Something you want to read shouldn’t be set aside because of language

You find exactly the material you need. It might be a paper related to your research, a manual that explains a technical problem, or a design proposal worth learning from.

Perhaps you weren’t looking for anything at all. You happened upon a paper and its title caught your attention, or someone shared a point from it and you wanted to know what the author actually said.

You want to keep reading, only to discover that it is written in a language you don’t know well enough.

It isn’t completely beyond you. With translation, you could work through it. But finishing it means setting aside extra time and switching between the document and different tools. So you save the file for later—and may end up putting it aside for good, never opening it again.

Research has tried to quantify this extra burden. A study published in PLOS Biology in 2023 surveyed 908 environmental scientists across eight countries. Among researchers who had published just one English-language paper, it estimated that non-native English speakers from countries with moderate and low English proficiency needed about 47% and 91% more time, respectively, to read papers in English than native speakers. Read the study

The survey took place in 2021. It cannot represent every reader today, and it did not measure how often downloaded files go unread. But it offers a concrete reference for the additional effort of reading in another language.

We hope iSomor can help fewer people give up on reading because of language difficulties. Whether to read something should depend more on your interest in its contents than on the language it uses.

A translation can still leave reading interrupted

Putting sentences into a familiar language solves part of the problem, but it doesn’t necessarily make a whole document easy to follow.

Numbers in a table need to be read with their column headings. A caption beside a figure may explain something difficult in the main text. If translation separates these relationships, the reader has to find and reconnect them. That is also why the original layout is worth preserving in translation.

There is another kind of interruption: you understand a sentence but aren’t sure it retains the original meaning. Was a condition left out? Does a technical term mean what you think it means? You may need to return to the source to check.

At that point, language is no longer the only difficulty. There is also the extra work of organizing information, recovering context, and checking meaning. Attention that could have gone toward understanding is spent on preparation.

This is the unnecessary effort we want to reduce, so reading is easier to begin and easier to continue.

With AI helping us read, why still read the source?

When faced with a long document today, we can ask AI to outline its main ideas, explain difficult passages, locate information around our questions, or help explore what else is worth asking. That support can run throughout the reading process; it does more than shorten a document.

This raises a natural question: if we can ask directly and keep asking follow-up questions, why read the source ourselves?

If you only need to find one piece of information, an answer may meet your immediate need. But when you want to understand an argument, learn a method, or make a judgment based on it, getting an answer and developing an understanding are still different things.

For example, we can ask AI why a method in a paper works, then ask about its limitations. Its explanation can help us approach those questions. Returning to the relevant sections lets us read that explanation alongside the author’s reasoning: what exactly is the evidence, how do the steps connect, and is the conclusion narrower than we understood? Reading the source is not just about checking whether AI is wrong. It is also about seeing how a claim is supported.

Some questions only emerge as we read. An example may bring your own project to mind; a note may change how you understand a number. AI can suggest new questions too, but its reading cues need not become the only path. You can still discover for yourself which parts matter to you and where to pause, rather than leaving that entirely to a round of questions and answers.

We don’t have to choose between using AI and reading ourselves. We can ask AI to organize and locate information, then read the relevant passages; when something is unclear, ask again and return to the context. Not every document needs to be read word for word. But when we need deeper understanding, a quotation, or a basis for judgment, the source still serves a purpose that a summary or a few exchanges cannot fully replace.

By reading the source, we don’t mean everyone should avoid translation and struggle through an unfamiliar language. Reading the material through a translation, then checking the original wording when a question arises, is also a way of approaching the source. We want AI to reduce the language burden in that process, so people who want to continue can do so instead of stopping at prepared answers because the source is difficult to read.

Easier reading doesn’t mean there is no need to think

We don’t believe that translating a document means it has been understood. Once the sentences make sense, questions remain: how does this connect with what I already know, and how might it help me understand the problem in front of me?

Not all reading effort comes from the same place. Some comes from looking up words, switching tools, or finding where you left off. Some comes from grasping an unfamiliar concept, comparing explanations, or reconsidering your view. We want to reduce the first kind so people have energy for the second—not treat every effort as a problem to eliminate.

A report, for example, might describe an approach as “more efficient.” Even after understanding it, you still need to ask whether the report values speed or cost, and whether your own priorities are accuracy, maintenance, or usability. Even a sound comparison may answer a different question from yours. Applying knowledge to your studies or work takes more than accepting a conclusion; it means understanding how it relates to your needs.

Returning to the source does not make its claims necessarily correct, either. The strength of the evidence and the soundness of the reasoning still need judgment. In an unfamiliar field, you may need more background knowledge or another source for comparison. Where evidence is insufficient, you can withhold judgment rather than rushing to accept or reject a claim.

Sometimes, finishing an article doesn’t give you a ready-to-use answer. Instead, you discover that your original question was too broad, or that an important piece of information is missing. That need not mean the reading achieved nothing. Understanding what you don’t yet know can be a starting point for further learning.

AI can take part in this process by explaining concepts, organizing leads, or offering another way of thinking. But both translations and AI answers remain open to checking and questioning. We want tools to create better conditions for thought, not let a fluent answer become the point where thinking ends.

iSomor starts with a PDF

For now, we start with text-based PDFs, offering translation, comparison with the original, and translated-file export while trying to preserve the document’s organization. These capabilities respond to the same need: making material in another language easier to understand, check, and use in subsequent learning or work.

This does not remove every obstacle to reading. Technical terminology, complex layouts, and specific contexts can still cause problems, and important content needs checking. We don’t want “translation complete” to stand in for that step.

Whether something is ultimately worth reading is for you to decide. We hope that when you do want to read it, the language alone won’t send it back to your “read later” folder.