• Journal of KIISE:Software and Applications
• /
• v.37 no.6
• /
• pp.446-454
• /
• 2010

According to the original chart parsing algorithm, a sentence is parsed in a strict left-to-right order. The modified chart parsing algorithm proposed in this paper breaks the strictness. With the proposed algorithm, a sentence is parsed in a random order. Conventional left-to-right parsing is also possible, since left-to-right parsing is a special case of random-order parsing. The proposed parsing algorithm is an extension of chart parsing algorithm and its control structure is very simple, so that it is easy to implement the algorithm.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.8
• /
• pp.2849-2864
• /
• 2021

The emerging of cloud data sharing can create great values, especially in multi-cloud environments. However, "data island" between different cloud service providers (CSPs) has drawn trust problem in data sharing, causing contradictions with the increasing sharing need of cloud data users. And how to ensure the data value for both data owner and data user before sharing, is another challenge limiting massive data sharing in the multi-cloud environments. To solve the problems above, we propose a Uniform Encoding and Parsing Framework (UEPF) with blockchain to support trustworthy and valuable data sharing. We design namespace-based unique identifier pair to support data description corresponding with data in multi-cloud, and build a blockchain-based data encoding protocol to manage the metadata with identifier pair in the blockchain ledger. To share data in multi-cloud, we build a data parsing protocol with smart contract to query and get the sharing cloud data efficiently. We also build identifier updating protocol to satisfy the dynamicity of data, and data check protocol to ensure the validity of data. Theoretical analysis and experiment results show that UEPF is pretty efficient.

• Proceedings of the Korean Information Science Society Conference
• /
• 2000.10b
• /
• pp.176-178
• /
• 2000

자연어 파싱에 많이 사용되는 Earley 파싱 알고리즘은 입력문장에 에러(error)가 있으면 즉시 종료되기 때문에 견고한(robust) 파싱을 구현하기 어렵다. 본 논문에서 우리는 Earley 파싱 알고리즘을 보다 견고한 파싱 기법으로 만드는 방법을 제안한다. Earley 파싱을 하다가 멈추면 파싱 모드를 지역적 양방향 분석으로 전환시킨다. 에러 위치 다음에 나타나는 단어를 아일랜드(island)로 정한다. 아일랜드를 지역적으로 양방향으로 확장시켜 에어 위치까지 도달하게 한 다음 에러의 종류를 파악하고 이를 복구하는 기법을 사용함으로써 견고성을 얻을 수 있다.