translation agency

Artificial intelligence – the latest localisation superpower

22.3.2024 by Delingua

Large language models – what are they?

13.3.2024 by Sohvi Kähärä

Over the past year, it has been difficult not to come across content created by generative AI.You may have seen a LinkedIn post where OpenAI’s AI-based ChatGPT has been asked to give good tips on how to get a salary raise. Perhaps you have encountered a chatbot when you have needed help with booking an appointment online. You may have even laughed at images created by DALL·E, where the faces of familiar celebrities have been digitally added to classic works of art.

Generative AI is developing at a tremendous pace and language technology is also benefiting from these leaps of development. From the perspective of language technology, the most important role is played by large language models (LLMs).

Large language models need massive amounts of data

Generative AI solutions created for producing and processing language are called large language models. They are statistical models that calculate the probabilities of the occurrence of words or parts of words. The simplest language models calculate the probability simply by calculating how many times the word appears in the given text materials, without taking the surrounding words and their meanings into account.

Large language models, on the other hand, are advanced language models that are based on neural networks and, as their name suggests, are trained with large amounts of data. The training of the language models of previous neural machine translation engines often requires a dataset of hundreds of millions of variables but large language models require billions of them.

Current neural network-based language models, including large language models, are most often based on transformers. Transformers are neural network architectures that consist of many overlapping layers. Previous machine learning models focused on individual words in their training data, while transformers are capable of simultaneously processing and taking into account words within a sentence or a longer text and analysing, among other things, broader relationships within the sentence.

Large language models use huge text materials to learn which words appear in which context and which words often follow each other. This allows large language models to predict what words and phrases would appear in an answer suitable to a certain type of question, how a beginning of a poem could continue or how a certain matter would be expressed in another language in the context in question.

A one-size-fits-all solution is not suitable for all cases

Since the amount of data needed for training large language models is enormous, a lot of generic content will inevitably have to be used. Large language models can also be customised, or personalised, to make them more suitable for different tasks and needs. This can be done by adding additional training data in a more limited subject area and modifying the model’s weightings. It is also important to choose a suitable customisation strategy, such as building a ready-to-use model or learning through use with the aid of prompt engineering.

Especially when large language models are used for translation, it is useful to personalise language model with high-quality, fully matching bilingual materials. It is especially good if the texts used in personalisation and the new texts to be translated are from as similar subject areas as possible. For this purpose, training language models with existing translation memories is a good idea: all of the customer’s translations are saved in translation memories by language pair so these texts are always bilingual, are good matches and use the company’s desired style and terminology.

In this way, the language model learns what the customer’s texts are like and is able to imitate their style better in the future. This improves, among other things, how machine translation produced by large language models is aligned with the company’s style. This, in turn, speeds up translating as higher-quality machine translation that is customised according to the customer’s needs reduces the time involved in any post-editing required.

The downsides of AI

Although continuous learning and adaptability associated with large language models are in many respects good, they also have their downsides. Language model-based solutions in general use learn as they are running so they can also be intentionally misled and taught false information. If you repeatedly teach a language model that a pike is a bird instead of a fish, it will start believing this to be true. Large language models also suffer from so-called hallucinations, which you can read more about in our article on getting the most out of machine translation.

In addition to invented facts, another risk associated with language models trained with a large amount of data is that confidential information ends up in the wrong place. Large language models in general use are open to everyone and, when using them, it is not possible to be sure whether your data will be made available to others. Therefore, it is important to choose a reliable partner for whom data security issues are a priority.

When training large language models with translation memories, it is necessary to take care of data security and ensure that customer-specific data is used only in each customer’s own model. This way, you can avoid the situation in which the language model would introduce another customer’s confidential information into the translation. We can build a customer-specific language model directly as part of LanguageWire’s highly secure ecosystem. Read more about LanguageWire’s insights into the data security in large language models here.

Large language models as part of the future

Training large language models for translation needs is a giant leap forward as they can provide more accurate and natural translations into multiple languages. These models make it possible to produce translations quickly while taking into account the desired style and terminology. In the future, this development can break down language barriers and foster interaction between people around the world by strengthening understanding and cooperation on a global scale.

5 reasons why integration is a good idea

5.3.2024 by Delingua

LanguageWire’s APIs and connectors are a time- and effort-saving solution, helping you avoid copying text unnecessarily from one place to another and import any changes, no matter how big or small, to all language versions of your website in almost real time.

A connector is a kind of system integration that allows the customer to order a translation of the content through an existing publishing system. In practice, a customer who uses a publishing system to process their website content can use our connector to order website translations from within their content management system. Once the translations are ready, they are imported into the system as localised copies of the original page.

We have listed five key reasons why integration is a good idea.

1. Work in a familiar publishing system

LanguageWire connectors and custom API work with your CMS, PIM or DAM. You can order translations and track the progress of projects directly through the digital platform you are already using.

Integration can be built with nearly any generally used content management system. We have Adobe, Drupal, inRiver, Optimizely, Sitecore, TYPO3, Shopify, Umbraco and WordPress connectors – and many others! If you didn’t find an option suitable for your platform among our connectors, you can use our API to create a custom integration between your digital platform and LanguageWire.

2. Automate your translation management process

Copying source texts from websites and emailing them to be translated is a time-consuming process. A smarter and more efficient way to manage large volumes of website content and numerous language versions is to integrate the systems so that the process for managing the translations is as highly automated as possible. Managing online store product information in different language versions, for instance, is much easier when the processed are automated.

LanguageWire’s APIs and connectors make full automation of the process possible. Simply connect your application to a tech-enabled translation management tool, and your translations can be imported directly from the translator to the page.

3. Save time and reduce workload

By using integration, you can publish content in every language as simply as in one. Integration is a better option than time-consuming manual work, and you receive ready translations right into your own publishing system. Managing translation orders is quick and easy with a user-friendly interface. This allows you to save time and money and, above all, gain additional peace of mind when the most monotonous part in the language version release process can be eliminated. With aid of a connector, you can import any changes, no matter how big or small, to all language versions of your website or webstore in almost real time.

4. Minimise risks

Furthermore, the likelihood of making mistakes increases when a large number of files are sent back and forth, with internal editors’ modification proposals. With LanguageWire, you don’t need to make any compromises on security. HTTPS-encrypted transfer of data increases security and you can manage the integration between LanguageWire and your system with modern access control.

Our APIs are built to be very stable so you don’t lose any jobs or complete translations in the event of a network outage. With integration, you can send texts for translation securely without a flood of emails and a need to worry about human errors.

5. Higher translation quality

We work as an extension of your business by building a permanent team of project managers, linguists and engineers. By connecting your digital platform to LanguageWire’s system, you will have access to our network of over 7,000 language experts. Our experts translate more than 450 million words per year, in 500 different language pairs. No matter what languages you are looking for or what sector you work in, we’ll bring you in touch with a language expert who best meets your company’s needs.

Interested? Contact us – let’s make translation management easier!

Financial period ends – how can we help you?

21.2.2024 by Delingua

For many companies, the financial period is a calendar year, which means that it ended just at the turn of the year. At the end of the financial period, companies must remember to provide their auditor with, for example, receipts, possible loans, investments or changes in the activities of the company as well as to carry out an inventory.

The new financial period also brings certain things to remember. At the beginning of the year, an annual report on the previous year will be drawn up and a meeting of the Board of Directors and an annual general meeting will be arranged. It is a good idea for companies operating internationally to prepare the annual report in all of the languages used in the company to allow owners and investors to get acquainted with the figures and events of the previous year. If the company has foreign owners, it would also be advisable to arrange annual general meetings in such a way that everyone attending them understands what is being said. In such cases, professional interpreting is justified.

Delingua is happy to help when you need to have the annual report translated or interpreting arranged for the annual general meeting. From us, you can order a translation of your annual report and interpreting for the annual general meeting. We have compiled a few tips below to keep in mind when the annual report and the annual general meeting are topical.

Tips for having an annual report translated

When ordering a translation of the annual report, it is worthwhile to get in touch early enough. Annual report projects are ordered and scheduled well in advance, and professional translation agencies will be busy in the first months of a new year. Be sure to be on time, allocate enough time for the translation and involve the translation partner already in the planning phase of the annual report.

Reference material makes translation easier, and annual reports are no exception. You can provide your translation partner with previous annual reports or other material in the target language, for example. This gives the translator an indication of the company’s terms, specialised terminology and style of the translation.

Also, decide in advance on the file format of the annual report. If possible, translations should be made directly in the original file format. The translation partner should also be informed of any exceptions, for example regarding the figures.

The annual report is often a highly confidential file and should not necessarily be delivered for translation by email. The file can be delivered over a secure connection through the translation agency’s ordering portal, for example. If the company does not have a written agreement with the translation agency, it is advisable to sign a possible non-disclosure agreement well in advance of the start of the translation project.

You will find more practical tips in our annual report guide (in Finnish).

Tips for interpreting for the annual general meeting

If international participants are attending the annual general meeting, they should be provided with professional interpreting. Conference interpreting, which practically means simultaneous interpreting, is the most suitable method of interpreting for an annual general meeting. In simultaneous interpreting, the interpreter interprets at the same time as the speaker speaks. The speech to be interpreted is transmitted to the interpreters’ headphones with the aid of interpreting equipment and interpreting is then transmitted by microphone to the listeners’ headphones. Interpreting can be arranged on site, as a hybrid model or entirely remotely. Interpreting can be arranged whether all of the participants or only some of them are physically present at the meeting or whether the meeting is held remotely.

When planning interpreting arrangements, the following factors, among others, must be taken into account: which languages the meeting will be interpreted into, what the main language of the meeting will be, and where and when the meeting will take place. It is also a good idea to provide the interpreters with advance material that will facilitate successful interpreting. Advance material may include the schedule as well as the documents and presentations of the meeting. Any other material in the target language also helps the interpreter to maintain terminological consistency.

Interpreting for foreign-language participants makes them feel that they have genuinely been taken into account in the arrangements of the meeting. We have prepared guides on interpreting for an event (in Finnish) and on remote interpreting. The guides will provide you with comprehensive information on what to take into account when arranging on-site and remote interpreting.

Delingua helps you at the end of the financial period

Now is a good time to order a translation of the annual report and arrange (remote) interpreting for the annual general meeting; we are happy to help. Contact us via email in matters concerning translation and interpreting; we promise to get back to you as soon as possible!

What is generative AI?

29.1.2024 by Delingua

We’re on the cusp of a technological revolution. One that promises to change how we research, work, and even live. 2023 was the year of Generative AI.

Generative AI isn’t just a buzzword; it’s a powerful force that has the potential to transform the way we communicate with each other and create content. It’s like having a creative genius, a tireless assistant, and a subject matter expert at your fingertips, ready to help your business thrive in an era where innovation is the name of the game.

So, whether you’re an industry leader looking to revolutionize your business or simply someone curious about the many possibilities of technology, get ready to discover just what exactly Generative AI has to offer.

Let’s dive in.

What is Generative AI?

Generative AI, short for Generative Artificial Intelligence, is a cutting-edge technology that empowers machines to autonomously create content across various domains. Unlike traditional computer programs that follow predefined rules, Generative AI models are designed to simulate human-like creativity, decision-making, and problem-solving processes.

At the heart of Generative AI lies the concept of machine learning, particularly deep learning, a subset of artificial intelligence. These AI models are capable of not only processing and analysing vast amounts of data but also generating new content based on the patterns and knowledge they’ve gleaned from the data they are trained on.

Versatility of Generative AI

Generative AI is incredibly versatile and can be applied in a variety of different ways to aid content creation and research. For example:

Content Generation: Generative AI can create written content, including articles, blog posts, product descriptions, and marketing materials. It can generate text in different styles, tones, or languages, making it a valuable tool for content marketers and copywriters.

Art and Design: In the creative world, Generative AI can produce art, design concepts, and even music compositions. Artists and designers can use AI-generated content as a source of inspiration or to augment their existing creative processes.

Simulation and Prediction: In industries such as finance and logistics, Generative AI can simulate scenarios, predict outcomes, and provide recommendations. This assists businesses in making data-driven decisions and allows data analysts to optimise their workflows.

Personalisation: Generative AI analyses user data to tailor content and recommendations. For instance, e-commerce platforms use it to suggest products based on a customer’s browsing and purchase history, enhancing the shopping experience for a user.

Language Translation: Language models like GPT-3 can translate text from one language to another with a high level of accuracy in many language pairs and can be further personalised with business-relevant language using termbases and translation memories as training data.

How Does Generative AI Work?

At its core, Generative AI relies on the power of large datasets and complex algorithms. These AI models are trained on massive amounts of data, learning to mimic human creativity and decision-making processes. For instance, OpenAI’s ChatGPT, currently the most well-known Generative AI model, has been trained on a diverse range of internet text, making it capable of producing high-quality content across various domains.

Generative AI operates on a blend of advanced algorithms, vast datasets, and deep learning techniques. Let’s delve a little deeper into the technicalities of Generative AI to help you understand its fundamental mechanisms.

1. Data and Training

As we said above, the core of Generative AI is the process of training. Imagine it as teaching a machine to understand and generate content by exposing it to a massive amount of data. For instance, in the case of language generation, an AI model is trained on vast collections of text, encompassing everything from books and articles to websites and social media posts.

2. Neural Networks

The central component of Generative AI is a neural network, which mimics the structure and function of the human brain. Neural networks are composed of interconnected nodes, or artificial neurons, organised in layers. There are typically three types of layers in a neural network:

Input Layer: This is where data is initially fed into the network.
Hidden Layers: These intermediate layers process and transform the data.
Output Layer: The final layer produces the desired output, which, in the case of Generative AI, is the generated content.

3. Learning and Weights

During training, the AI model learns the relationships and patterns within the data by adjusting the strengths of connections between artificial neurons. These connection strengths are referred to as “weights.”. The learning process involves iteratively adjusting these weights to minimise errors and make the model better at predicting or generating content.

4. Generating Content

Once the AI model is sufficiently trained, it can generate content by taking an initial input, often referred to as an “input” or “prompt,” and processing it through its neural network. The model then leverages the patterns and associations it has learned from the training data to produce coherent and contextually relevant output.

5. Fine-Tuning

To make the generated content even more accurate and tailored to specific tasks, businesses often engage in a process called “fine-tuning.” This involves training the model on a narrower dataset that aligns more closely with the desired application. For example, a company might fine-tune a Generative AI model on its product descriptions to ensure they match its brand style and tone.

How Businesses Can Leverage Generative AI

Generative AI’s ability to create content autonomously is a game-changer for businesses. It can automate content generation, reduce operational costs, enhance personalisation, and accelerate innovation.

The business advantages of Generative AI are multifaceted. Firstly, it offers cost-effective content generation. Imagine automating the creation of product descriptions, social media posts, or even legal documents. This not only saves time but also reduces operational costs.

Moreover, Generative AI enhances personalization and customer engagement. By analysing user data, it can generate tailored recommendations, advertisements, and even chatbot responses, leading to higher customer satisfaction and loyalty.

Additionally, Generative AI accelerates product development. It can assist in generating design concepts, and prototypes, and even simulate product performance under different conditions, helping businesses bring new products to market more quickly.

The benefits and use cases for specific industries are still being explored but the potential of this new technology seems limitless. For example, the healthcare industry has already started utilising Generative AI for drug discovery, predicting disease outcomes, and even generating medical reports.

Concerns and Challenges Surrounding Generative AI

While Generative AI offers tremendous potential for businesses, it’s essential to address certain concerns and challenges to ensure responsible and ethical usage rather than blindly embracing new technology without proper scrutiny. There are some issues you need to consider before implementing Generative AI into your processes.

Ethical Considerations

Generative AI can sometimes produce content that is misleading, offensive, or biased. This raises ethical concerns, especially when businesses use AI-generated content in customer-facing communications or decision-making processes.

Privacy and Data Security

Generative AI relies on extensive datasets, and using this data responsibly is critical. Mishandling data can lead to privacy breaches and regulatory violations, potentially damaging your business’s reputation.

Regulatory Compliance

AI, especially in highly regulated industries like finance and healthcare, must comply with a myriad of regulations. Ensuring that AI-generated content meets these standards can be challenging.

Training Data Quality

The quality and representativeness of training data can influence the effectiveness of Generative AI. Biased or incomplete data can lead to biased or inaccurate AI outputs.

Generative AI holds immense promise for businesses, but it also presents challenges that must be addressed responsibly. By implementing ethical guidelines, prioritising data security, defining accountability, and remaining vigilant, businesses can harness the power of Generative AI while minimising risks. The key is to strike a balance between innovation and responsibility, ensuring that AI serves as an asset to your organisation without compromising ethics and integrity.

The Future of Generative AI and Language Service Providers

At LanguageWire we have already talked in depth about how we foresee Large Language Models being the future of not only content generation but also machine translation.

Embracing this new technology will be the next step in innovation in this field. The field of Generative AI is constantly evolving, and we anticipate even more remarkable developments in the future.

This article was originally published by LanguageWire (Verity Hartley)

Captions vs Subtitles: What is the difference?

24.1.2024 by Delingua

The terms ‘caption’ and ‘subtitle’ are often used rather interchangeably. Each is a transcription of an audio or video recording that displays the spoken language as words on a screen.

This serves as an excellent tool to make recorded media more accessible to viewers in various circumstances. For example, it is well known that many people view online videos without sound, meaning they read the text on their device because they don’t want to disturb others nearby, or have physical hearing impairments.

In fact, “more than 100 empirical studies document that captioning a video improves comprehension of, attention to, and memory for the video. Captions are particularly beneficial for persons watching videos in their non-native language, for children and adults learning to read, and for persons who are D/deaf or hard of hearing.” (Gernsbacher, 2015)

By creating the transcription of the words spoken (and/or displayed on screen as graphics or titles) in their original language, this transcription can then be used to generate a localisation (translation) to appeal to many more audiences from different international markets.

The transcription could be generated as a single block of text, typical of written prose. But for the purpose of joining the text with certain images on a screen, it is necessary to generate coordinated time code. This is where a transcription becomes much more sophisticated, requiring professional software and services.

When we look more closely at the different potential consumers of video and audio content, there are subtle and important differences between captions and subtitles.

Similarities between captions and subtitles

Each is a text file accompanied by time code for synchronization to media. The most common format is SRT.

At a minimum, the text includes a transcription of what is being spoken. Text may also include on-screen text for subsequent translation (localization).

Differences between captions and subtitles

Subtitles generally only include a transcription of what is being spoken in the video or audio recording. These are often translated into various languages.

Captions can include other elements that describe sounds and music (for the hearing impaired) and can be toggled on and off (closed captions). Of course, captions can also be translated.

There are very many more detailed differences and options when configuring captions and subtitles, and many sources to learn about these differences.

Important considerations when ordering subtitling or captioning services

Is there existing on-screen text that needs to be included in the transcription for the purpose of translation?

Will the subtitles/captions cover up existing text (such as names and titles of speakers)? In this case, there may need to be another step involving specific placement of a caption on the screen.

Should the transcription include indications of other sounds (such as music) for hearing-impaired viewers?

Subtitling and captioning services are an excellent way to add to the value of your media content. It makes vastly more accessible to a much wider audience. It may also improve to increase the comprehension of viewers, even if they might otherwise understand the original language, as many people are more visual that aural learners.

This article was originally published by LanguageWire Company (John Jones)