Article Preview
Top1 Introduction
Question Answering (QA) is the fusion of natural language processing (NLP), Machine Learning (ML) and Semantic Analysis. It is used everywhere in various domains such as medical, education systems, personal assistants. In the last decade, there has been significant growth in a new part of the internet, namely the semantic web. The semantic web is developed with the motive to link the data available across multiple web pages, organize them in such a way that the data is directly readable by machines. It includes data sources in different forms. The structured database contains information that is organized and easy to access.
A substantial increase in the amount of research work over semantic web data creates an interaction model that enables people to benefit from semantic web standards. The QA process over dynamic data sources appears to be the most optimistic method to access information. At the same time, Question Answering systems hide their complexity behind an intuitive interface that is easy to use. The spontaneous increase in the semantic web data has resulted in heterogeneous data, as a result of this many systems has to compete with the types, quantity, of data sources.
Most Semantic Web Resources use RDF Query language to access the database, which implies that the user has to learn a query language to scan the semantic web. To facilitate that, the developed Question Answering System that helps people to retrieve data from the database they want. It can be useful in preserving their private Knowledge Bases(KB) within the organizations. They can trust the system to access the data sources, other than depending on additional services. The Question Answering system can enact any data source, function better because it has developed to manage independent and domain-dependent applications. It varies depending on the types of inputs, maybe a keyword or a description. It also requires factoid, affirmation type of questions, and have a clear understanding, reasoning of reality, the origins of data from various domains. Thus, the scope of QA is enormous and widely acceptable.
In Real time application, the Text Retrieval Conference (TREC) (Voorhees, 2001) was the first large scale evaluation of domain-independent, and it consists of open-domain, fact-based questions with broad semantic categories. Questions may also require a specific order of outcome, or aggregate or filter results. The information conveyed in several languages on the internet. While RDF (Lassila & Swick, 1998) data used to represent the tags in several languages, there is no popular language used in Web documents. People have different native tongues. Using QA systems which can handle several input languages and differ from the language used to communicate information is a more flexible solution. Ambiguity is the phenomenon of different definitions of the same word. It may be textual and syntactic, or functional/semantic and lexical.
The QA system begins the process by reviewing the query as an input, then selecting the appropriate KB in which the relevant answer is then to extract the information from the KB and address the user. Based on KB, modern technology used to store complex structured and unorganized data. In the KB, the fact expressed by a triple as subject, predicate, and objects. The predicate word represents the relationship among the entities. The most popular KBs are Dbpedia (Auer et al., 2007), Yago (Suchanek et al., 2007), Wikipedia (Völkel et al., 2006), Freebase (Bollacker et al., 2008), SPARQL (Cyganiak, 2005) is the standard way to access KB. It is a tedious and challenging process for end-users to access/use due to the complexity of understanding the schema and syntax of SPARQL. The common challenges for developing the QA project are complex queries, where the corresponding SPARQL query contains multiple generic graph patterns and specialized approaches needed to obtain the actual query layout.
The intend of the proposed work is to hide the structure of the model by providing a user-friendly system to the clients with high-performance, and it automatically generates the SPARQL query to retrieve the answer from KB. The primary concept of the approach is a template-based selection that improves the accuracy of an answer in minimal time. These templates are also defined for comparatives and superlatives to cover all the cases. The performance level determines the level of accuracy of the answer given by the system.