ISBN-13: 9783659536496 / Angielski / Miękka / 2014 / 100 str.
Automated question answering (QA) has been a hot topic of research and development since the earliest AI applications (A.M. Turing,1950). A keyword based monolingual restricted domain procedural question answering (PQA) system for cooking recipe in English has been presented as case study. The architecture of the PQA system has been proposed. Also, eight question classes have been proposed to classify procedural text question types. A focused crawler of depth one has also been developed to automatically gather relevant documents from the web. The relevancy of the collected web documents are verified using domain related word frequency. Procedural text structure is analysed and stored in web server. The users of this QA system submits their natural language question into the system from connected machines, and a swallow parser analyzes and classifies the question according to the proposed eight question classes. The system ranks the available documents by relevance score based on two assumptions. Suitable evaluation metrics are also described for procedural questions.
Automated question answering (QA) has been a hot topic of research and development since the earliest AI applications (A.M. Turing,1950). A keyword based monolingual restricted domain procedural question answering (PQA) system for cooking recipe in English has been presented as case study. The architecture of the PQA system has been proposed. Also, eight question classes have been proposed to classify procedural text question types. A focused crawler of depth one has also been developed to automatically gather relevant documents from the web. The relevancy of the collected web documents are verified using domain related word frequency. Procedural text structure is analysed and stored in web server. The users of this QA system submits their natural language question into the system from connected machines, and a swallow parser analyzes and classifies the question according to the proposed eight question classes. The system ranks the available documents by relevance score based on two assumptions. Suitable evaluation metrics are also described for procedural questions.