CMP211 Artificial Intelligence In Language Translation Assignment Sample

I. Introduction to Artificial Intelligence In Language Translation Assignment Sample

Intelligent systems which operate similarly to human thinking, are at the heart of advanced computer use these days. Among the various tasks they handle, artificial intelligence has achieved great results in language translation. In the past, using rules to translate was not very effective since often, both the words and the grammar were wrong or confusing. AI’s introduction, especially by means of neural networks and deep learning models, has introduced major changes in the way translation is done. Thanks to language translation, people from different cultures are able to communicate, international businesses can function smoothly and content is accessible to users of different languages. Because there are more multilingual messages on the internet and people require quick translations, it is more necessary than ever to use accurate and contextual translation systems. Translation tools based on AI have become useful solutions because they are capable, quick and always improving compared to human translation. The rising impact of AI in language translation is not only a technical change, but also comes from the importance of connecting global agents and encouraging digital inclusion. This article is intended to explore what is new in using AI for translation, look at how these advances are put into practice and discuss both the current uses and the prospects for future growth of these methods. Gaining knowledge of these changes helps people recognize how intelligent systems may imitate the intelligence and decision-making processes of humans in complex language-related tasks.

II. Literature Review

According to Mohamed et al., [1], language translation started out with systems that depended on rules of grammar and lists of words. Even though these first[1] Mohamed AI models were pioneers, they could not change with the way human language often changes. SMT was later introduced as an improved option since it uses probability to anticipate translations from bilingual data. Even after progress, SMT had problems with idioms, long-stretched connections and the diversity of syntax in different languages. Neural machine translation (NMT) brought major changes to the industry. Instead of the older approaches, NMT makes use of artificial neural networks to learn the full translation connections all at once. Originally, these systems relied on Recurrent Neural Networks (RNNs), but soon LSTM units were introduced to straighten out the vanishing gradient issue. They included more context in translation and became SMT’s competitors, even though they still had difficulties processing long phrases and complex sentences.

development of the transformer architecture brought about the introduction of self-attention. This new approach meant that the model considered every word in a sentence together which helped it produce fluent and accurate translations. Since BERT and GPT showed excellent performance in understanding and creating natural language, their abilities were soon applied to the field of translation. They might model connections between words that involve idioms and instances when a word’s meaning changes depending on the context. Google Translate, DeepL and Amazon Translate are some of the open-source frameworks and APIs that have applied these models to useful applications. The platforms translate in real-time and they offer support for many different languages. AI in translation has been boosted by new developments in computer processing and the presence of many multilingual data sets.

According to Genovese et al., [3], there are still some problems despite the advances made. Emailing results that are based on biased training data can result in messages with cultural or gender prejudice. Besides, working with user-generated data raises fears over privacy and who owns the data. The challenge of context sensitivity is increased with languages that offer vague grammar or the same word can be interpreted in several ways. Lack of training data on such dialects and languages prevents them from being included in mainstream systems which decreases the level of translation. A further important aspect to consider is how well translated material is captured. Usually, the BLEU, METEOR and TER metrics are used to see how well a machine performs in translation. However such numbers usually do not reflect the human touch in translations like tone, fluency and how culturally relevant they are. New research is suggesting that, in some cases, it’s better to add people’s opinions or use a hybrid system that combines automated and qualitative ratings.

According to Nasrullah et al., [4], discussions about ethical issues in AI have become more common these days. More and more people are worried about losing jobs in translation, using translation tools for wrong purposes and model transparency. Since deep learning models are difficult to understand, it becomes hard to know why certain translation results are produced which makes it harder for people to trust them. Recently, pre-trained multilingual models have been developed in AI for translation, so that they can support translation among many languages with just one model, instead of a specific model for each language pair. The shift is clear with models like mBART and mT5, as they are both faster and perform better when understanding many languages. They have shown that they work well for translation and also for creating, classifying and summarizing multilingual texts.

According to Wang, [5], a new approach is to use reinforcement learning and prompt tuning to make translations in context better. Models in reinforcement learning can adjust themselves using feedback that comes from users, specific tasks or what human people prefer. Sophisticated adjustments of this kind help systems perform well with legal, medical or financial texts which rely heavily on getting both accuracy and terms correct. Studying how zero-shot and few-shot learning works in translation has become important. These methods make it possible for models to translate from and to languages they have not learned specifically, solving the problem of using low-resource languages. Even though the results in these situations are not as high-quality as they could be, further developments can help more people use language technology.

According to Papastratis et al., [6], teaming up with other institutions and groups, tech companies have built major collections of texts in many languages such as OPUS, Common Crawl and FLORES. Because of these resources, models can learn to work well with many different languages. Likewise, steps are being taken to save endangered and indigenous languages by using both digital technology and AI. The inclusion of speech translation in AI has enlarged the number of applications it can be used for. Systems that join ASR and NMT, for example speech-to-text-to-speech, make it possible to communicate orally in real time among different languages without any problems. When past, present and emerging AI systems meet, it supports important sectors such as conferences, emergency management and education, so that their translation is intuitive for users.

All in all, researchers discover that AI has made language translation better than before. Even as technology enhances speed and accuracy, it is still necessary to look into ethical issues, add more languages and better understand different settings. As things develop, translating using AI will become very important, thanks to more understandable machines, personalized solutions and smooth integration across platforms and devices.

III. Experimentation Planning Overview

Evaluating AI language translation systems can be done by carrying out a test that measures the results produced by several translation tools in the same conditions. The experiment will look at translating content with Google Translate and a transformer-based language model integrated with a big language model interface [6]. The intention is to verify the precision, cultural understanding and total quality of the output text for various combinations of languages. A representative dataset of 20 short texts from the five areas of news, healthcare, legal, technical and informal communication was chosen for the experiment. The texts are available in Spanish, Hindi, Chinese and Arabic after being translated from English [7]. These are popular choices because people speak these languages across the globe and their rules can be simple or complex. Both AI tools work on the same translation and the results are gathered to study the differences between them.

Tests use automated systems and manual reviews from people to gauge performance. The main use of the BLEU score is to give a standard number for comparing how much n-gram overlap there is between machine output and human translations [8]. Furthermore, a panel consisting of bilingual evaluators scores each translation based on fluency, accuracy and contextual appropriateness on a five-point Likert scale. The tools necessary for this experiment are an online interface that connects with both translation systems, SacreBLEU for calculating BLEU scores and a worksheet to track scores from the evaluators [9]. With Python, the scoring process can be done automatically and the data shown through bar charts and radar plots.

The content is checked to make sure that it does not reveal any identifying personal data or sensitive information. Before evaluating, the evaluators understand the study and are promised anonymity when they give their feedback [10]. When people translate from other works, there must always be appropriate citation and credit given. There are certain drawbacks with human reviews since people’s opinions differ and knowledge of a specific industry may hinder proper translation. These influences are mentioned and to reduce bias, the investigation draws from a diverse range of articles and makes sure at least two evaluators check each language combination [11]. This approach lets assessors systematically figure out what an AI-based translation system can do and how well it works.

IV. Results

When the experiment happens as expected, it should be obvious that the transformer-based language model works differently from the conventional neural translation tool. In areas where much context is needed such as legal and healthcare documents, the transformer model can likely produce translations of better quality. Thanks to previous studies and the abilities of transformers, the transformer-based system is expected to achieve higher BLEU scores which shows it is more in line with human translations. An example is that in healthcare, transformers often score more than 50 on BLEU, whereas traditional models often stay in the mid-40s. Likewise, translating legal texts may result in similar patterns, as the transformer is able to more precisely keep the original tone and language of the document [12]. Meanwhile, the gap between the systems usually becomes smaller when talking informally or using ordinary language.

Experts may identify qualitative variations between the translations done by machine and those done by humans. Experts believe that the transformer model will achieve better scores for fluency, correct grammar and appropriate language. Experts may observe that this model is able to interpret expressions and references from another culture differently from a straight translation, thus preventing misunderstandings. As an example, turning an English idiom such as ‘Every cloud has a silver lining’ into an Arabic or Hindi phrase that locals would understand. For short sentences that do not include challenging meanings, a traditional translation works well, yet it can struggle with more complex messages. These kinds of errors such as gender disagreements, missing modifiers or clumsy sentence structures, are likely to occur more often with GTAP’s outputs [13]. On the other hand, the transformer model tends to have fewer mistakes and sentences that are more clearly put together.

Showing evaluation scores on a radar plot or bar chart should clearly reveal that the transformer model is preferred for several factors. It would be most noticeable for these languages, where both word formation and grammar are complex: Arabic and Hindi. Transformers can often outperform other models in various areas and keep their accuracy high when dealing with texts made for a specific group. Both systems are likely to have their own positives and negatives, but the transformer-based one is expected to beat the traditional system in providing better quality translations [14]. The results shown here reveal that transformers help offer more accurate and fluent translations while understanding context better, proving why they are used more in today’s language-processing systems.

Conclusion

Realizing a better understanding of language translation with AI clearly shows the huge difference intelligent systems can bring to multilingual communication. Developments in technology, like changing from rule-based to transformer-based approaches, have made machines capable of translating accurately and fluently. Many studies discuss frequent updates to models, metrics and application software, demonstrating AI is maturing in wireless communications. It became clear from planning the experiment and analyzing the results that transformers are more effective at dealing with complex grammar, words used in specific domains and common idioms than other AI approaches. While the results prove that modern AI translates well from one language to another, they also highlight some continuing issues, among them, translation bias, misunderstanding of context and weak capabilities with less common languages.

Society and ethics are still important topics when it comes to using AI for language translation. Ensuring transparency, including everyone and being responsible is key to giving fair access and handling things properly. Experts from many fields such as linguistics, ethics and computer science should join forces to address these obstacles. The development of models that work in multiple languages and have better context understanding, the support of languages with few resources and the enhancement of evaluation techniques to resemble human-like translation should be prioritized in the future. If cultural sensitivity and proper safeguards are built into AI translation systems, the progress in technology will be accessible and suited for everyone worldwide. This study shows that AI has an important part in the future of language translation and highlights the significance of being responsible with AI.

Acknowledgment

“I would like to acknowledge all those who have been of great help to me in the course of completing the dissertation. In the first place, the work would not have been possible without the contribution of (project teacher name) of the university (insert college/university name). I am indebted to (insert other teachers who have some contribution) who have offered continuous support, meaningful suggestions, and criticism while preparing the project. They devoted their time, knowledge and moral support whenever it was needed to this study and its development and for that I am grateful. I am also grateful to all those with whom I had the opportunity to do the work and complete the project. Each member of the dissertation committee has offered and provided professional guidance and have given me great advice while completing the project. I would like to extend my sincere appreciation to my classmates, for their valuable time, valuable advice and natural constructive criticism that helped in developing this study. On a personal note, I am also grateful to my family members who have offered me continuous support while I was completing the project. Without the help and support of them, this project would not have been completed.”

Read My Last Sample: https://www.newassignmenthelp.co.uk/game-theory-in-terms-of-reporting-tax-and-auction-strategies-assignment-sample-53418

References

• [1] Mohamed, Y.A., Khanan, A., Bashir, M., Mohamed, A.H.H., Adiel, M.A. and Elsadig, M.A., 2024. The impact of artificial intelligence on language translation: a review. Ieee Access, 12, pp.25553-25579.
• [2] Kolhar, M. and Alameen, A., 2021. Artificial Intelligence Based Language Translation Platform. Intelligent Automation & Soft Computing, 28(1).
• [3] Genovese, A., Borna, S., Gomez-Cabello, C.A., Haider, S.A., Prabha, S., Forte, A.J. and Veenstra, B.R., 2024. Artificial intelligence in clinical settings: a systematic review of its role in language translation and interpretation. Annals of Translational Medicine, 12(6), p.117.
• [4] Nasrullah, N.S., Herdina, G.G.H. and Aini, N., 2023, April. Retrans: Regional Language-English Translator Machine Based on Artificial Intelligence (AI) for The 10 New Bali Regions. In Proceeding International Conference on Religion, Science and Education (Vol. 2, pp. 767-771).
• [5] Wang, Y., 2023. Artificial intelligence technologies in college English translation teaching. Journal of psycholinguistic research, 52(5), pp.1525-1544.
• [6] Papastratis, I., Chatzikonstantinou, C., Konstantinidis, D., Dimitropoulos, K. and Daras, P., 2021. Artificial intelligence technologies for sign language. Sensors, 21(17), p.5843.
• [7] Malik, A.R., Pratiwi, Y., Andajani, K., Numertayasa, I.W., Suharti, S. and Darwis, A., 2023. Exploring artificial intelligence in academic essay: higher education student's perspective. International Journal of Educational Research Open, 5, p.100296.
• [8] Kenny, D., 2022. Machine translation for everyone: Empowering users in the age of artificial intelligence. Language science press.
• [9] Alyasiri, O.M., Selvaraj, K., Younis, H.A., Sahib, T.M., Almasoodi, M.F. and Hayder, I.M., 2024. A survey on the potential of artificial intelligence tools in tourism information services. Babylonian Journal of Artificial Intelligence, 2024, pp.1-8.
• [10] León, C.O., Allcca, F.Y. and Claudio, B.M., 2023. Using Artificial Intelligence for sign language translation: a systematic literature review. Salud, Ciencia y Tecnología-Serie de Conferencias, (2), p.223.
• [11] Moons, P. and Van Bulck, L., 2023. ChatGPT: can artificial intelligence language models be of value for cardiovascular nurses and allied health professionals. European journal of cardiovascular nursing, 22(7), pp.e55-e59.
• [12] Aziza, S., 2024. Artificial Intelligence in Translation: Benefits and Drawbacks. International Journal of Scientific Trends, 3(11), pp.70-76.
• [13] Baidoo-Anu, D. and Ansah, L.O., 2023. Education in the era of generative artificial intelligence (AI): Understanding the potential benefits of ChatGPT in promoting teaching and learning. Journal of AI, 7(1), pp.52-62.
• [14] Jiang, R., 2022. How does artificial intelligence empower EFL teaching and learning nowadays? A review on artificial intelligence in the EFL context. Frontiers in Psychology, 13, p.1049401.