Our start-up of the week this week is KantanMT, whose machine translation technology developed in conjunction with Dublin City University (DCU) is targeting the US$40bn-a-year language service and localisation industry.
KantanMT is a cloud based statistical machine translation platform that was developed to make secure and customised machine translation more accessible for smaller Language Service Providers and internal language departments. Members pay a monthly subscription fee for an account on the KantanMT platform.
Because the platform is cloud based, members need no special hardware or software to get started which greatly reduces the overall cost involved.
With this account they can develop, measure, refine, and deploy their own private machine translation engines, and using the RESTful API members can integrate machine translation into translation management systems, content management systems and web applications.
Other companies offering a similar service include Microsoft Translator Hub, and Let’s MT, which was development by the Latvian Research group, Tilde.
No longer lost in translation
According to founder Tony O’Dowd the main distinguishers between KantanMT and other providers are analytics, speed of deployment, customer support and ease of use.
“KantanMT was built with usability in mind. Members can be up and running in a matter of hours, rather than weeks, as is the case with some other machine translation solutions. Traditional machine translation systems and platforms are very technical, and require a considerable amount of skill and intelligence to apply.
“This means that an organisation would need to invest a large amount of money to either; hiring consultants, or PhD students. With KantanMT, anyone with basic IT skills can build, measure and manage their own suite of machine translation engines.”
KantanMT’s Analytics suite was developed in partnership with the CNGL Centre for Global Intelligent Content at DCU. It includes KantanWatch which allows members to track engine quality (based on industry standard metrics – BLEU, F-Measure and TER) over time, Build Analytics, which reports on the quality of engine training data and offers members suggestions to increase their quality scores, and KantanAnalytics, which O’Dowd says is the industry’s first non- reference, segment level quality analysis technology.
“These analytics technologies offer our members much more transparency than any other machine translation provider in the industry. KantanMT’s goal is to put the control back in the hands of the engineers and project managers, so they can develop fresh, competitive solutions for their clients and companies.”
The market
KantanMT is selling to Language Service Providers and large organisations that have internal language departments.
The market for translation is growing rapidly, primarily because of the massive increase in the amount of content being created by international companies and organisations.
The Common Sense Advisory Board has indicated that the localisation industry will reach approx. €29 billion (US$40bn) by 2014, and TechNavio predicts the Global machine translation market will increase 18.05pc Compound Annual Growth Rate (CAGR) between 2012 and 2016.
Types of companies who have engaged with KantanMT: LSPs (Language Service Providers), software companies, data mining companies, automotive companies, online comparison sites, gaming companies.
The founder
Tony O’Dowd is a well-known industry entrepreneur having founded Alchemy Software Development, developers of Alchemy CATALYST, a market leader in visual translation memory systems.
Having sold Alchemy to the world’s largest privately owned localisation service provider Translations.com, he set up KantanMT to make machine translation technologies easy to access and easy integrate into corporate communication channels.
Prior to Alchemy he was executive vice president of Corel Corporation and director of technology at Symantec Ireland Ltd.
The technology
“Machine translation engines are generated by statistically mapping one language onto another, usually within a specific domain, like legal, financial, automotive, conversational etc.,” O’Dowd explained.
“Members ‘build’ machine translation engines by uploading translation memory files (collections of bi lingual data sets) within their dashboard, or via the API.
“After building their engines, members can integrate MT into their localisation workflows and content management system using the KantanMT API. All accounts are completely secure, no engine data is shared,” O’Dowd said.
“Our vision is to make machine translation technology accessible and easy to integrate into the heart of localisation workflows and communication channels. This reduces the barriers to international trade and helps our clients expand their businesses globally.”
Progress
O’Dowd explained that since the launch of the platform in early 2013 over 32 billion words have been uploaded by members to the KantanMT.com platform. Over 7,000 machine translation engines have been deployed, and over 350 million words have been translated.
“Our service is now available worldwide 23x7x365 and has a 99.99pc uptime. We offered the platform largely as a free service in 2013, and are now focused on converting these into paying customers.”
Challenges
O’Dowd said that from a technical point of view, the main challenge was to move machine translation technology to the cloud, and to make it scale both horizontally and vertically at an economic price point.
“From a business point of view, we had to convince our members that you could develop, improve and deploy MT engines, using a cloud-based platform and that they would no longer need to invest in expensive hardware, software and technical.
From an investment point of view the company had a good head start compared to other companies as O’Dowd was well known in investment circles and the localisation business in general.
“The Irish start-up scene is vibrant and energised. We attended the recent Dublin Web Summit and were blown away with the talent and technology. Investment access is doable – if you’ve got a good idea and a team behind you. The investment community are eager to look at deals – but the deal size is still small scale (for the majority of deals) compared to Silicon Valley, and to some extent England.
“The Irish Government has some great schemes in place to get companies with existing products onto the cloud, and the team at Enterprise Ireland are great at helping indigenous companies embrace export-led sales and international markets.
“All very positive and needed to keep Ireland ahead and at the top in software innovation and technology,” pointed out
“At a Government level, I think more flexibility and innovation could be brought to focus on the indigenous software market. The construction Industry thrived on capital based investment schemes, perhaps something similar needs to be explored for the high-tech software industry in order to continue boosting its potential and its job creation capabilities.”
‘Be first to market’
O’Dowd concluded with some advice for fellow start-ups. “My one piece of advice is ‘Be first to market’. Being first into the market means you are market leader from the very start and you will be viewed as the thought leader by your peers.
“There’s no point being perfect and late to market – you’ve lost first leap advantage and will immediately be viewed as a follower and not an innovator.
“Gaining traction fast, especially in cloud-based businesses is vital. Trade off short-term revenue gain for footprint and then mine this quickly and thoroughly,” O’Dowd said.