Comparison and evaluation services for machine translation (MT, Machine Translation)

Comparison and evaluation services for machine translation (MT, Machine Translation)

Human Science proposes Machine Translation (MT/MachineTranslation) as one of the solutions to challenges such as reducing translation costs and compressing delivery times.
While it differs from the quality of human translation, it is possible to translate large volumes of manuals and documents in a short time and at low cost, depending on how the "quality standards" are defined.


For more information about our machine translation services in human science, please visit this machine translation feature website.

Cost reduction and shortened delivery time with machine translation (MT, Machine Translation)

Machine Translation (MT) has been a significant trend in the translation industry in recent years. It is said that machine translation is at a practical level, especially for European languages.

It is effective not only for cost benefits but also when speed is required for simultaneous product releases in Japanese, English, and multiple languages.



Case Studies

  • ・Office Equipment Instruction Manual
  • ・Languages: French, Italian, German, Spanish
  • ・Word count: Approximately 75,000 words
  • ・Corpus: Approximately 45,000 segments
  • ・Quality Standards Setting: At the same level as human translation



Cost Reduction Rate and Delivery Time Reduction Rate

Machine Translation_Delivery Time Reduction_2 Machine Translation_Cost Reduction



By focusing on key points, you can build a machine translation workflow.

  • 27.5% Cost Reduction
  • 48% Reduction in Delivery Time

Achieved.

Initial costs such as the introduction of the MT engine are not included.

*The cost reduction rate may vary depending on the document to be translated, the desired translation quality, and the status of past translation assets.


For more information about our machine translation services in human sciences, please visit this machine translation feature website.



Machine translation is not suitable for all translations – To succeed with machine translation -

Machine translation is not suitable for all texts.



It is believed that machine translation is most effective when applied to materials with a limited target audience and a large volume of text, such as support information web pages and documents for developers.

At Human Science, when considering the introduction of machine translation, we leverage our extensive experience with various machine translation systems to propose document analysis and solutions tailored to the target users.

  • Benefits include lower costs and faster delivery.
  • Disadvantages may include negative impacts to customer satisfaction, brand image, and after-sales support.

We will compare and examine the introduction of machine translation from various perspectives, including the aforementioned viewpoints.



Machine Translation of Human Sciences: Comparison and Evaluation Services

The optimal engine and process vary depending on the desired quality level. The real key to a successful machine translation lies in the agreement between client and LSP on how much the automatic results should be edited. Human Science will propose the best solution based on our extensive achievement record.

 

Services Offered

  • Machine translation implementation consultation
  •   ・Setting Quality Standards
     ・Selection of MT Engines
     Select and evaluate 2 to 3 types from 5 types of rule-based and statistical MT engines
     ・Cost-effectiveness Verification
     ・Workflow Construction
  • Pre-edit of source text to improve machine translation quality
  • Post-edit
  • Machine translation quality evaluation

Workflow Example

Machine Translation_Workflow

Key Points for Successful Implementation of Machine Translation

Selection of MT Engine

We will select and propose an engine that meets your requirements, including the items to be translated, volume, and implementation and operational costs.

We will select 2 to 3 types from a total of 5 types, including 3 statistical-based MT engines and 2 rule-based MT engines, and report and propose the evaluation of machine translation to our customers.

*The quality varies depending on the amount of TM (corpus).

*The quality varies depending on the amount of TM (corpus).


Rule-Based Characteristics and Benefits

  • ・It is relatively easy to introduce machine translation by simply preparing a glossary and example collection.

Characteristics and Disadvantages of Statistical Machine Translation

  • - It is necessary to collect a large amount of parallel corpora and perform tuning.
  • ・Requires a significant amount of time and cost for tuning
Translation memory (TM)
use
By combining traditional TM data with machine translation, we can utilize existing TMs without wasting them.
For passages where we can reuse previous translations, we will use translations from the TM. And for new passages that require translation, we will use automatically translated text. This optimizes project efficiency while ensuring consistency with existing translations and cutting costs.
Dictionary use
By continuously utilizing dictionaries, we can improve the quality level of translated texts. Existing glossaries can also be incorporated.
Post-editor training
Not only do post-editors need to understand the characteristics of machine translation, but they also need the ability to quickly determine where and how to correct automatically translated sentences to make them accurate. Post-editing is not something that just anyone can do; it requires skilled translators or reviewers to perform the task. Our company has established a training program to develop our post-editors.
Post-editing
guideline development
At Human Science, we have established our own guidelines for post-editing based on our previous experience with MT. With this as our foundation, we provide post-editing services tailored to our clients' needs while ensuring a consistent level of corrections after machine translation.

Machine translation consulting achievements

Overseas Software Manufacturer
  • MT engine and process consulting for knowledge base translation and survey response translation in various languages
  • Language: (Knowledge Base Translation)
  •     English → French, Italian, German,
    Chinese (Simplified, Traditional), Korean
  •    (Survey Response Translation)
  •     Japanese, French, Italian, German, Chinese, Portuguese
  •     Chinese (Traditional, Simplified), Korean → English
Domestic Medical Device Manufacturer
  • MT engine and process consulting for translation of user manuals in various languages
  • Languages: > English → French, Italian, German, Spanish, Portuguese

 

Overseas MT Engine Vendor
  • Quality Assessment Survey of MT Engine
  • Languages: English → French, Italian, German, Spanish

For more information about our machine translation services in human science, please visit this machine translation feature website.



For those who want to know more about translation into English and other languages

Tokyo: +81-3-5321-3111 Nagoya: +81-52-269-8016

Reception hours: 9:30 AM to 5:00 PM JST