← Latest papers
💬 NLP

Between Century and Poet: Graph-Based Lexical Semantic Change in Persian Poetry

This paper investigates lexical semantic change in Persian poetry by modeling it as the rewiring of local semantic graphs within aligned Word2Vec spaces across centuries and poets, revealing that neighborhood structural shifts better capture historical and relational meaning than simple vector displacement.

Original authors: Kourosh Shahnazari, Seyed Moein Ayyoubzadeh, Mohammadali Keshtparvar

Published 2026-04-09
📖 5 min read🧠 Deep dive

Original authors: Kourosh Shahnazari, Seyed Moein Ayyoubzadeh, Mohammadali Keshtparvar

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are walking through a massive, ancient library filled with Persian poetry. In this library, certain words are like famous actors who have been on stage for a thousand years. Words like "Night," "Wine," "Earth," and "Heart" appear in poem after poem, century after century.

You might think, "If the word is the same, the meaning must be the same, right?"

This paper says: No.

Just because an actor plays the same character name doesn't mean they are playing the same role. A "King" in an epic battle poem is very different from a "King" in a love poem. A "Wine" in a party song is different from a "Wine" in a spiritual prayer.

The researchers wanted to figure out how these words change over time and across different poets. Instead of just looking at the words in isolation, they used a clever new method to see who the words are "hanging out with."

Here is the breakdown of their study using simple analogies:

1. The Old Way vs. The New Way

  • The Old Way (Vector Drift): Imagine tracking a person's GPS location. If they move from New York to London, you know they traveled. But you don't know who they met in London, or if they changed their job, or if they started a new hobby. This is how most computer studies of language work: they just measure how far a word "moved" in a mathematical space.
  • The New Way (Graph-Based Neighborhoods): This study is like looking at a social network. It doesn't just ask, "Did the word move?" It asks, "Who are the word's friends now?"
    • If the word "Night" used to hang out with "Stars" and "Sleep," but now it hangs out with "Separation" and "Waiting," its meaning has changed, even if the word itself looks the same.
    • The researchers built a map where words are dots, and lines connect words that appear together. They watched how these lines got cut and rewired over 1,000 years.

2. The Two Forces Shaping the Words

The study found that words are pulled by two different forces, like a tug-of-war:

  • Force A: Time (The Century): Some words change because the whole world changes.
    • Example: The word "Day" or "Fire." As history moves forward, the way people talk about the sun or fire changes naturally. These words are "Time-Sensitive."
  • Force B: The Poet (The Voice): Some words change because a specific poet has a unique style.
    • Example: The word "Earth." In one poet's hands, "Earth" might mean "dust and humility." In another poet's hands, it might mean "the ground we walk on." These words are "Poet-Sensitive."

3. The "Wine" Lesson (The Best Analogy)

The researchers looked at two words for wine: "Mey" and "Baadeh." This is the most important part of the story.

  • "Mey" (The Wild Party): This word is like a wild, chaotic party. It shows up everywhere. Sometimes it's about getting drunk at a tavern, sometimes it's a metaphor for spiritual love, sometimes it's a joke. Because it does everything, its "friend group" is huge and messy. It's hard to track its history because it's always changing its outfit.
  • "Baadeh" (The Elegant Dinner): This word is like a refined dinner party. It mostly hangs out with words like "Cup," "Cupbearer," and "Toast." Its circle of friends is smaller and more consistent.
  • The Takeaway: The researchers found that "Baadeh" was actually easier to study! Because its "friends" stayed consistent, they could clearly see how its meaning shifted over time. "Mey" was too noisy to tell a clear story.
    • Lesson: Sometimes, to understand history, you don't want the most famous word; you want the one with the clearest story.

4. The "Heart" and the "Dervish"

  • The Heart ("Del"): This word is the emotional anchor. No matter which century or poet you look at, "Heart" is always about feelings, pain, or love. It's very stable. However, the specific feelings around it change (sometimes it's sad, sometimes hopeful). It's like a house that stays in the same spot, but the furniture inside gets rearranged.
  • The Dervish ("Darvish"): This word is a chameleon. In some poems, a Dervish is a holy man. In others, he's a beggar. In others, he's a metaphor for a soul. This word changes its "friends" completely depending on who is writing the poem. It is heavily shaped by the poet's voice.

5. Why This Matters

The authors argue that we shouldn't just look at words as moving dots on a map. We need to look at their social circles.

  • Persistence: The word stays the same.
  • Migration: The word moves to a new group of friends.
  • Mediation: The word becomes a bridge between two different groups.

The Bottom Line

This study is like a detective story about words. It tells us that in Persian poetry, meaning isn't just about the definition of a word. Meaning is about who the word is standing next to.

A word can survive for a thousand years not by staying still, but by constantly making new friends and changing its neighborhood. By looking at these "neighborhoods" (graphs) instead of just the words themselves, we can understand the soul of the poetry much better.

In short: Words are like people. They don't just change because time passes; they change because of who they hang out with, and sometimes, they change because of who is telling the story.

Drowning in papers in your field?

Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.

Try Digest →