Split Text by Meaning — Not by Word Count

What does "split text by meaning" mean?

Most text splitters cut by character count. Every 1,000 characters — chop. This produces garbage: sentences split mid-thought, topics scattered across chunks.

Splitting by meaning (also called semantic splitting or topic segmentation) works differently. It analyzes the actual content of the text — which words appear together, where the vocabulary shifts — and finds the natural boundaries where one topic ends and another begins.

The result: segments that respect the structure of the original text. Each segment is a coherent topic unit — perfect for an atomic note, a flashcard, or a database entry.

Why word-count splitting fails

When you split text by word count, you're making an arbitrary decision: "every N words, cut." But text doesn't work that way. Topics don't align with word counts.

How semantic splitting works

TopicSplit uses lexical cohesion analysis — the same technique used in computational linguistics for topic segmentation. Here's the algorithm:

1. Split text into sentences — using punctuation boundaries
2. Extract content words — filter out stop words (the, and, is, etc.)
3. Measure overlap — compute how many content words are shared between adjacent sentences
4. Find breakpoints — when overlap drops significantly, mark a topic boundary
5. Group segments — sentences between boundaries become one topic segment

How to use it for your PKM workflow

Whether you use Obsidian, Logseq, Notion, or a plain text file, TopicSplit fits your workflow:

• Obsidian: Paste an article → split → copy markdown → create atomic notes with ### Topic N headers
• Logseq: Split a research paper → paste blocks directly into your journal
• Notion: Split meeting notes → organize into database entries by topic
• Researchers: Split interview transcripts → segment by theme for analysis
• Writers: Split long-form content → restructure into sections
• Students: Split lecture notes → create study flashcards by topic

TopicSplit vs other text splitters

FAQ

Is this really free?

Yes. MIT licensed, forever free. No paywall, no premium tier, no "pro" upsell. If it saves you time, a $1 Ko-fi tip keeps me building more free tools.

Does it work in languages other than English?

The algorithm is language-agnostic for the splitting logic, but the stop-word list is English. It works reasonably well for Romance languages (Spanish, French, Italian) and decently for Germanic languages. For best results, English text is ideal.

What's the maximum text length?

There's no hard limit — it runs in your browser, so it depends on your device. Articles up to ~10,000 words work instantly. For very long documents (50k+ words), the Pro CLI is faster.

Can I use the output commercially?

Yes. MIT license. Use the output however you want. No attribution required (but appreciated).

How is this different from AI summarizers?

AI summarizers condense text — they rewrite it shorter. TopicSplit segments text — it keeps every word, just organizes it by topic. You get the full content, cleanly structured.