Character-Level Language Model (LM) Training Algorithm

A Character-Level Language Model (LM) Training Algorithm is a language modeling algorithm that operates that the string character-level (to produce a character-level LM).

• Context:
  • It can be implemented by a Character-Level LM System (to solve a character-level LM task).
• Example(s):
  • a Neural-based Character-Level Language Modeling Algorithm, such as: (Karpathy, 2015).
  • a MaxLikelihood-based Character-Level Language Modeling Algorithm, such as: (Goldberg, 2015).
  • …
• Counter-Example(s):
  • a Word-Level Language Modeling Method.
• See: Maximum Likelihood-based Character-Level Language Modeling Algorithm.

---

---

References

2015

• (Karpathy, 2015) ⇒ Andrej Karpathy. (2015). “The Unreasonable Effectiveness of Recurrent Neural Networks.” In: Proceedings of Blog post 2015-05-21.
  • QUOTE: ... By the way, together with this post I am also releasing code on Github that allows you to train character-level language models based on multi-layer LSTMs. You give it a large chunk of text and it will learn to generate text like it one character at a time. …

---