Month: May 2017

What could be the perfect user interface for interactions between humans and machines? The IT industry has been pondering this question ever since the advent of advanced computer systems. When it comes to digital language learning, however, the answer is obvious – The ability to input language verbally offers unique possibilities for learners, something that keyboard-driven written input simply cannot match. Unfortunately, early attempts at tapping the potential of this input method tended to fail because of the rather underdeveloped technology that was available at the time.

How, then, does Avallain use Avallain Author and Avallain Unity voice technology to individualise and improve learning experiences and results as decisively as we do today?

How voice technology supports the learning process

Traditional user interfaces based on visual feedback, such as keyboards and touch screens, have one noteworthy disadvantage when used for learning a language – Users can only enter their words into the e-learning software via a keyboard. Consequently, they are able to train their reading, listening and writing skills in their language of choice, but they cannot effectively train their speaking skills. This is where integrating voice technology into the language learning process opens up a whole new set of possibilities:

• Users can actively train their speaking skills.
• Learners can receive immediate individualised feedback.
• Errors in pronunciation can be detected before they fossilise.
• Language teaching approaches no longer need to be based solely on written language.

Avallain has recognised the advantages that voice technology can offer early on. We have been following the developments in this field for many years, keeping an eye out for voice technology advancements which can be integrated into our products to improve them in a meaningful way. However, we do not blindly follow any new approach, as underdeveloped technology could impact both the user experience as well as the ultimate goal of learning a new language very negatively.

The three milestones on the road to a mature voice technology

Only 15 years ago, no one would have believed that voice technology – which, back then, was rather limited and frequently unreliable – would mature into advanced voice recognition software such as Siri and Alexa. Today, people all over the world can control these virtual assistants by speech, using countless different languages and dialects. And when these assistants are asked questions, their ability to give fitting answers has become remarkably reliable.

Today, voice technology is advanced enough to meet our standards of quality, which means it can offer significant benefits to users of Avallain Author and Avallain Unity. Our products allow educators to create and publish interlocking e-learning solutions which would have been considered science fiction, even in the nineties. In addition, research suggests that real breakthroughs in some of the most complex areas of voice technology have been made over the last couple of years as well.The path to the current state of the art has been very demanding, requiring us to overcome three key milestones.

First milestone: Audio recording and playback

From today’s point of view, recording and playing back speech over the internet could be considered the easiest of the three milestones to be overcome. This technology constitutes the basis of all further developments in the field. And since it already managed to meet our standards of quality in 2002, we could successfully integrate it into our Macmillan English Campus project even back then.

Today, our software supports a variety of recording and playback methods and applications. For example, Avallain Author allows user input and pre-recorded files to be merged into complete dialogue sequences, both online and offline. Thanks to Avallain Unity, the final recordings can then be sent directly to teachers for feedback, or simply as a means of verbally communicating outside of the fixed course locations.

Second milestone: Synthesised language

Compared to recording and playing back speech, creating synthetic voices audio is a much more complicated topic. For this reason, we waited until the technology had sufficiently matured in 2004 before integrating speech synthesis as a feature into our e-learning software. Back then, we entered into several partnerships with leading speech synthesis companies such as Acapela Group. These cooperations allowed us to offer exciting new features to our customers, all based on language synthesis technology. Some of these features include:

• The ability to translate pre-written text into synthetic audio (text-to-speech).
• Adjusting specific features of recorded speech – e.g. changing dialects to fit specific markets.
• Controlled and scalable production of audio recordings.
• Learners can directly translate their own written input into audio using text-to-speech (TTS).

Our current voice technology is flexible enough to create and play back authentic speech using various languages and dialects. Even UNESCO has noted how useful this feature has been for literacy programmes all over the world, describing our successful efforts to provide Swahili education to the coastal population of Kenya on page 16 of their report on international education initiatives.

Third milestone: Speech recognition

Until recently, speech recognition and assessment has always required the attention of an educator, particularly because of the great variety of possible pronunciations. And even the specialised software solutions of today, which can be valuable to automated pronunciation analysis, still only apply to very specific cases of language analysis.

Because of these limitations, our focus has been on general voice recognition features, meaning the ability of the software to recognise and accurately transcribe spoken language. In the early days, even technology developed by giants such as Microsoft, Apple and Google could fail, regardless of the quality of speech. The first companies to offer significant advances in this area were niche software providers such as Nuance who would often go on to be market leaders in areas such as recording dictations and automating customer services. However, their solutions usually required users to be trained in using their specific software.

For us, this approach to voice recognition was not an ideal solution for the education sector, as the need to teach learners how to use a learning software only creates additional obstacles on the way to education success. For this reason, we initially concentrated on using less technologically ambitious, more intuitive e-learning solutions. For example, when working on the learning platform iwdl.de, we deliberately limited voice input options in gamified exercises to individual words instead of entire sentences. Using this approach, iwdl.de has managed to become the first ever digital learning tool to be approved by the German Federal Office for Migration and Refugees for use in immigrant integration courses.

2017 – Visions of the future become a reality

Now that we have successfully overcome these milestones, voice technology opens up a whole new world of exciting possibilities to learners and educators alike. In the summer of 2017, for the first time, we will introduce an Activity Type which allows the software to give direct feedback to learners regarding the quality of their pronunciation of texts within an Activity. After that, the next important step is to introduce the ability to use spoken language within any given Input Activity. This could be used for exercises in which learners have to enter elements verbally or even for gamified exercises in which learners can simply utter the answers to specific questions – these are the near-future milestones for Avallain.

To achieve these goals in 2017, we are currently working primarily with Google’s Speech API, however we will be using three of the key software solutions in this area in the near future.

What’s next for Avallain? That will be decided through constant communication between us and our customers. But one thing is certain – as always, our customers will be the first to be able to offer the latest technologies to their end-users in a comprehensive and fully reliable form. Together, we will make individual education more comprehensive, more efficient and more exciting.

Even in our tech-savvy modern world, the term “big data” can divide opinions. To some, the term is synonymous with surveillance, to others it represents the unique opportunity to analyse and gain a thorough understanding of complex situations by means of technology. Within the educational sector in particular, big data can significantly improve the effectiveness of differentiated and individualised learning processes.

But how is it possible to observe the various ethical and legal restrictions of the international marketplace at the same time?

Information overload in the digital age

Detached from its moral and political connotations, the term “big data” simply describes sets of data which can only be processed by technological means. In most cases, the information is in a state of constant flux or it is too unstructured to be successfully processed by the human brain. Often, the datasets are also simply too large.

Only a few decades ago, this would have been an insurmountable problem, leading to the loss of valuable information. But today, even particularly large and complex datasets can be collected and processed thanks to advanced digital systems. Now more than ever, software is being used to perform tasks that are simply too difficult for humans to perform at the same level. Notably, such tasks include:

• Collecting vast datasets
• Analysing and correlating individual pieces of information
• Interpreting all available information

Processing information this way saves both time and resources – while yielding surprisingly precise results. One need only look at the institutions that already rely on such systems to realise just how useful big data processing can be.

The economic and social role of big data

Advertising companies were among the first to use computers to collect, analyse and process big data. This entails automatically correlating highly complex subjects such as personal thematic preferences as well as patterns of media consumption and purchasing behaviour. The resulting insights are then used to create advertising strategies designed to psychologically affect a target audience by catering to its specific communicative needs.

Today, even taxi companies rely on collecting and processing big data to optimise their workflow. And since the US election in late 2016, it has become obvious that big data is not only processed by big business, but by political and social entities as well. Big data has become an integral part of modern society, and it is a subject of much discussion in the educational sector as well.

Can the EdTech Industry take advantage of abundant information while avoiding the risks involved?

There are two major discussion points: complexity and the legality. Big data processing solutions are highly complex by nature, which is why there are ethical as well as legal considerations to be made. For example, many countries do not allow for large individual sets of learner data to be merged with each other for analysis, on grounds of privacy protection.

Of course, there are good reasons for extensive privacy protection legislature. For example, the ethical concerns over the complete surveillance of classrooms practised by some US start-ups are completely reasonable. However, prohibiting big data processing techniques such as merging learner data across municipal borders can cause the artificial intelligence behind the system to form conclusions based on insufficient information. Consequently, any resulting analysis of such a limited dataset would be distorted, negatively impacting the individual learning experience.

The Avallain solution – One comprehensive system, fully customisable

We at Avallain have been tackling the topic of “big data” for a very long time. How can the available big data be collected to improve individual learning experiences – while simultaneously observing ethical and legal restrictions, which may differ from country to country?

Does big data provide big advantages to e-learning?

One of the unique benefits of education technology is its ability to adjust to the needs and preferences of individual learners in order to support individualised and differentiated learning approaches even within heterogeneous groups. However, when developing our own system, we found that most contemporary tools are solely focused on collecting trivial information, such as counting click-through-rates via Google Analytics, collecting administrative statistics as well as individual test results and tracking the frequency in which learners use particular pieces of content. Using such tools, one may be able to calculate average scores, but it is impossible to easily determine which educational approach has most significantly supported a learner in their studies. Finding such information usually requires a lot of additional work on the part of educators.

Even simple tasks like merging large datasets across platforms and devices using xAPI has not yet become an established standard for these tools. Such a limited and unstructured approach to big data does not yield useful information; it only contributes to “data lakes“. This means that the software collects input and simply stores it without analysing or collating it. Because of limited resources, such information tends to only accumulate over time without ever being processed.

The Avallain system – Think big

Our software solution, first released in 2016, is the result of our observations. It was designed to process big data efficiently without crossing ethical boundaries. Our system is based on the very same philosophy that Avallain has been following for more than 10 years now. Combining the most advanced technologies with the maximum amount of user friendliness thanks to complete flexibility. Our system cannot only be adapted to the personal needs of individual users, but also to the ethical and legal restraints of any given country.

Our system combines an xAPI learning record store with additional event storage capabilities. Not only does this allow learner progress to be recorded, but it also allows any given event to be recorded, ranging from menu usage to voice recordings in interactive audio tasks. Such events are stored in a separate big data warehouse system and can be processed using any common business intelligence suite. We also focus on usability when retrieving information – the process is particularly quick and does not require any additional training. This allows for effective big data mining from the very first day of using our software, giving immediate answers to questions such as:

• Which content is especially popular?
• Which content is popular with which age group?
• Which content may be too hard or too easy?
• Which content yields the best learning results?
• Which channels of communication are being used within the learning platform?

Both the event storage system as well as the business intelligence suite can be completely customised according to the customer’s requirements. Our big data analysis system is compatible with database systems ranging from open source software to professional cloud storage solutions. In addition, it can operate with any of the most common analysis tools such as Amazon Elastic Map Reduce. Thanks to the cross-industry approach of our architecture and the established standards which it is based on, we can quickly adjust to future developments and new methods of big data analysis as well. These can be integrated into the software at any time, while remaining cost-efficient.

Responsible use of big data supports individual paths to learning success

Avallain provides systems for efficiently using big data in digital education – which is done without crossing any ethical or legal boundaries. Our intelligent systems offer educators, institutions and corporations the ability to collect, analyse and interpret information in a very cost-effective manner.

We have recognised that by using big data consistently and responsibly, we can revolutionise individualised and differentiated learning approaches. To learners, this paints a particularly bright future, as their education will be more individualised, more interesting and more effective thanks to big data.