6 min read
Nikita Teslenko Grygoryev: 07/17/23
Transfer learning from large language modelsis behind a multitude of successful launches and applications in recent times. While transfer learningas...
1 min read
21/09/2013
A recent question on LinkedIn's Bay Area Machine Translation User Group asked what is best, a DIY MT solution or a SaaS model. It is a good question as the market has moved very fast from "MT-denial" to almost widespread adoption to several ways of adoption. These are some of the recommendations from Pangeanic about implementing a DIY solution in-house and being in full control of your engines, re-trainings, etc or on the other, hire a hosted SaaS MT solution. This is the quote which we use for illustration purposes. "I mentioned to Uwe I was surprised he did not mention PangeaMT in the article as DIY, when (as far as I know) we coined the concept and released the first DIY versions back in 2011. See http://www.businesswire.com/news/home/20110602006541/en/PangeaMT-launches-DIY-Solution-Machine-Translate-MT Approaches differ according to customization and user needs. Just like there is no "one good approach" to all language pairs in MT, few clients have similar needs. Customization is the key. Like Uwe, I represent a commercial company so I tend to talk about my experience and what I know best. The trend is towards SaaS nowadays (http://www.pangeanic.com/news/2012/saas-power-at-localization-world-paris-2012.html) because it takes a lot of the hardware hassle and infrastructure pain. However, both corporations and LSPs are still afraid of transferring data or providing data for customization. That is where DIY comes in, as you can customize as much as possible and leave the system ready so the user is empowered to run the project.
Installing DIY at a client's end means a certain level of training and knowledge acquisition on the part of the client. It is not just another tool, it is a new technology. Installing a solution and walking away from it would be like leaving my car running for my 11-year old to drive off (harsh comparision but as with anything in life, not understanding certain complexities about a system can lead to disaster). After a proper training course, running DIY in-house or using the SaaS model are pretty straight forward. In short, our recommendation is DIY (meaning "self-sufficient") to MT users who want independence or are very concerned with privacy, SaaS if you are not."
Installing DIY at a client's end means a certain level of training and knowledge acquisition on the part of the client. It is not just another tool, it is a new technology. Installing a solution and walking away from it would be like leaving my car running for my 11-year old to drive off (harsh comparision but as with anything in life, not understanding certain complexities about a system can lead to disaster). After a proper training course, running DIY in-house or using the SaaS model are pretty straight forward. In short, our recommendation is DIY (meaning "self-sufficient") to MT users who want independence or are very concerned with privacy, SaaS if you are not."
Related Articles
10 min read
Konstantinos Chatzitheodorou: 04/20/23
On March 22, 2023, we attended SlatorCon Remote, an online event where language industry experts and leaders came together to discuss this...
36 min read
Manuel Herranz: 11/07/23
LLMs or Large Language Models (LLM) are advanced deep learning algorithms capable of performing a wide range of tasks related to natural language...