-
Notifications
You must be signed in to change notification settings - Fork 5
System Components
In this system, the question type that we are about to solve is YES/NO.
Some examples of YES/NO question are “Is thrombophilia related to increased risk of miscarriage?”, “Does HER2 under-expression lead to favorable response to trastuzumab?” and “Are there any DNMT3 proteins present in plants?”.
Unlike information retrieval systems and Web search engines, which typically return lists of relevant documents, QA systems for document collections aim to return exact answers (e.g., names, dates) or snippets (e.g., sentences) that contain the answers sought, typically by applying further processing to the user’s question and the relevant documents that an information retrieval engine has returned.
The user provides the question that need to be answered. At first the system will analysis the question itself and determine that which type it is. Then the system need to figure out that what content is the question asking for. Moreover, the system need to generate a retrieval query with several retrieval component for the following processing.
In details, the questions need to be translated into collections of terms and phrases, and each term or phrase can be seen as a concept which should be stored in the cas. The main idea of generate the concept is to provide them for the document retrieval engine and retrieval the related documents.
The main function of document retrieval is to retrieve a list of documents that related to the question. As the question analysis part already generate a series of concepts for the user’s question, the document retrieval engine will make use of these concepts and return a list of relevant documents for the question. For different kinds of document retrieval engine, we could get either an unranked set of documents that are likely to contain an answer, or a ranked list of documents by their possibilities of containing the target answer.
The main function of the document processing component is to search the list of documents that provided by the document retrieval component and the concept query that provided by question analysis component. Then The a list of candidate answers will be provided to the answer selection component.
The main function of answer analysis is to select the final answer from the list of candidate answers that provided by the document processing component. And then provide this answer to the user.