LSTM-based Character-Level Language Model (LM) Training Algorithm

A LSTM-based Character-Level Language Model (LM) Training Algorithm is a neural character-level LM algorithm that is an LSTM-based LM algorithm.

• Context:
  • It can be implemented by an LSTM-based Character-Level Language Modeling System.
• Example(s):
  • the one proposed in (Karpathy, 2015).
  • …
• Counter-Example(s):
  • an MaxLikelihood-based Character-Level LM Algorithm.
  • an LSTM-based Word-Level Language Modeling Algorithm.
• See: Language Modeling.

---

---

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. …

---