Future Development Trends

Baidu, Tencent, Ali, Sogou, Youdao, iFLYTEK and a large number of Internet enterprises are vigorously promoting the concept of "AI+" in voice search, human-computer interaction, intelligent translation and other fields, and have taken the development of intelligent translation business as a new entry point and a new hand to drive the future development of business, which will objectively promote the great development of translation technology. The great development of AI technology has injected a strong driving force for the development of translation technology, and the translation technology in the new era presents many new changes, which deserves the attention of the academic community.
1. Continued integration of resources
Language services are on the path of transformation towards digitization and intelligence. New technologies represented by cloud computing, big data, AI, IoT and blockchain drive the opening, flow and sharing of translation data. Language resources have become the most critical means of production for translation technology enterprises. Language resources are recycled in the translation process, which brings new power and direction to the language service market. By processing and integrating massive language resources, translation technology providers analyze the inherent laws of translation data, optimize translation models, and release the value of data. Intelligent translation systems need to continuously acquire new translation data for continuous and in-depth learning, and resource integration of translation technology has become a general trend.
Machine translation providers build translation ecosystems by continuously integrating translation resources. Take Baidu Translation as an example, Baidu Translation provides text translation, text translation, AI simultaneous interpretation, video translation, subtitle production, video editing, human translation, translation plug-ins, app translation, web page applets, and bilingual dictionaries. Its machine translation, on the other hand, can be subdivided into verticals such as biomedical, electronic technology, and hydraulic machinery. Similar to Baidu Translation, Ali, Sogou, Youdao, Yunyi, Xinyi, Maverick Translation, MedPeer, etc., are committed to integrating multimodal translation resources, building an integrated resource ecological platform, and effectively avoiding the waste of technical resources.
2、Functions continue to expand
The complexity of translation needs inevitably requires the complexity of translation tool functions. Translation tools are upgraded from single-function to multi-function or full-function to cover the diversified and multi-level translation needs of clients. Translation technology providers seamlessly integrate data processing, translation search, memory matching, content recommendation, machine translation, post-translation editing, real-time hotspots and other functions to provide an integrated service portfolio.
In the complex translation assistance environment, translation tools and translation environment are highly integrated, and CAT software provides a broad application scenario for various types of machine translation engines. In the machine translation post-translation editing mode, the translator (Post-Editor) edits, improves and confirms the translation in the visualized editing environment according to the rules and processes of post-translation editing. With the optimization of the post-translation editing environment, this model gradually evolves into "MTM+PE (Machine Translation + Translation Memory + Post-translation Editing)" . More and more CAT tools (e.g., SDLTrados, memoQ, Memosource, MateCAT, Wordbee, etc.) deeply integrate the multimodal assisted translation mode to provide translation quality assessment, post-translation editing time, post-translation editing percentage, pre-translation and post-translation editing differences, and new media localization.
The deep involvement of AI technology has led to an increase in the automation level of translation technology services, with the "CAT+TMS+CMS "1 model being the most prominent. This model takes auxiliary translation, translation management and content management as the basic information architecture, highlighting collaboration and integration. Once the system detects a project update, it will automatically extract the new content and process it in accordance with the pre-defined workflow, including word count, duplicate matching, difficulty analysis, task assignment, task notification, etc., so as to reduce the workload of translators and project managers and shorten the project cycle. Taking Cloud Translator as an example, the system can realize online collaboration and freelance translation workflow, and with the support of AI technology, it allows robots to translate with translators and teams as partners, and deeply integrates in multiple aspects such as manuscript allocation, management, and translation collaboration.
3. From desktop to cloud
With the development of complex language service projects, the demand for translation data processing has exploded, and the demand for software and hardware management and maintenance has increased dramatically, the traditional desktop-level CAT solution has highlighted serious problems, such as high overall cost, information fragmentation, high management risk, poor flexibility and other issues. The "cloud integration" translation solution has gradually become mainstream, solving the original high cost problem. With the large-scale application of cloud OA, cloud CRM, cloud ERP, cloud SCM② and other technologies and products in various industries, the trend of cloud-based translation technology is becoming more and more significant. From desktop-based CAT tools to web-based versions to various types of apps, the cloud translation model is making its presence felt in the field of language services.
In the cloud translation service model, the system foregrounds intelligent algorithms, liberates the center's computing resources, accelerates processing speed, realizes flexible application, simplifies network and storage configuration, and allows users to use it directly by simply logging in. Cloud translation technology integrates private clouds, cloud computing interfaces, cloud shared resource platforms and cloud language service industry chain, which can substantially improve translation production efficiency. The cloud translation platform will intelligently adapt to the scenarios, and tends to be lightweight, SaaS ③ and ecological development. With the technical characteristics of "Internet+" and the advantages of mobility and socialization, language service enterprises will provide more and better services at lower cost and in shorter time. In the future, the translation industry cloud ecology will focus on industrial integration, promote the change of translation business models and the construction of digital platforms and ecosystems; the translation application cloud ecology aggregates advanced cloud applications and provides one-stop sustainable cloud application services; and the translation technology cloud ecology provides a comprehensive cloud translation technology platform. The cloud translation technology is more open and the applications are more diversified, which will bring brand-new vitality to the whole language service industry.
4. From human to machine translation
Based on the degree of intelligence, AI is usually broadly categorized into weak AI, strong AI and super AI. Weak AI (Artificial Narrow Intelligence) focuses on accomplishing a specific task, specializes in a single aspect of skill, and does not have the ability to think, such as the AlphaGo system, which beat the world Go champion, and Google Translate, a fast machine translation system.Strong AI ( ArtificialGeneral Intelligence) is similar to human-level AI, capable of independent thinking, abstract thinking, understanding complex ideas, learning quickly and from experience, etc., and will have its own values as well as a worldview. According to Nick Bostrom (2014:22), an Oxford philosopher and leading AI thinker, "Artificial Superintelligence (AI) is much smarter than the smartest human brains in almost all domains, including the domains of science and innovation, general literacy, and social skills. " From the perspective of automation and intelligence in translation, the translation model can be divided into four stages of development (e.g., Table 1), each with different characteristics and degrees of automation.

Table 1 Characteristics of translation patterns at different stages Stage Translation characteristics

Limited by technology and data and other issues, we are still in the stage of simple human-computer interaction translation in the era of weak artificial intelligence, and the future development of AI technology will substantially liberate the amount of translation labor. From traditional human translation to machine translation post-translation editing mode, the degree of human-machine interaction is getting higher and higher. With the innovation and breakthrough of the Internet of Everything and brain-computer interface and other cutting-edge technologies, the highly intelligent translation system will automatically connect all the resources needed for translation, giving full play to the advantages of AI empowerment, and the translator's wisdom will be focused on more creative work.

Reference: [1] Wang Huashu and Wang Xin. Research on Translation Technology in the Age of AI:Application Scenarios,Existing Problems and Trend Prospects[J]. Foreign Language,2021,37(01):9-17.