95
2#	78	92
3#	62	92
4#	84	94
5#	77	97
6#	65	96

By using different systems to translate the three test samples, it is concluded that the bleu index value of the designed system is larger and the translation speed is faster, the translation accuracy of the proposed English translation research method is 25.33% higher than that of the English translation research method. The designed system solves the lag and error problems of the current translation system and improves the interaction between users. In this paper, two different methods of English translation research are used to analyze the changes in translation accuracy in the simulated environment. The test results are shown in Table 3.

Through the arithmetic mean value processing of the proposed English translation research method and the conventional English translation research method, it is concluded that the translation accuracy of the proposed English translation research method is higher than that of the English translation research method. During the experiment, two different English translation research methods was also used to analyze the change of translation response time in the simulated environment. The comparison curve of test results is shown in Fig. 7.

Figure 7.

Comparison of response time of translation system.

Compared with the traditional translation system, in the same environment, the corresponding speed of the proposed system is significantly faster, and the amount of translation information processing is also significantly increased. The translation response time of the designed system not more than 50 s. Therefore, it is proved that the gray clustering evaluation based English communication language computer-aided translation system is feasible. The language computer aided translation system has high effectiveness in practical application and fully meets the research requirements.

4. Conclusion

The traditional language translation system is mainly based on the lexical structure analysis method for deduction, which has the problem of poor standardization of translation results, which leads to the inherent defects of this translation method. Based on this, this paper designs the computer-aided translation system of English communication language based on grey clustering evaluation, explains the principle of grey clustering evaluation in detail, and proves it through experiments. Compared with the traditional translation method based on lexical structure analysis, this method has many advantages.

In the future research, we should focus on the proportion of the total number of verbs in the total number of verbs and adjectives in the process of English translation, so as to make the translation results more targeted.

References

Last

P.J.

Engelbrecht

H.A.

and Kamper

, Unsupervised feature learning for speech using correspondence and Siamese networks, IEEE Signal Processing Letters 27(10) (2020), 421–425.

Gao

S.S.

Liu

K.H.

X.F.

Haider

D.A.H.M.

Kong

F.S.

Liu

Cory

R.E.E.D.

Sun

M.C.

and Yu

Y.Q.

, Structural seismology: exploring the correspondence between surface geological features and heterogeneities in the earth’s crust and mantle, Abstracts of the 9th World Chinese Conference of Geological Science (2019), 2.

Raheli

Rezaei

R.M.

Jadidi

M.R.

and Mobtaker

H.G.

, A two-stage DEA model to evaluate sustainability and energy efficiency of tomato production, Information Processing in Agriculture 4(4) (2017), 342–350.

Chatterjee

and Gupta

, Efficient phrase table pruning for Hindi to English machine translation through syntactic and marker-based filtering and hybrid similarity measurement, Natural Language Engineering 25(1) (2019), 171–210.

Jia

Y.F.

Carl

and Wang

X.L.

, Post-editing neural machine translation versus phrase-based machine translation for English-Chinese, Machine Translation 33(1–2) (2019), 9–29.

Khan

N.S.

Abid

and Abid

, A novel Natural Language Processing (NLP)-Based machine translation model for English to Pakistan sign language translation, Cognitive Computation 12(2) (2020), 1–18.

English

Son

J.M.

Cardamone

M.D.

Lee

and Perissi

, Decoding the Rosetta stone of mitonuclear communication, Pharmacological Research 161(12) (2020), 105161.

Mardani

and Salarpour

, Measuring technical efficiency of potato production in Iran using robust data envelopment analysis, Information Processing in Agriculture 2(1) (2015), 6–14.

Zheng

Wang

Zheng

and Zheng

, Unpaired photo-to-caricature translation on faces in the wild, Neurocomputing 355 (2019), 71–81.

10.

Kulakov

S.M.

Musatova

A.I.

and Kadykov

V.N.

, Multivariate estimation of the production time for steel-wire batches by means of situational-normative models. Part 1, Steel in Translation 49(6) (2019), 390–396.

11.

Barroga

and Mitoma

, Critical thinking and scientific writing skills of non-anglophone medical students: a model of training course, Journal of Korean Medical Science 34(3) (2019), e18.

12.

Geffert

, Unary coded PSPACE-complete languages in ASPACE (loglog n), Theory of Computing Systems 63(4) (2019), 688–714.

13.

Santino

T.A.

Alves

R.E.F.M.

Monteiro

K.S.

Okelo

S.O.

Patino

C.M.

Alchieri

J.C.

and Mendonça

K.M.P.P.

, Psychometric evaluation of the Brazilian version of the pediatric asthma control and communication instrument, Pediatric Pulmonology 55(1) (2020), 15–26.

14.

Thompson

A.L.

and Do

, Unconventional spoken iconicity follows a conventional structure: Evidence from demonstrations, Speech Communication 113 (2019), 36–46.

15.

Luo

and Xiang

, Application of data mining methods in internet of things technology for the translation systems in traditional ethnic books, IEEE Access (99) (2020), 1.

16.

Chandramoorthi

and Thittai

A.K.

, Enhancing image quality of photoacoustic tomography using sub-pitch array translation approach: simulation and experimental validation, IEEE Transactions on Biomedical Engineering 66(12) (2019), 3543–3552.

17.

Azarbonyad

Shakery

and Faili

, A learning to rank approach for cross-language information retrieval exploiting multiple translation resources, Natural Language Engineering 25(3) (2019), 363–384.

18.

Giron

J.F.

Ramsey

S.D.

and Temple

B.A.

, Conditions for translation and scaling invariance of the neutron diffusion equation, Progress in Nuclear Energy 110 (2019), 333–340.

19.

S.H.

and Lai

Y.K.

, Simultaneous multiã€¢ttribute image-toãŠ£mage translation using parallel latent transform networks, Computer Graphics Forum 39(7) (2020), 531–542.

20.

Tian

Guan

Xing

and Fraundorfer

, Efficient ego-motion estimation for multi-camera systems with decoupled rotation and translation, IEEE Access (99) (2020), 1.

21.

Fitzpatrick

Chapman

Grossman

R.C.

and Brady

R.R.

, Randomized controlled trial of plain English and visual abstracts for disseminating surgical research via social media, British Journal of Surgery 7(10) (2019), 181284–181294.

22.

Zhang

Gao

and Lai

, Detail-preserving CycleGAN-AdaIN framework for image-to-ink painting translation, IEEE Access (99) (2020), 1.

23.

Blanks

R.G.

Wallis

M.G.

Alison

R.J.

and Given-Wilson