How Fast Should You Post-edit Machine Translation? Here's What the EU Thinks - Slator - Translation

The Translation Centre for the Bodies of the European Union (a.k.a. the Translation Centre) is putting a number on it — how many pages translators and post-editors should produce per day.

(Post-editing machine translation, referred to variously as PEMT and MTPE, has become the industry standard and a skill in its own right.) 

In a new, EUR 2.8m call for tenders, the Translation Centre establishes multiple framework contracts for translation and post-editing services provided by freelancers “working on their own equipment, but connected to the Centre’s IT environment.”

Due to the confidentiality of the documents to be translated and post-edited, only the translators named in the contract will be permitted to handle the work.

According to the call for tenders, “as a rule, each project will always be processed through Trados Studio by the Centre.”

Contractors will receive a package containing a bilingual file with MT proposals and any relevant translation memory suggestions. Their assignments include verifying “each and every segment, regardless of its status, and mak[ing] the necessary changes.”

The RFP distinguishes between light post-editing, which does not entail stylistic corrections, and full post-editing, which requires attention to detail in terminology and register.

All final deliverables must be usable as received, “without any further revision or correction by the Translation Centre’s services.”

Tenderers can propose a maximum of two translators for each lot and must demonstrate their individual experience translating and post-editing in the language combination.

Just how much experience does the Centre want? At least 500 pages on topics related to European affairs or public administration matters. 

Translators Crunch the Numbers

The RFP states that a linguist working full-time should be able to translate 500 “standard” pages, or post-edit 800 pages, within six months. (The Centre estimates a standard page to include 1,500 characters, excluding spaces, in the source language.)

For the work handled through these framework agreements, the Centre “generally expects” one contractor to translate up to 10 pages daily; full post-edit up to 15 pages per day; or light post-edit up to 20 pages.

Slator Machine Translation Expert-in-the-Loop Report

60-page report on the interaction between human experts and AI in translation production, including AI-enabled workflows, adoption rates, postediting, pricing models.

Speaking with Slator, postediting trainer Enrico Antonio Mion pinned his own hourly output at 700 words for full post-edit and 1,000 words for light post-edit for a total of 5,600 or 8,000 words daily, respectively.

Bidders are required to submit a price offer, in Euros, per eight-hour workday for each lot — an interesting requirement considering the Centre’s estimates for services sought. With multiple units to consider, calculating rates can quickly become complex, and linguists turn to colleagues for advice. 

In a January 2023 forum on translation marketplace ProZ, Belgium-based French – Dutch translator Lieven Malaise wrote, “I apply a MTPE discount of 30-40% and that still allows for my hourly income of 35 euros. I don’t lose money by practicing MTPE.”

Jo Macdonald, an Italian – English translator in Spain, said he was uninterested in “PEMT jobs paid at your hourly rate calculating 10,000 words/hour.”

English – Swedish translator Thomas Johansson wrote that in his three years of working “with EU translations,” PEMT jobs have increasingly replaced translation. “Linguists are expected to accept a 30% discount on their rates for such jobs,” he explained, adding that “in effect, my hourly income often came out to about 40% of what it should have been.”

In an April 2021 thread about hourly rates for PEMT, Spanish – English translator William Bowley recommended working backwards: “If you’re working with agencies, they’ll often have their own expectations of x words per hour, upon which you can then base your quote.”

Hourly rates for translation and editing, including PEMT, are a departure from the traditional per-word rate, which Japanese – English financial translator Matt Stanton identifies in his advice to freelancers as crucial to success.

“Never forget that you’re paid by the word – not by the hour,” he wrote in a 2021 LinkedIn post. “This gives you immense power, as your hourly income is in your hands.”

But English – Dutch translator Gerard de Noord neatly summed up the paradox in a comment in the April 2021 ProZ discussion. 

“When you’re being paid per hour and your client tells you how many units you have to churn per hour, you’re not being paid per hour, you’re being paid per unit.”

Adblock test (Why?)

Use Google to Translate Text in Any Image - Lifehacker - Translation

Screenshot: Jake Peterson

Let’s say you have an image you need to translate. Maybe it’s a still from a show, and it uses subtitles in another language. Perhaps it’s a photo of a sign with instructions or directions that you can’t read. You might have a screenshot of an article on your phone that you’d like to learn about, but obviously cannot. Whatever the case, you can try using Google Translate’s new “Image” section to understand what the picture is saying.

Watch

• Off
• English

view video

Five Sites to Score Free Stuff (Other Than Facebook Marketplace)

Google Maps Is Android’s Answer to Apple’s 'Find My'

January 17, 2023

5 of the Most Useful Google Maps Settings

May 2, 2022

The new option is powered by the AR Translate Tool, the same tool that Google Lens uses. You’ve been able to use this feature on mobile for a while now, but its debut on desktop today means it’s easier than ever to translate just about anything you need to. Sorry, Duolingo.

But the best part of the feature isn’t the simple translation itself. Google Translate overwrites the original text, replacing it with the translation, as if it were the original language all along. Well, kinda. The translations show up in borders, as if someone took printed versions of the words and pasted them over the originals. In some cases, it’s quite seamless, while in others, it’s a bit like an arts and crafts project. Still, it makes it easy to read the translations, which is really the whole point.

To give it a shot, go to Google Translate in your web browser or through the mobile app. In your browser, choose the “Images” option that now appears, while on mobile, choose “Camera,” then tap the images icon on-screen. Upload the photo from your device (you may have to grant permission to your photo library depending on the device you’re using), and Google will get to work.

Once Google translates the text, you can choose to select portions of it to copy elsewhere, have Google Translate read it out loud (mobile), download a copy (desktop), or send back to the Translate app (mobile).

Screenshot: Jake Peterson

Adblock test (Why?)

Rotary's 16 years of dictionary gifting in schools coming to an end - Stuff - Dictionary

Supplied

Donovan Primary School Children lining up to receive their Illustrated dictionaries from Rotary Club of Invercargill North.

Pupils at Donovan Primary School were some of the last to benefit from a Rotary Club tradition this week when they received their Usborne Illustrated Dictionary as part of the Dictionaries-In-Schools project.

Rotary Club of Invercargill North programme convenor Anne McCracken said every Year 4 pupil at all Invercargill primary schools has received a dictionary since the project was first introduced in 2007.

More than 12,000 dictionaries had been handed out, which cost more than $100,000.

McCracken said each pupil writes their name and the date on their own copy, shakes the hand of the Rotarian who gave the dictionary to them, and then has to find a nominated word.

READ MORE:
* Aoraki pupils receive dictionaries through Rotary initiative

“The pleasure on their faces is a joy to watch,’’ she said.

Adblock test (Why?)

Google Can Now Translate Text in Photos from Your Browser - PetaPixel - Translation

Google Translate is already pretty good at automatically translating an entire web page’s text from one language into another, but a new update lets it do the same for text that lives inside of photos.

The capability of translating text in pictures was already available using Google’s augmented reality system within its Translate smartphone app — which uses a smartphone’s camera to “see” text on any surface and provide translations in real-time — but was missing from the company’s browser-based translation tools.

But as Engadget reports, the ability to translate text seen in photos has been added to Google’s Translate web interface. The tool is very straightforward and can translate the text — either using auto-language detection or it allows a user to specify the language — from any uploaded photo.

Below is an example photo with a lot of text in it, which Google’s new system can now translate:

After the translation takes place, which only takes a few seconds, users are given the option to download the translation or copy the text. The first option downloads a photo that has all of the areas translated in the original resolution that was uploaded, though the file is compressed.

The latter option provides the entire amount of translated text and copies to a computer’s clipboard that, when pasted, appears as a simple list. The issue with this particular method is that it isn’t particularly clear what is being translated. In the example photo above, the first thing to show in the list is “¥190” followed by “Edo village,” which seems to indicate the translation is just going from top to bottom, left to right. For most users, the view of the translated words overlaid on the original image are going to be the most useful.

At the time of publication, the photo translation feature wasn’t available in Google’s Translate Chrome extension, which still only shows “translate this page” as an option and hasn’t previously had the ability to translate the text in images.

Google appears to be taking more steps to improve its translation tools. Engadget reported in early February that the company was making changes to its system that were powered by artificial intelligence and would help users find alternate, and possibly more accurate, translations to words that might have multiple definitions.

---

Image credits: Photos licensed via Depositphotos.

Adblock test (Why?)

Gen Z dictionary shows larger generational divide than ever - Washington Examiner - Dictionary

Are you someone’s “sneaky link” or “OTP”? Have no idea what that means? Dating as you get older is hard. Trying to understand the younger generation on the dating app? Harder. Well, now Tinder has released a dating dictionary to help millennials understand what their Generation Z love interests are trying to say. It might show an even larger generational divide than ever.

Tinder research shows 62% of people ages 18-25 feel they speak a completely different language compared to older singles they meet online. While most people date in their generation, as millennials have gotten older and married, those who are still single may look to a younger generation to find love.

'ZOMBIE' DRUG OF HORSE TRANQUILIZER-LACED FENTANYL RAVAGES US AND ALARMS HEALTH OFFICIALS

While the dictionary includes some well-known millennial-coined terms, “cat fishing” and “vibe,” the Tinder dictionary outlines new phrases and even new types of relationships Gen Z has made up, one being a “situationship,” which is described as “a romantic juncture that hovers somewhere between a steady relationship and an affair or friendship and remains undefined throughout.” Many definitions of relationships were described as noncommittal, as the “hook up” culture seems to be growing and more widely accepted than ever before. People are now “cushioning,” or have many people on the side if their main relationship doesn't pan out.

Tinder has released a Gen Z dictionary to help other generations understand their younger love interest language.

Tinder Press Room

Gen Z seems to be a coin-conscious generation, as some unfamiliar terms are about saving money. “Affor-dating” and “frugal flirting” refers to “affordable dates” such as going for walks or coffee.

Some other unknowns were “kitten-fishing,” wherein the date makes themselves out to be better than they are online, not necessarily in looks but in job and lifestyle. Another is “Beige flags,” in which someone is incompatible because they seem boring or basic. The example given is “I got immediate beige flags when her bio said, live, laugh, love." 

Many of the terms focus on sexual preferences and gender ideology, such as transgender, queer, nonbinary, pansexual, polyamory, and ethical nonmonogamy. Notably, many terms have to do with wanting to be environmentally friendly, showing Gen Z values climate change as a high priority when dating. “Eco-dumping” refers to dumping someone who thinks climate change is fake. “Green dating” is when you specifically date someone who shares your values about the environment.

CLICK HERE TO READ MORE FROM THE WASHINGTON EXAMINER

This dictionary release shows your values can affect the language you use. This could help you understand Gen Z better — that is, if you want to.

Adblock test (Why?)

Microsoft Says Large Language Models Are SotA Evaluators of Translation Quality - Slator - Translation

On February 16, 2023, Microsoft announced that large language models (LLMs) can achieve high machine translation quality, mainly for high-resource languages. Building on this finding, Tom Kocmi, Senior Researcher at Microsoft, and Christian Federmann, Principal Research Manager at Microsoft, investigated the applicability of LLMs for automated assessment of translation quality. “If the model can translate, it may be able to differentiate good from bad translations,” they said.

In their research paper published on February 28, 2023, they proposed GEMBA, a GPT estimation metric-based assessment method. Kocmi and Federmann evaluated the ability of seven different GPT models, including ChatGPT, to assess translation quality using GEMBA.

According to the researchers, LLMs demonstrate “state-of-the-art capabilities” in translation quality assessment at the system level. However, they emphasized that only GPT 3.5 and larger models are capable of achieving state-of-the-art accuracy when compared to human judgments. Those findings provide “a first glimpse into the usefulness of pre-trained, generative large language models for quality assessment of translations,” they said.

“Unexpected” Performance

Kocmi and Federmann outlined the requirements for assessing translation quality using LLMs. These include: a prompt variant, a source language name, a target language name, a source segment, a candidate translation, and a reference translation — which is optional for quality estimation.

The researchers experimented with four different prompt types, modeling two scoring tasks and two classification tasks. This was done because “scoring of translation quality may be an unnatural task for an LLM.” The scoring tasks were based on direct assessment and on scalar quality metrics, while the classification tasks involved rating translation quality using a one-to-five stars system and labeling translation quality as one of five discrete quality classes. Moreover, they evaluated these four prompt variants in two modes: with a reference translation and without a reference translation (in a quality estimation setting).

Slator Game Localization Report

100-page report on the fast-growing game localization vertical. Includes market size, game development, language tech, GaaS, and in-depth localization process guide.

The researchers assessed the performance of GEMBA by comparing it to other top-performing automatic metrics such as COMET and BLEURT. They used data provided by the WMT22 Metrics shared task, which compares these metrics against human ratings for the English into German, English into Russian, and Chinese into English language pairs.

According to the researchers, GEMBA demonstrated “unexpected” levels of metric performance. More specifically, GEMBA outperformed all other reference-based metrics while also achieving the highest performance in the quality estimation mode. However, the results also showed that GEMBA is not yet reliable enough on the segment level and should only be applied for system-level evaluation.

Progress in Document-level Evaluation

Then, Komci and Federmann evaluated the performance of seven different GPT models using GEMBA, their proposed metric-based assessment method. The GPT models ranged from GPT 2 to the latest ChatGPT model. 

The researchers found that Davinci-002 and Davinci-003 (also known as GPT 3.5) and ChatGPT demonstrated great performance in the translation assessment task for all of the prompt variants, with Davinci-003 achieving the best performance.

ChatGPT performed slightly worse than the other two models, often providing an explanation of its scoring. The researchers suggest that this may be due to the prompt format not instructing ChatGPT not to generate an explanation, and different prompts could potentially improve the model’s performance.

The researchers made their code and prompt templates publicly available, along with all corresponding scoring results, “to allow for external validation and reproducibility.” They concluded that GPT-enhanced evaluation metrics could lead to progress in document-level evaluation due to their ability to use much larger context windows, which “could be beneficial as there is little research into document-level metrics.”

Adblock test (Why?)

Get Ready to Take Your Business Global with These $109.99 Translation Earbuds - Entrepreneur - Translation

Disclosure: Our goal is to feature products and services that we think you'll find interesting and useful. If you purchase them, Entrepreneur may get a small share of the revenue from the sale from our commerce partners.

If you want to grow your business even more, the natural next step may be to start expanding internationally. Up to 75% of the world doesn't speak English, but that doesn't have to stop you. With a pair of Mymanu CLIK S Translation Earbuds, you can translate over 37 languages in an instant.

These earbuds were featured at this year's Consumer Electronic Show, and they've won prestigious awards and distinctions like the CES and Red Dot awards. So if you want to carry a few dozen languages in your pocket, grab a pair while they're only $109.99 (reg. $220).

You don't need to learn another language to work with an international business partner. Just use a pair of earbuds instead. The Myman CLIK S could help you understand and communicate with individuals and groups of speakers in Arabic, Turkish, Thai, Chinese, German, Japanese, Spanish, French, and more.

Download the MyJuno app onto your Android or iOS device, and pick the language you're using and the language you need to translate. When you speak, you'll hold down a button in your earbud to provide an immediate audible translation on your phone. When your conversation partner needs to talk, they'll speak into your phone. A translation plays in your earbuds and displays on the app screen for you to read.

If you need to decompress after a long work day, you can also use your translator earbuds to listen to music or make calls. A single charge could get you up to 10 hours of use, and you can store up to two full recharges on the carrying case.

Get a pair of Mymanu CLIK S Translation Earbuds on sale for $109.99 (reg. $220) until March 5 at 11:59 p.m. PT.

Prices subject to change.

Adblock test (Why?)