Minimum Discrimination Information-based Language Model Adaptation Using Tiny Domain Corpora for Intelligent Personal Assistants
Cited 0 times inCited 0 times in
- Minimum Discrimination Information-based Language Model Adaptation Using Tiny Domain Corpora for Intelligent Personal Assistants
- Jang, Gil-Jin; Kim, Saejoon; Kim, Ji-Hwan
- Adaptation methods; Conditional distribution; Constraint estimation; Discrete distribution; Language model; Language model adaptation; Minimum discriminationinformation; Personal assistants; Relative performance; Tiny domaincorpus; Word frequencies; Word similarity; Wordnet
- Issue Date
- IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
- IEEE TRANSACTIONS ON CONSUMER ELECTRONICS, v.58, no.4, pp.1359 - 1365
- This paper proposes a novel Language Model (LM) adaptation method based on Minimum Discrimination Information (MDI). In the proposed method, a background LM is viewed as a discrete distribution and an adapted LM is built to be as close as possible to the background LM, while satisfying unigram constraint. This is due to the fact that there is a limited amount of domain corpus available for the adaptation of a natural language-based intelligent personal assistant system. Two unigram constraint estimation methods are proposed: one based on word frequency in the domain corpus, and one based on word similarity estimated from WordNet. In terms of the adapted LM's perplexity using word frequency in tiny domain corpora (ranging from 30 similar to 120 seconds in length) the relative performance improvements are measured at 13.9%similar to 16.6%. Further relative performance improvements (1.5%similar to 2.4%) are observed when WordNet is used to generate word similarities. These successes express an efficient ways for re-scaling and normalizing the conditional distribution, which uses an interpolation-based LM1.
- ; Go to Link
Appears in Collections:
- ECE_Journal Papers
can give you direct access to the published full text of this article. (UNISTARs only)
Show full item record
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.