Updated time: 2023-05-29 10:13:55
05A-1081111
2019-12-19
TW I735380、US 11,568,151 B2
A modifier of a word X is often semantically compatible with X. A complex sentence can usually be composed from a simple sentence by adding compatible modifiers, sub clauses, and complements. Let FB(X) be the collection of compatible modifiers of X. Utilizing FB and sentence structure, we can conversely deduce the original simple sentence. In this invention, we describe how to derive all compatible FB pairs in a sentence, which naturally produces a dependency parse tree (DPT). The action of merging a modifier with X is called reduction. Sentence reduction is performed from the leaves of a DPT recursively to the root, its simplest form. Our DPT can simultaneously help resolve word segmentation, and natural language generation.
- Chinese-word-segmentation and parsing used to be two separate systems. However, the errors of the segmentation system will propagate to the parsing system.
- Provide a new method for named entity recognition, based on a dictionary of label sequences.
- Can handle colloquial and ungrammatical sentences.
- Service chatbot: Help companies answer customers' questions and promote products.
- Machine reading: Help client understand specific Internet text; help transform medical records to database
- Education assistant: Help students with their homework; reinforce the part that students feel difficult.

Fig1. The reduction process of modifiers of the noun“apple.”

Fig2. The incidence map of the whole sentence
Hsu, Wen-Lian
If you are interested in this technology, please fill in the information on the webpage under the manufacturer option (Request for Technology Licensing, Material Transfer, or Sponsored Research) on the website of this office

Reduction in Natural Language Processing