Enhancement of a language model using two separate corpora of distinct characteristics

언어 모델은 음성 인식이나 필기체 문자 인식 등에서 다음 단어를 예측함으로써 인식률을 높이게 된다. 그러나 언어 모델은 그 도메인에 따라 모두 다르며 충분한 분량의 말뭉치를 수집하는 것이 거의 불가능하다. 본 논문에서는 N그램 방식의 언어모델을 구축함에 있어서 크기가 제한적인 말뭉치의 한계를 극복하기 위하여 두개의 말뭉치, 즉 소규모의 구어체 말뭉치와 대규모의 문어체 말뭉치의 통계를 이용하는 방법을 제시한다. 이 이론을 검증하기 위하여 수십만 단어 규모의 방송용 말뭉치에 수백만 이상의 신문 말뭉치를 결합하여 방송 스크립트에 대한 퍼플렉시티를 30% 향상시킨 결과를 획득하였다.

Language models are essential in predicting the next word in a spoken sentence, thereby enhancing the speech recognition accuracy, among other things. However, spoken language domains are too numerous, and therefore developers suffer from the lack of corpora with sufficient sizes. This paper proposes a method of combining two n-gram language models, one constructed from a very small corpus of the right domain of interest, the other constructed from a large but less adequate corpus, resulting in a significantly enhanced language model. This method is based on the observation that a small corpus from the right domain has high quality n-grams but has serious sparseness problem, while a large corpus from a different domain has more n-gram statistics but incorrectly biased. With our approach, two n-gram statistics are combined by extending the idea of Katz's backoff and therefore is called a dual-source backoff. We ran experiments with 3-gram language models constructed from newspaper corpora of several million to tens of million words together with models from smaller broadcast news corpora. The target domain was broadcast news. We obtained significant improvement (30%) by incorporating a small corpus around one thirtieth size of the newspaper corpus.