An Interview with Flora from SUMM AI on automatic Translation in Easy Reading

This is the transcript for the German podcast. All typos and inaccuracies are mine.

Domingos: Welcome to a new podcast on digital accessibility. Today I have another exciting guest with me, namely Flora Geske from the company SUMM AI. The company SUMM offers a tool with which you can automatically translate texts into easy language. But Flora can explain that better. First of all, thank you for taking the time. Just introduce yourself.

Article Content

Information about Flora and its relation to easy language

Flora: You're welcome, thank you very much for the nice intro and also thank you for allowing me to be here today. My name is Flora Geske. I am CEO and founder of SUMM AI, so ultimately managing director and responsible for strategic issues, i.e. where is the journey going with plain language, where is the journey going with our tool.

I strive to create partnerships that are in line with our mission, namely to make the world understandable for all people and also with public relations to the outside world. That means, for example, coming to very nice podcast invitations and explaining what we do.

My background is that I am a business informatics specialist, which means that I studied business informatics at the Technical University of Munich and that is exactly where I met my two co-founders during my studies. It was a while ago now that I was studying for a master's degree, but that's when we basically made the plan , that it should be a startup and also a startup in the area of digital accessibility.

Domingos: Yes, super exciting, how did you even get to the topic of easy language? When you hear business information systems, I don't think there's that much about easy language.

Flora: It's actually a shame that there hasn't been that much going on at this interface between inclusion and tech, I think it's on the rise a bit now. This may also be disappointing if you work in this area yourself. But I already have the feeling that it's a trend And we ultimately found the interface so that my co-founder Vanessa and I did an AI project during our master's degree with the University of Toronto. We dealt with AI and natural language there and somehow figured out, a little before many people had ChatGPT in their LinkedIn newsfeed, that natural language and AI go together quite well and that the technology has a lot of potential and we want to use it Find cases, find use cases in which we not only use AI, but also AI for socially relevant purposes, i.e. do something useful for our society with AI, because there are many commercial applications and they are also exciting, but that was a bit like that our wish and I didn't actually have to look long to find such an application, because my aunt is actually the easy language target group, i.e. she has a physical disability and was previously unable to go to school for so long, unfortunately because of this disability.

And yes, that's how it happened that I've known the topic of easy language for a long time, because she's interested in politics, because she surfs the Internet and always comes across these language barriers and doesn't understand things that she actually wants to be involved in. And yes, I knew then that easy language wasn't available in many places yet. On the other hand, I knew what great things AI could do, and then we found out that with our AI tool, we can produce easy language, make this translation automatically, and that we can then make a real difference, and so on the idea came about.

Domingos: Very interesting. Most people listening will probably know what plain language actually is, but maybe you can explain it in a few words and also what target group it has.

Flora: Absolutely, easy language is actually a fixed concept, a fixed term, so the L for easy is also capitalized in this case because it basically references a concept that already exists and it is a kind of language style, in the area like that in terms of competence level A1 to A2, which is really very simplified. These are sentences, there are no commas, so use sentences without subordinate clauses, with a very simple choice of words, if possible no foreign words and really in a way that you can follow very easily and very clearly. There are also no abstract formulations and no idioms, so that maximum comprehensibility can be achieved. I like to say linguistically the lowest common denominator, so that people with learning difficulties, with educational disadvantages but also people who learn German as a foreign language, for example, can find access to our language and understand a text very well.

We do a lot in the public sector, for example today I have a text on the table on the subject of property tax reform. Terms are simply used and very long sentences are formed, so almost everyone gets out of it a bit. This is a case where a very complicated language is made very simple and truly accessible with a translation into easy language.

The company SUMM and its translation tool

Domingos: Yes, thank you very much for the introduction. Then I would like to talk about your company. How long have it been around?

Flora: We actually just had our anniversary in April, when we officially started as a company, so we've been working on the topic for quite a while, because of course technical developments like this take time, you have to deal with the topic first learn easy language. First of all, we networked in the easy language community with a lot of relevant people who already knew a lot about the topic, because we were basically new to it. So from the day we were a real company and worked, now it's our anniversary. Time flies, a lot is happening and it's great fun to move forward and also see: people notice us, perceive our mission and also support us.

Domingos: How many are you now? I've now noticed that you were three founders, but I think there are a few more of you now.

Flora: exactly, our team is 14 people. In fact, not all of them are full-time employees, so we also have great working students and interns who support us. But over that time we have already grown to a considerable number of employees because there is a lot to do, because we are starting a lot of great projects, because our solution is being integrated into websites in many places, for example at the city of Hamburg on the Web portal hamburg.de. So there is a lot to do and in this respect we are growing.

We pay a lot of attention to diversity, that is very important to us. We have, I would say, two-thirds of the founding team are female and we try to keep this in mind within the team, we have employees with severe disabilities, so we really try to keep in mind as we grow, that we want to be diverse and that we have as many perspectives on the matter as possible who we want to take with us to work.

Domingos: yes, very cool. You already said that the tool automatically translates texts into simple language. Hardly any of us will probably understand the technical background, but perhaps you can briefly explain how the process works. So I enter my text into a web form and it is then automatically translated or how does that work?

Flora: There are different ways to use the tool. We started with a very simple web application, i.e. just a page that you can call up in the browser and that looks very similar to something like Google Translate or deepL, as we like to say. So there is an input field where you can copy complicated text, then click on translate and then you will get a translation suggestion in Easy Reading. Then we always recommend taking another look at it, because just as we humans feel when we read something, our AI may interpret something a little differently and you might then want to make a small change. We have also created one or two tools for revision and that is actually the basic application, which is not particularly complicated in principle.

We then said that people who work with text might want to do it in Microsoft Word. That means we have integration. We have a small page panel in Word to generate the translation. And we also have a programmatic interface, now it's getting technical again, i.e. easy, that you can also integrate our translation option into the IT system and then, for example, editors on hamburg.de can create accessible text with one click, without that they have to copy something in and out again. You simply notice when we work with large institutions or companies that the convenience of the process is also a priority.

Domingos: Absolutely. Who are you actually targeting with your tool? Is your target group the people who rely on easy language, is it the translators or the large institutions like the city of Hamburg?

Flora: So our approach is that we say, well, those who are responsible for building the barrier, we would like to include them when we dismantle the barrier. This means that we are aimed at public institutions, the public sector and all those who have a barrier and want to dismantle it. This means that a city of Hamburg, for example, or other cities and municipalities or authorities use our tool to simplify texts and convert them into easy-to-use language. Then the users take another look to see whether the content fits, and then the texts can be published on the respective websites or in brochures, depending on where the text is supposed to go. And the good thing about it is that we also say very clearly that someone who knows the content of the text will take a look again and make sure that there is no false information in it, that it all really fits. That means, on the one hand, we have the advantage that a review is taking place and, on the other hand, that those who are trying to break down barriers on their website and their flyers or brochures where the barriers exist There are also translators who use our tool because we also say: Ultimately, we are not interested in replacing manual work, but rather a maximum of a third of institutions and companies or actually only public institutions offer easy language at all, although there is even a legal obligation because manual translation of texts into easy language takes so long and is so expensive. That means we are a supplementary tool to manual translations. And that's why translators can also use our tool when they do translations and get there faster when there are a lot of orders.

Companies will also now have a legal obligation from 2025, the (German) Accessibility Strengthening Act, which will also ensure that products and services are offered accessible in plain language. That means we expect some demand there too.

Domingos: how would you rate the quality of the translation. So do you think that your texts can be published directly, so to speak, or should someone with translation skills or easy language target groups take a look at it?

Flora: yes, that's a very good question and of course we try to be objective in order to assess for ourselves how good our tool is. We have a small feedback function. This means that whenever someone generates the text with our tool, you can also report back: That was something or that was nothing and we are at 80 85% very good usability, so these texts can then be published more or less immediately. You have to look again at the other portion.

What we do in advance, when we give out our tool, is a training course, an introduction to the topic of easy language and also with a small checklist so that you just know: okay, if I want to publish a text like that now What do I have to pay attention to so that this is really a good and understandable text? This means that we make sure that the users also know the weaknesses or the things that need to be taken into account manually, that they know them and that really good texts come out. And then we're somehow at 5 minutes in post-editing, so that we can publish the text very quickly and we're essentially breaking down not only the barrier to the text, but also the barrier in the process of publishing the text.

And when it comes to the topic of target groups and also people who rely on easy language, we do this in this step, especially when developing the tool. Our texts, with which our tool is improved, with which our tool is trained, we include the so-called test groups, i.e. people who need easy language and who can say, I understand this or I don't understand that and also give texts regularly there to simply understand: What can our tool do well? When it comes to the fact that it produces really comprehensible text and where do we perhaps still need to do some homework and improve it further.

Domingos: A question that comes to mind: plain language texts are often accompanied by visualizations that are intended to support the understanding. Is this even possible with your tool or can you somehow purchase them additionally?

Flora: Very good question: Actually, we are primarily active in the text area, so we take care of text simplification. When it comes to visualizations, there are great databases that you can purchase as a supplement and then use them, because visualization is of course also an important point, but at least at the moment we are primarily focusing on the text.

Domingos: A general question, since you are an expert: How do you generally see the future of automated comprehensibility of texts? I would say around 85 percent of people have problems understanding official texts, for example.

Flora: So the future of the topic is above all a huge mountain of existing texts and texts that are produced every day, which are very complicated and which have to be converted into simpler types of text, so simply a huge mountain of work that needs to be done do is. I think what will actually differentiate a bit are different levels of simplification. What we sometimes hear is that The Easy Language sometimes reads a bit bumpy or at least unfamiliar to someone who doesn't necessarily need it. There are then beautiful intermediate levels such as Plain Language, which allows for a little more complexity than Easy Reading. This means that we differentiate even more: Who actually needs what, so that in the end we might even have a topic in the vision where you open your laptop, and then there is something that is understandable for me and for each individual someone else then just plays out information on a different text level. Or perhaps not just text, but also media forms are included at a level of abstraction or complexity, which is then individually understandable. So I think there are still a lot of shades that can be depicted. But for us it's important now that we reach the lowest common denominator in terms of language, because then everyone really comes along. And if you have to read a bumpy text, I personally find it very pleasant and a very good level of language. So our immediate homework is to first make all of these texts accessible and understandable and then I think there will be a lot of differentiation, a lot of individual understanding and individual taste.

Flora's tips for social startups

Domingos: yes, thank you very much for this insight. Maybe the penultimate question: There are many young people like you who would like to do something useful and perhaps also want to get involved in this topic of digital accessibility. With your experience over the last few years, what tips would you like to give them?

Flora: great question. In fact, I think I may have answered part of this before. Namely this topic of networking with the community. So I think that at the beginning you might have the urge to look at the topic and say: I'm building something, that's great, everyone will snatch it out of my hands because it solves the problem so well. But experience shows that you have to start the conversation right from the start and, above all, include people who have already dealt a lot with the topic. And that's what we did at the beginning, especially with the translation community for Easy Reading, and we're continuing to do that. This is not a completed process, because then you can best fit into the world as it really is and not as you imagine it. That was a huge success factor for us that I would recommend to everyone, even though the product might look completely different in the end or the proposed solution that you come away with. It's about solving a real problem and moving us forward as a society, and that's only possible if we take what's happening around us into account.

Domingos: Yes, thank you very much for these tips. The final question would be: How can we continue to track your progress or the development of the SUMM company?

Flora: Also a great question. We do some things on social media, particularly on LinkedIn. We have a company page where we always give updates and let our customers and users have their say, where you can see very well, even if there are improvements to the text, for example, and we have a monthly newsletter We always curate it with great care, which we really enjoy because it is also a document for us of how our company is developing. You can register for it on our website summ-ai.com and stay up to date with how it is continuing.

Domingos: Yes, thank you very much for this insight. The need for understandable language is of course really big and that's why I'm really excited to see how the company and the quality of the text and the whole thing develops further. And thank you for taking the time this afternoon to do this interview and good luck with your company.

Flora: Thank you Domingos for the great questions and yes, your interest in our solution. And you just mentioned the team is growing. We're also really happy when there are people who say: Man, that sounds like exactly where I want to work. If you are interested in becoming part of our team, please take a look at our website. We are happy when the team grows and we can continue to make the world understandable together.

More information

More Talks with Accessibility Specialists