What are the key differences between human translation and machine translation?

This is guest blog posting from Sharon Wilson.

1. Human translators do the thinking. As they aren’t programmed like MT, they can engage in translating the more complex text which has nuances and slang in it. MT has difficulty with thinking as it only translates what it has been programmed to do.

2. A human translator is flexible when it comes to time and can spend more time translating a text if it’s more complex. On the other hand, MT is programmed at the vocabulary level and can do the translation at breakneck speed, but it doesn’t have the ability to check the translation and write any of the words or phrases differently.

3. A human translator can turn out to be quite expensive so that’s a good reason to turn to MT as it’s much cheaper.

The process used by a human translator

If it’s accuracy you are after you will choose a human translator because a human translator knows when a translation is good and will fulfill a client’s requirements. The first thing a human translator will do is read the source text. From this s/he will see what the text type is, determine the writer’s style and purpose; see what the quality of the writing is like and see how the different parts of the text are arranged. From this information, the translator will know whether the translation falls within knowledge range or whether some research is required to say become familiar with some specific vocabulary.
Current MT software doesn’t have the knowledge to think out of the box so a mish-mash of words could be found in a legal document that isn’t linked properly so the translation doesn’t sound quite right. It’s also quite common in a language for a word to have many synonyms and a human translator will know exactly which one fits the best while MT would have no idea.

Proofreading a translation

Most MT software won’t have the skill to complete an accurate proofread unless they have been well trained in a particular specialized area. A human translator can quickly spot misspelled words and punctuation errors.

Editing a translation

The experienced translator should have the required knowledge of style, punctuation, and grammar for the genre(s) they are translating so if the wrong synonym has been used it will be quickly substituted for a more appropriate one. A machine translator won’t be able to do that.

So, what’s the best, machine translation or human translation?
Machine translation without a doubt is quicker but the translation isn’t so accurate so can’t be relied upon except when only a vague meaning is required.

Author Bio:
A Blogger from Aussie Translations, Sharon Wilson has specialized in writing blog posts and have worked with entrepreneurs, executives, industry experts and many other professionals in writing and publishing, blogs, newspaper articles, SEO web content, and more.

How Accurate is Machine Translation?

This is guest blog posting from Sharon Wilson.






Machine translation is not considered to be as accurate as a human translator when it comes to more complex translations. Typically, it only has the capacity to translate simple word for word translations. The accuracy depends on the languages translated and the complexity of the words translated.

What is Machine Translation (MT)?

MT is a method used independently of a human translator to translate text or the spoken word from the source language to the targeted language. It’s been developed to speed up the translation process and has been especially useful for businesses that are wishing to expand overseas to countries that don’t speak their company language. This trend is set to procure this year.
MT is quite simple software as it typically translates word for word without seeing the connection the word has in the sentence. This means the translation isn’t of high quality as it doesn’t recognize a complete phrase or sentence and how this conveys meaning. Of course, there are situations such as a tourist who wants to find out the meaning of a road sign and inputs it into MT software. The end result will probably be something the tourist can at least understand. Other instances where MT may be useful are tweeting, blogging and texting, answering adverts, booking hotel rooms online and taking part in online forums.

Advantages of MT

  • It’s readily available, like Google Translate;
  • It’s easy to use and fast;
  • It’s either free or low cost;
  • A human can read through the translation and improve it;
  • Translation memories (TM) can be used to recall words that have already been translated;
  • It’s useful for translating web pages.

Disadvantages of MT

There is no hard and fast way of assessing how good MT is as a translator, but depending on the pair of languages subject to a translation there is an accuracy rate of between 60 and 80%. There are other uncertainties about MT too such as:

  • not all content will translate well with MT;
  • different languages translate better, depending on the MT engine;
  • some MT engines are better at translating specific subject matter;
  • most MT tools have difficulty translating slang, technical terms, or nuances;
  • some MT software produces translations that don’t flow well so they sound stilted.

Basically, MT translation can’t be relied upon, but human translators can.

Author Bio:
A Blogger from Aussie Translations, Sharon Wilson has specialized in writing blog posts and have worked with entrepreneurs, executives, industry experts and many other professionals in writing and publishing, blogs, newspaper articles, SEO web content, and more.

eBay Uses Machine Translation to Sell More

Machine translation has its opponents and supporters but everyone agrees that eBay adopting machine translation (MT) is becoming a huge success story. They really are selling more with the help of MT. Continue reading eBay Uses Machine Translation to Sell More

Can I Print Only Selected Parts in PDF Files?

PDF files have been used to send and view documents for quite some time now. But as great as the format is, it has its limitations. While PDF is not always the most flexible format, people have also false assumptions on what cannot be done with a PDF.

One of the most common issues that users face is trying to print off a small section of a larger document. It’s frustrating to have to print an entire 150-page document when all you really need is a few paragraphs from some individual pages. The good news is that there is a way to print PDF that will not only save you a whole bunch of printer ink, but a great deal of time in the process. Continue reading Can I Print Only Selected Parts in PDF Files?

Real Life Example of Software Localization Process

This is a real life example of translating a small Windows application called Multilizer Translator Gadget, which is a Windows Sidebar Gadget. Although every software localization project is unique, there is a universal path which is followed by majority of localization cases. Let’s go through this path with this case.

Pre-Localization Phase

Before any localization project can be started, one needs to define the goals for the project. Localization is not a simple term and to avoid any misunderstandings it is important to clarify what it means in this specific case. In our case the localization was reduced to translating the user interface into multiple languages.

#1 Set up the goals for the project.

After one knows what is the desired outcome of the project, it is recommendable to take a look at the software. There are many things software developers can do to ease the actual localization process. (Due to the technical aspect of this stage, the following few paragraphs contain some technical jargon. If it’s too complicated, just jump straightly to the “Localization Phase” or ask a nerd for some assistance.)

In our case we found out that gadgets typically consist of HTML, CSS, and JavaScript. And the texts to translate can be in HTML files, or sometimes also in the JavaScript. This applied also to the gadget that we developed.

The sources were organized in a typical way in the following folders:


The root contains the HTML and the gadget manifest. CSS contains the style sheets. Images contains the images used by the gadget. And js contains the JavaScript that adds the desired functionality in the gadget.

#2 Check the software architecture.

The Gadget takes a use of the Microsoft MUI architecture; there can be folders named by the ISO locale, such as ‘es-ES’. On a Spanish system the Sidebar would first look for files under ‘es-ES’ directory. If not found, then ‘neutral’ files would be loaded. Based on this architecture, translation would involve that all the files would be copied under a new directory named by the ISO locale. After that the files with texts in English would be translated to another language. This approach, however, is not so elegant. A lot of files were duplicated as such.

To overcome this, we decided to internationalize the code. This means that we removed all English texts and placed them in one common resource  file named ‘resourcestrings.js’. And the user interface is now populated at start-up by calling a getResourceString() function.
With this approach the translation of the gadget is simply reduced to translation of ‘resourcestrings.js’.

We also added two extra functions to make the programming part even more convenient:

  • setElementText(elementid,resourcekey)
  • setAttributeText(elementid,attributename,resourcekey)

#3 Internationalize the code (if possible)

These relatively small changes in code ease and simplify the next phase of the localization project significantly.

Localization Phase

When all the preparative work is done, the actual localization can begin. One can do the localization phase manually or semi-automatically with a special localization tool. Manual work requires a lot more technical knowledge about software development than working with a localization tool. The most significant differences between these two methods are time and security. A localization tool speeds up the process by automating certain tasks and secures the software by eliminating the possibility to make fatal changes to the code during a localization process.

We did our project with our own, advanced localization tool called Multilizer Enterprise. It is a versatile tool for both small and big projects. We first created a new Multilizer project and scanned our software with the tool. Our localization tool recognises which part of the code should be translated and shows only these strings. After that we exported the translatable material automatically to Google Translator Toolkit with MOTO.

#4 Prepare the project for translation.

At this point of the project it’s time to translate. Depending on the goals of the localization project, one should decide how the translation is made. The best output is likely to be made by a professional translator who is specialized on the topic and on software localization. However there are also situations when e.g. crowdsourcing translation is the best solution. Again the goals and targets of the project should be taken into account. In our project we used an online employment platform called Elance to find translators.

The translators worked on our project with Google Translator Toolkit. This kind of online outsourcing is simple for the translators, because they can work online where and when they want to without any need for downloading or installing programs to their own computers. To ensure the quality, there should be a good communication channel between the participants.

#5 Recruit translator(s) and communicate with them.

When the translations are finished, they can be integrated to the code. In our project this was made by importing the translations automatically back from Google Translator Toolkit to Multilizer Enterprise.

#6 Integrate the translations to you project.

When all this have been made, one can finish the project and build the localized software. In this stage there is a temptation to mark the project as finished and move to the next one. However there’s still one important thing to do.

After-Localization Phase

After the software is localized, one should test it properly. It’s a good idea to let the translator(s) see the outcome as well. All the possible mistakes or typos are relatively easy to fix now before any further marketing actions.

This is how our localization project proceeded. Would you do something differently?


Grammar Vs. Understandability: Which Is More Important In Terms of Translation Quality?

This article is originally published at translation-blog.multilizer.com.

Translation quality is a hot topic especially when talking about machine translation. It is true that machine translation is not the same as professional human translation, but can we say that the quality of machine translation is worse than the quality of human translation. The common opinion seems to agree with this claim. However I’m not sure if the issue of quality is this simple at all.

While the use of social media and Internet is booming, the criteria of written text is evolving. Anyone of us can write and publish online no matter how bad our language skills are. It was argued already when SMS was the hot new innovation that the languages will suffer when people start to communicate innovatively. Since then the technological development has taken giant leaps and the world has changed.

I believe that there are different criteria for (translation) quality depending on the place or channel. Grammatical mistakes in a novel are probably received differently than in social media. The target of today’s e-communication is to share ideas in a way that people can understand the message. A major reason behind translation projects is also sharing information to a wider audience. This stretches the limits of the concept of quality.

The quality of translated information is hard to define. Let’s take a real life example:

A person from our development team was the other day involved in translating a manual in PDF format. The manual was related to building an electronic gadget, and the translation was made with Multilizer PDF Translator.

Multilizer PDF Translator automatically translates PDF documents using machine-translation, so the translation quality is not perfect.  Unlike other solutions for PDF translation, however, Multilizer PDF Translator preserves the layout. During the translation project it became clear that the layout of the translated document was more important than perfect grammatical quality.

As a proof for our claim, see the Spanish translation for “Important! Before turning on power, ensure that capacitors C7, C9, and C15 are connected as shown in the diagram.”

Without Image With Image
¡Importante! Antes de encender la unidad, asegúrese de que los condensadores C7, C9, y C15 se conectan como se muestra en el diagrama. Circuit diagram with Capacitors

Importante! Antes de encender la impresora, compruebe que los condensadores C7, C9, C15 y estén conectados como se muestra en el diagrama.

The left text without image has a better translation from grammatical point of view as the right one which is far from perfect. I bet many of us would however choose the incorrect translation if it shares the information in an understandable way.

Have you ever read a grammatically correct text without understanding what it is really trying to say? At least I have.

Translate PDF Files Automatically. Try today!