www.witinall.com
Your German language service
  • Home
  • About
  • Contact
  • Terms
  • AGB
  • Blog
  • Shop
  • Languages
  • Site Privacy Policy
  • Ask the Experts

Have your say...

No nasty comments, please!

email

Machine translation: why you shouldn't make it too easy for the machine.

7/26/2018

2 Comments

 
Picture
#5 Punctuation

Did you know that a typo can land you in jail? Oh yes, you could end up like this poor fellow:

       Walker sorgt sich um die Sicherheit seiner Freunde und wird des Mordes beschuldigt.
      
[Walker is concerned about his friends’ safety and is charged with murder.]

And just because someone left out a pair of commas:

       Walker concerned about his friends’ safety gives himself up and is charged with murder.

In the original (source text) sentence, Walker gives himself up to the police. There is no mention of that in the translation. In fact, the translated sentence taken out of context suggests that his being concerned for his friends is the reason or cause for his being charged. (I once saw a similar sentence in which a rocket was being built, which ended up being fired (!) in the translation. More than wishful thinking.) What ever happened? 

Put the commas where they should be and the reason becomes clear (and you also get a perfectly acceptable translation): 






​


​
Losing bits of information is one of the most common ‘errors’ in automated translation, and one of the most annoying. And the reason is not always an erroneous input. Let’s look at another example, taken from the Wikipedia page on the serial comma:

       My usual breakfast is coffee, bacon and eggs and toast.

Google translates this correctly as: 
​
      Mein übliches Frühstück ist Kaffee, Speck und Eier und Toast.

What happens if we decide do add the serial comma that groups the bacon with the eggs?
​ 
       My usual breakfast is coffee, bacon and eggs, and toast.

Well,…nothing. It’s still:

       Mein übliches Frühstück ist Kaffee, Speck und Eier und Toast. 

The machine doesn’t understand cultural references of any kind. It doesn’t know how ‘bacon and eggs’ are eaten or that in German you could use the phrases ‘Eier mit Speck’ or ‘Speck mit Ei’ to indicate the proximity of the foodstuffs on the plate. All it does is look up the words, the frequency of appearance in a corpus and the most likely combination. Which is why the serial comma is simply ignored: it isn’t mandatory and so isn’t used by all speakers of English.

You can leave out commas to your heart’s content where there is no danger of a misunderstanding:

       Your new financial friends will be abstract, obsessed with numbers [,] and keen on comparing you with your                  competitors.

But (!) be aware, that the machine has a creative mind of its own and will add a little variety:

Ihre neuen finanziellen Freunde werden abstrakt sein, von Zahlen besessen und darauf bedacht sein, Sie mit Ihren Konkurrenten zu vergleichen. 

Deine neuen finanziellen Freunde werden abstrakt sein, von Zahlen besessen und darauf bedacht sein, dich mit deinen Konkurrenten zu vergleichen.

Apart from the fact that ‘financial friends’ is not normally translated as ‘finanzielle Freunde’, I don’t see why the comma should lead to the informal ‘du’ being used all of a sudden. Do you?

Commas also serve to indicate relative clauses. In the following sentence the noun ‘fingers’ is modified by a relative clause:

The flame on the match head danced and jumped in her fingers, which had yet to stop shaking.

The translation I received from one of the free programs available on the net was:

Die Flamme auf dem Spiel Kopf tanzte und sprang in den Fingern, die noch zu schütteln aufhören musste.

You had better be able to read German if you wanted to use this. Not only is ‘Spiel Kopf’ (not even a word) not even close to ‘match head’, but there is no subject-verb agreement, i.e. the number of the finite verb ‘musste’ is singular rather than plural (‘fingers’)! Well, you could always count this as a typo, I guess. ;-)

Perhaps the comma in the previous example caused some confusion. Let’s now look at the parentheses (or round brackets) and dashes indicating an interruption in the following sentence:

They don’t have to reveal their agenda (are they long or short your shares?) and try to tease out gossip with which to speculate.

translated as:
Sie müssen ihre Agenda nicht offen legen (sind sie long oder short ihre Aktien?) Und versuchen, Klatsch zu entlocken, mit dem zu spekulieren.

And with a dash (single or double):

They don’t have to reveal their agenda - are they long or short your shares? - and try to tease out gossip with which to speculate.

translated as:
Sie müssen ihre Agenda nicht offen legen - sind sie long oder short ihre Aktien? - und versuchen, Klatsch herauszupicken, mit dem man spekulieren kann.

In the second version (the one with the dashes) the ‘und’ is correctly reproduced in lower case and the clause ‘mit dem man spekulieren kann’ has the correct finite verb form ‘kann’. There is no obvious reason for the lexical variation. Perhaps you have an idea. I’d love to hear from you, so just write your comments in the field provided below. 

Picture
2 Comments

Machine translation: why you shouldn't make it too easy for the machine

7/26/2018

2 Comments

 
Picture
#4 The active and the passive voice

The best advice for budding writers and translator I’ve come across is that you should go over your text with fresh eyes and ruthlessly cut out anything you feel is redundant or superfluous. For me, it’s often cut, cut, cut, with my grey cells screaming, ‘You want to get rid of WHAT?! You CANNOT be SERIOUS!’ So, when I write in this blog I tend to just let loose and give in to the inspiration of the moment without regard to good writing practice. I hope you will forgive me. And I hope you’ll find something interesting to read. Continuing in the series on machine translation this week, I would like to talk to you about two interesting things I’ve noticed coming up time and again. First, let’s look at the passive voice.   

You can make writing clearer by using verbs in the active voice and avoiding the passive. For instance, if you want to make it clear who is doing (or supposed to be doing) an action:

       Make sure you update your virus protection programs regularly.
       Do not continue to press the On/Off switch after the machine has been automatically switched off.
       Before you can claim your tax refund, you will need to verify your identity.

This is a sensible precaution when conveying vital information. Passive voice verbs can keep readers guessing as to who is doing what if no ‘agent’ (or ‘doer’) is mentioned. Consider the following sentences:

       It must be insured that the virus protection programs are updated as quickly as possible on all computers when a        new virus is detected. 
       Appropriate measures must be prepared and tested.

Do you know who is doing the action? Do you know who is supposed to be doing it? (Could be you.) What about the following example?

       If the cable is damaged or cut during work, do not touch the cable.

It seems we don’t have to know. This is because sometimes the context makes it irrelevant as to who is doing the action conveyed by the verb. For example:

       Do not continue to press the On/Off switch after the machine has been automatically switched off.

You don’t need do know who did the switching off, the machine itself or maybe some external electronic system. 

​Let’s now look at a different example and ask the computer to do some translating for us.

       The tax may be refunded if the traveller has exported the purchased goods outside the European Union not later          than on the last day of the third month following the month in which the purchase was made.



​





​The English sentence contains two passive voice verbs (‘refunded’ and ‘made’) and a verb in the active voice (‘exported’) plus an agent given as ‘the traveller’.

       The tax may be refunded if the traveller has exported the purchased goods outside the European Union not later          than….

The traveller is clearly doing the exporting. What about the ‘refunding’? Do we need to know who is doing it? No, but we are told something about the restrictions that apply:

       …if the traveller has exported the purchased goods…not later than on the last day of the third month in which the        purchase was made.

Again, the passive ‘was made’ does not tell us anything about the agent of the action – it could be ‘the traveller’ or somebody else –, but we do know that it was a ‘purchase’ that was made.

And now let’s put the machine to the test with a sentence that isn’t formed quite so well. The following English sentence came up in a text recently:

       A tax refund can be made if the traveler purchased goods exported outside the EU no later than the last day of the third month following the month in which the purchase was made.

As you can clearly see, there has been a mix-up and the sentence no longer makes sense. A human being will recognise this error and rephrase the sentence accordingly because he or she understands the meaning. The machine, by contrast, takes the clauses provided and translates them literally to read:

       Eine Steuerrückerstattung kann erfolgen, wenn der Reisende Waren, die außerhalb der EU ausgeführt werden,          spätestens am letzten Tag des dritten Monats nach dem Monat, in dem der Kauf getätigt wurde, gekauft hat.

As in the source text, the goods are being purchased after having been purchased. Makes your head spin, doesn’t it? Could these issues have been avoided by putting everything in the active voice, for example, ‘You can claim a tax refund if…’? Maybe.

In any case, before you hand your text to the machine it makes sense to:
  • determine whether, if at all, you need passive verb phrases;
  • turn them into active phrases to ensure that the person doing the action is clear;
  • avoid errors which will be translated literally.

Easier said than done? Well, it’s usually not too difficult. And it’s worth it. Next time, I’ll tell you about how leaving out seemingly insignificant commas will make the machine change the meaning of your text completely. Lots of fun to come!

Picture
2 Comments

Machine translation: why you shouldn't make it too easy for the program

7/16/2018

0 Comments

 
Picture
3: Chopping up snakes 

One of the tenets that most proponents of an easily comprehensible style of writing seem to agree on is that

     sentences should not be too long.

The Plain English Campaign, for example, recommends that clear writing should have an average sentence length of 15 to 20 words. Complex sentences can cause havoc because the structure and meaning may not be clear.

​Let’s put a machine to the test. The following German example (in box on right) contains four short sentences.

As you can see, Google Translate is quite capable of delivering a correct English version that makes perfect sense.

​What happens if we turn three of those sentences into one long ‘snake’ of a sentence? Let’s add some conjunctions:

       …eine einfache Deinstallation, denn die Macher…, sondern bauen eher darauf…

Again, the result is satisfying:

Picture
​









​Note how the program even offers a bit of stylistic variation by differentiating between “are likely to believe” and “rely on”.

This result doesn’t really surprise, given that machine translation tools are being developed with ever greater computing power “brains” and learning mechanisms modelled on human neural networks. Services such as Amazon Translate are cheap to use and capable of handling large batches at high speed. Cryptic product descriptions such as the following will soon be a thing of the past:​
Picture
How about fiction? Let’s put another free tool to the test:
Picture
This isn’t such a long sentence – and you could find longer ones, containing several subordinate clauses, in some kinds of fiction or philosophical texts, for example –, but let’s not dwell on that. What we can see is that syntactic complexity is handled quite well.

So why do blunders and howlers still occur? Any idea? And why is the following sentence (from the AWS website) not perfect? (Clue: it has to do with the translation of ‘simply’.)
      
Wenn Ihr kostenloser Nutzungszeitraum abgelaufen ist oder Ihre Nutzung das Kontingent übersteigt, zahlen Sie einfach die üblichen     nutzungsbasierten Preise.


Send me your answers and let’s start the discussion. Alternatively, come back in a few weeks’ time when we’ll take the topic further to talk about nominalizations, semantics (vocabulary) and the role of punctuation in machine translation. And on how to avoid losing vital bits of information. See you then!

References:
Duden-Ratgeber Technische Dokumentation. Beschreibende und anleitende Texte erstellen. 2012. (von Andreas Schlenkhoff in Zusammenarbeit mit der Duden-Redaktion) ISNB: 978-3-411-74721-4
Martin Cutts. 1996. The Plain English Guide. How to write clearly and communicate better. ISNB: 0-19-860049-6
Picture
0 Comments

Tips to Improve Productivity of Underperforming Employees

7/2/2018

0 Comments

 
by Samuel Gitukui

There are many situations in the workplace that can affect the performance of employees. Conflicts are one of them. For some managers rushing in and firing an employee seems to be the first and only way to solve the issue.

However, the best business leaders seek to understand the underlying issue and find solutions. In this way they can increase the motivation of the employee and boost their performance. Let’s find out a few ways on how to improve the motivation for an underperforming employee.

       1.      Ask for a Private Meeting

The best way to handle an employee is by asking them in for a one-on-one meeting. You want to avoid addressing the issue in front of peers and colleagues, as that can create resentment and cause them to get into a defensive state.

Instead ask them to the office and talk about what is causing their lack of motivation. The employee is more likely to open up. Once you have leant the issue, take time to find the best way to handle it. This makes the employee feel part of the family and is itself the first step towards making progress.

       2.      Analyzing the Problem

Some issues are easy to solve. If the employee is regularly arriving late, then reversing this situation is only a matter of managing time. Helping the employee understand what he is doing wrong should not be a lengthy process.

For some problems however, it’s a bit more complicated. Assume a situation where one of the employees doesn’t get along well with the rest of the colleagues.

This will require you to take time to decide how best to approach the matter before calling the employee back in for a second meeting.

       3.      Some Issues Can be Ignored

The best leaders will tell their employees what to do as well as how to do it. However, it is not a bad idea to let employees find their own solutions. Some employees will have a messy desk and look disorganized but could end up being quite productive. 

Sometimes by letting such employees find their own solutions, you are able to determine whether they are affecting the productivity of the company. If not, then perhaps the issue is not worth calling them in for a meeting.

       4.      Don’t Waste Time

When a problem needs to be addressed, you need to do so with utmost speed. This will prevent the problem from persisting and spreading onto other employees.

When you solve the issue and the employee does in fact show progress, you also communicate to the rest of the workers that your instructions are to be adhered to and taken seriously and that you are not willing to accept complacency. 

       5.      Don’t Make it Personal
​

Problems need to be handled the right way and this means avoiding making the employee feel like you are attacking them. Pass on your views and suggestions in a mature and calm way that does not create resentment in the employee.
Picture
0 Comments

Machine translation: Why you shouldn’t make it too easy for the program.

7/2/2018

5 Comments

 
#2: Simplified English, Plain English & Co.

The basic idea behind writing in a simple style is not new and has been taken up by educators and technical editors worldwide. Simplified (technical) English has been around for a long time, helping staff in technical domains pass on their knowledge (in instruction manuals, product descriptions, safety guides etc.) to non-native users of English. Basic English once found a place in education as a simple style suitable for students learning English as a second language but has been widely criticised as being too restricted. It was consequently superseded by plain English in modern times, a much ‘broader’, much more flexible and versatile concept. Plain English was devised with the aim of making documents, especially legal and government documents and forms clearer and easier to understand. In Germany, the concept of Leichte Sprache (engl. easy language) is based on a similar premise, with accessibility and the inclusion of readers with disabilities very much in the foreground.

What these approaches have in common is that they provide guidelines for creating written texts in such a way that they will be more easily readable for a wide range of readers and thus more easily comprehensible to them. By restricting the use of certain constructions and terms, and by being consistent in the use of a preferred vocabulary and syntax, plain language is thought to become more easily recognisable. In other words, a certain level of standardisation can help make life easier for readers and translators alike. But what about machines? 
​
Picture
5 Comments

    Authors

    Alexandra
    Matthieu
    ​Sarah
    ​Samuel
    ​Summer
    ​Mike

    Archives

    May 2023
    April 2023
    March 2023
    January 2023
    November 2022
    September 2022
    August 2022
    July 2022
    May 2022
    March 2022
    February 2022
    January 2022
    December 2021
    November 2021
    October 2021
    August 2021
    July 2021
    June 2021
    April 2021
    March 2021
    February 2021
    January 2021
    November 2020
    October 2020
    September 2020
    August 2020
    June 2020
    April 2020
    March 2020
    January 2020
    December 2019
    November 2019
    October 2019
    September 2019
    July 2019
    June 2019
    May 2019
    April 2019
    March 2019
    February 2019
    January 2019
    December 2018
    November 2018
    October 2018
    September 2018
    August 2018
    July 2018
    June 2018
    May 2018
    April 2018
    March 2018
    January 2018
    August 2017
    March 2017
    February 2017
    October 2016
    May 2016
    April 2016
    February 2016
    January 2016
    December 2015
    December 2014
    September 2014
    February 2014
    January 2014
    September 2013

    Categories

    All

    RSS Feed

    Link to Delicious
    English Dictionary;
Proudly powered by Weebly