‘How can I translate a letter I’ve received in Spanish? ‘ Some useful tools and their pitfalls.
Christine Betterton-Jones – Knowledge junkie
Translation tools have improved enormously over the past couple of years, and there are some great new ones which use artificial intelligence and big data to get better and better results. However, language is a very complex thing and there are still many pitfalls.
If you only want to get some idea of what the letter says, the easiest way is to use the Google or the Microsoft Translate (Bing) app on your smartphone. Apple has its own app which runs on iphones.
Using a smartphone
The Google Translate app is probably the most powerful of the smartphone apps available. It can scan printed signs and printed text using the camera as a scanner. Once it identifies (or you tell it) the source language, the app recognises the words and unusual characters (like the ñ) and translates the text into a target language on the fly.
Other powerful features:
- Translate typed and handwritten text
- Convert the translation to speech
- Translate text in a digital photograph
- Translate dictation to text in the target language on the fly (transcribe)
- Translate dictation to spoken voice in the target language on the fly (conversation)
There are many other similar apps in the Playstore – most of these use the Google Translate “engine”
Online translation through your browser:
This is mostly applicable to copied and pasted text, or a letter you received as a DOC or editable PDF … it’s a bit more complicated if the letter is on printed paper!
In the latter case:
1. You need to convert the letter into a digital image by photographing it or scanning it.
2. Then get software which understands Spanish (or Valenciano!) to recognise the characters and words in the document and covert them to editable text in a process known as Optical Character Recognition (OCR). This can be done with some scanning software (e.g. NAPS)
Luckily there are free online services which can do his for you. Just upload the digital image to the website. e.g.
https://convertio.co/ocr/spanish/
…this can also do Catalan ( close to Valenciano !)..basque, galician….and loads of other languages.
Supported image file formats are: PDF, JPG, BMP, GIF, JP2, JPEG, PBM, PCX, PGM, PNG, PPM, TGA, TIFF, WBMP
There are many other free OCR online services with different capabilities. e.g.
3 Then download the foreign language document as editable text or in editable PDF format. (It is worth checking the text against the original letter to see if there are any mistakess
4. Now that you have the foreign language letter in editable text format, use an online translation tool to translate it.
This site summarises some of the major ones (Note: the article is from 2018)
https://lingohub.com/blog/2018/11/find-good-machine-translation-engines
The top three today are:
- https://translate.google.com Google translate
This can manage both text and uploaded documents in the following formats: .doc, .docx, .odf, .pdf, .ppt, .pptx, .ps, .rtf, .txt, .xls, or .xlsx 110 languages
- https://www.bing.com/translator Microsoft Bing .
This can translate text in more than 60 languages
This tool can manage text and uploaded docx and pptx documents. It also has a free windows desktop app. Its translations read more fluently than some of the others. 26 languages to date.
Some sites enable you to compare the results from different tools e.g. https://translation2.paralink.com You can compare the results for pasted or typed text translated by Google, Microsoft and other translation software. 91 languages.
Other sites can translate a document and preserve the original layout
https://www.onlinedoctranslator.com/en/ (this uses Google Translate)
If you want to write and translate a reply
- Write it in simple English with short sentences
- Use two different translation tools to translate the text into Spanish and than back again into English.
- Edit your English until the back translation makes sense
- Get someone with good Spanish to check the Spanish version
How translation tools work
From: https://translartisan.wordpress.com/tag/rule-based-machine-translation/
Rule based
This was the original method of machine translation, developed by the Russians in 1933 It relies on countless built-in linguistic rules and millions of bilingual dictionaries for each language pair. This is the basis of Prompt, Systran and Apertium (the latter is notable for dealing with different dialects).. the method is also the least accurate. https://www.apertium.org/index.eng.html
Statistical
At the turn of 1990, at the IBM Research Center a machine translation system was first shown which knew nothing about rules and linguistics as a whole. It analysed similar texts in two languages and tried to understand the patterns.
The machine needs millions and millions of sentences in two languages to collect the relevant statistics for each word. How do we get them? Well, let’s just take abstracts of the minutes of the European Parliament and the United Nations Security Council meetings —they are available in the languages of all member countries. Google Translate, Yandex, Bing, and other high-profile online translators worked as phrase-based statistical translators right until 2016. Google used to translate everything into English first, then into the second language – leading to many errors!
Neural machine
This is more difficult to understand. Using an analogy, I can describe the features of a dog to you, and you can “translate” these features into a drawing or painting of a dog.
Imagine the source text as a set of specific features. Basically, all it takes is to encode the text, then let the other neural network decode it back into text but in another language. The decoder only knows its own language. It has no idea of the origin of the features, but it can express them, e.g., in Spanish.
Neural network translation needs to be trained. It predicts the translation and sometimes gets it wrong. However, Google Translate and Microsoft Bing are starting to use this technique.
DeepL, the newest free translation tool was launched in 2017. It is run by a German company using something called “convoluted neural networks” and deep learning. The foundation of the tool is the company’s online multilingual dictionary Linguee, which compares the actual use of words and phrases. DeepL learns from the dictionary and phrase pairs. This means you won’t get a word-to-word translation with DeepL. This engine catches the nuances of language and provides a more natural translation. Translations are generated using a supercomputer that reaches 5.1 petaflops. It is operated in Iceland with hydropower. So it’s “green” to boot! https://en.wikipedia.org/wiki/DeepL_Translator
However, if you want to translate important documents then it’s still best to use a professional service. e.g. https://translationpal.com/services/medical-records-translation/spanish-to-english
Translation tools are neither equal nor infallible!
Here’s a comparison of different translation”engines” using a passage from El País , a Spanish national newspaper as the source text. Note the length of the complex, first sentence!
En su libro El futuro por decidir (editorial Debate), la que fuera la máxima responsable de la lucha contra el cambio climático de Naciones Unidas, la costarricense Christiana Figueres, considera que para alcanzar esas emisiones netas cero en 2050 hay que empezar por recortar a la mitad las emisiones de CO₂ de aquí a 2030 (pues la otra mitad será mucho más difícil de reducir), y propone que cada persona trace su propio plan personal para dividir por dos su huella de carbono a lo largo de esta década. Sin restar relevancia a las decisiones de cada ciudadano, Saldaña cree esencial que todas las personas se involucren, sobre todo, en los cambios colectivos. “A nivel individual podemos hacer pequeñas cosas, pero se consigue una disrupción más grande si trabajamos colectivamente. Por eso el ámbito local, como ayuntamientos o barrios, tiene un papel clave para realizar grandes transformaciones en un espacio corto de tiempo”, recalca. Y concluye: “Hay que buscar los cambios colectivos”.
In her book The Future to Decide (Editorial Debate), the one who was the United Nations most responsible for the fight against climate change, the Costa Rican Christiana Figueres, considers that to achieve these net zero emissions in 2050, it is necessary to start by cutting to CO₂ emissions will be half by 2030 (as the other half will be much more difficult to reduce), and he proposes that each person draw up their own personal plan to divide their carbon footprint by two throughout this decade. Without diminishing the relevance of the decisions of each citizen, Saldaña believes it is essential that all people get involved, above all, in collective changes. “Individually we can do small things, but a bigger disruption is achieved if we work collectively. That is why the local sphere, such as city councils or neighborhoods, plays a key role in carrying out major transformations in a short space of time ”, she emphasizes. And she concludes: “We must seek collective changes.”
DeepL
In her book El futuro por decidir (published by Debate), the former head of the United Nations’ fight against climate change, Christiana Figueres, from Costa Rica, believes that to achieve net zero emissions by 2050, we must start by cutting CO₂ emissions by half by 2030 (as the other half will be much more difficult to reduce), and proposes that each person should draw up their own personal plan to halve their carbon footprint over the course of this decade. Without downplaying the importance of individual citizens’ decisions, Saldaña believes it is essential for everyone to get involved, above all, in collective changes. “On an individual level we can do small things, but a bigger disruption can be achieved if we work collectively. That is why the local level, such as town councils or neighbourhoods, has a key role to play in bringing about major transformations in a short space of time”, he stresses. And he concludes: “We have to look for collective changes”.
Bing
In her book The Future to Be Decided (Debate publishing house), the former head of the fight against climate change at the United Nations, Costa Rican Christiana Figueres, believes that to reach these net zero emissions in 2050, we must start by cutting CO₂ emissions in half by 2030 (since the other half will be much more difficult to reduce), and proposes that each person draw up their own personal plan to divide their carbon footprint by two throughout this decade. Without detracting from the relevance of the decisions of each citizen, Saldaña believes it is essential that all people are involved, above all, in collective changes. “On an individual level we can do small things, but you get bigger disruption if we work collectively. That is why the local level, such as town halls or neighborhoods, has a key role to carry out major transformations in a short space of time, “he emphasizes. And he concludes: “We must seek collective changes.”
Apertium (rule based)
In his book The future for deciding (publishing Debate), the one who went the maximum manager of the fight against the climatic change of United Nations, the costarricense Christiana Figueres, considers that to reach these net broadcasts zero in 2050 it is necessary to begin by recortar to the half the broadcasts of CO₂ of here to 2030 (as the another half will be much more difficult to reduce), and proposes that each person trace his own personal plan to divide by two his footprint of carbon along this decade. Without subtracting importance to the decisions of each citizen, Saldaña believes essential that all the people involve , especially, in the collective changes. “To individual level can do small things, but achieves a bigger disruption if we work collectively. Therefore the local field, like city councils or neighbourhoods, has a key paper to make big transformations in a short space of time”, recalca. And it concludes: “It Is necessary to look for the collective changes”.