How to choose a quality and efficient machine translation (MT) tool?
Published on 10 August 2020 by Alice Chan
1. Identify the text type of the source and select a MT tool aimed at the matched domain
The ongoing problems facing by the users of general MT is mainly about its inappropriate wordings and unnatural tone. Before using general MT, you must first accept that sentences, phrases or words are easily misunderstood by general applications, particularly in professional and specialized fields with frequent usage of industry jargons such as AP, EBITA and ESG. Therefore, choosing a suitable MT tool aimed at the matched domain is essential to reveal the closest form of the source text by applying more accurate and specialized vocabulary and style. Also, natural language processing process (NLP) is used for analyzing large amounts of natural language data to enhance customer experiences.
DeepTranslate is designed to translate financial documents, especially for those are requested by the Stock Exchange of Hong Kong. Our database consists of a mass volume of financial-related sources particularly from the Stock Exchange of Hong Kong and other aspects surrounding the stock market. With the adoption of a “deep learning” practice, DeepTranslate system can spot the recurring patterns among sentences, phrases and words in this specialized domain through repeated calculations of financial materials, in order to achieve higher accuracy of financial translation than other general translation tools. Take the headline “董事长报告书” of financial reports as an example, our C-E translation is “Chairman’ Statement” while others machine translation services untranslate and keep the Chinese headline or produces an inaccurate translation of “director’s report”.
2. Use a MT tool with translation memory (TM)
A more efficient way of getting the best translation quality from MT is using TM technology. It acts as a quality control tool of the translation system which can be applied for the checking of completeness, formatting and glossary by storing the translations which have already been edited by human translators and reusing in the future. The MT of relative words and phrases will be replaced with the finetunes made in the vetting process. In other words, all MTs of the same sentences, phrases and words in future can be regarded as already being edited by human translators. However, it may take long time for companies to build their own TM for all of their financial reports in the past.
DeepTranslate’s AI experts and linguists can build the TM and provide customized trainings for your TM to meet your needs. Our TM service can save you a considerable amount of time and money on translation and post-editing. With excellent features including but not limited to data security and formatting preservation, DeepTranslate provides efficient and safe AI translation services.
3. Understand the principles and applications of the MT tools before selection
More translation tools on market are starting to transit to a neural MT system from the rule-based one in order to improve the translation quality. The latter approach requires a time-consuming handling process to add syntactic and grammatical rules by experts and linguists which leads to very complex system with not much data.
With in-depth understanding of the importance of AI technology to translation, DeepTranslate has adopted powerful and professional “deep learning” from its establishment. Our AI system models sentences based on predictive word sequencing which is processed by a large artificial neural network on the basis of a mass volume of sources from the stock market. This can bring our translation closer to actual speech of the financial world.
