Lex-POS Sequence and WEOE Embedding Task

A Lex-POS Sequence and WEOE Embedding Task is a GEC Sequence Tagging Task that produces Lex POS Sequences.

• AKA: Lex-POS, ex-POS Sequence Task.
• Context:
  • Task Input: a Word Sequence,
  • Task Output: a Lex-Pos sequence, WEOE Token Vector.
  • Task Requirement(s):
    • a Word Embedding One-Hot Encoding (WEOE) System,
    • a GEC Sequence Tagging System.
  • …
• Example(s):
  • Lex-POS-Approach - GitHub repository.
  • …
• Counter-Example(s):
  • GECToR GEC Sequence Tagging Task,
  • LaserTagger Sequence Tagging Task,
  • PIE-GEC Sequence Tagging Task.
• See: Feature-based Grammar Error Detection Task, Grammar Error Detection Task, Grammar Error Correction Task, Natural Language Processing Task, Part-Of-Speech (POS) Tag, Word Embedding, Word Embedding One-Hot Encoding.

---

---

References

2021

• (GitHub, 2021) ⇒ https://github.com/Machine-Learning-and-Data-Science/Lex-POS-Approach#lex-pos-approach Retrieved:2021-02-27.
  • QUOTE: Lex-Pos sequence has the potential to imbibe the specific nature of the linguistic words (i.e. lexicals) and generic structural characteristics of a sentence via Part-Of-Speech (POS) tags, and so, can lead to a significant improvement in detecting grammar errors. A new vector representation technique, Word Embedding One hot Encoding (WEOE) is introduced to transform Lex-POS into mathematical values.

2020

• (Agarwal et al., 2020) ⇒ Nancy Agarwal, Mudasir Ahmad Wani, and Patrick Bours. (2020). “Lex-Pos Feature-Based Grammar Error Detection System for the English Language.” In: Electronics, 9(10).
  • QUOTE: The main contributions of the work are summarised as follows.
    • A new sequence of the English sentence named Lex-Pos is proposed, which tends to infuse the specificity of linguistic and generalization of syntactic characteristics of a sentence;
    • A novel vector representation for Lex-Pos sequence of sentences named as Word Embedding One-Hot Encoding (WEOE) has been presented by combining the word embedding and one-hot encoded sequences;
    • The novel error induction methods have been proposed to create negative samples containing POS-tag errors for training;
    • The grammar classifier is designed using LSTM deep learning architecture;
    • Overall, nine experiments have been conducted on three designed datasets to reveal the potential of Lex-Pos sequences; and
    • A comparative study is presented where two replicas of existing grammar-aware systems are designed and experiments are conducted to further demonstrate the strength of Lex-Pos sequences.

---