Multiple Domain Answering by Analysing Semantic Relationship in QA Forum

J. Johnson Rajasingh
2018 International Journal for Research in Applied Science and Engineering Technology  
Recent developments made in the web services have applied to the Information retrieval tasks. Semantic matching is a critical task for many applications in several Natural Languages processing like question answering scheme, etc. Key phrases is the subfield that contains metadata that summarizes and characterize the documents. Though, previous techniques were introduced a key phrase extraction model, still the issues like word mismatching, misidentification of the words are not yet focused. In
more » ... ot yet focused. In this paper, we have proposed an efficient key phrase extraction model that efficiently retrieves the relevant data in lesser time. We have constructed machine learning models which build an index for every keyword. Firstly, the keyword is allowed for stemming process that eliminates the stop words in the sentences. Then, the stemmed words is further allowed to build into normalized words that combines with Medinet and Word net. By doing so, we have achieved faster-response time for query retrieval process of the Question Answering scheme. Experimental results have shown the efficiency of the proposed system. I. INTRODUCTION Semantic matching is a critical task for many applications in natural language processing (NLP), such as information retrieval , question answering and paraphrase identification. Taking question answering as an example, given a pair of question and answer, a matching function is required to determine the matching degree between these two sentences. Recently, deep neural network based models have been applied in this area and achieved some important progresses. A lot of deep models follow the paradigm to first represent the whole sentence to a single distributed representation, and then compute similarities between the two vectors to output the matching score. In general, this paradigm is quite straightforward and easy to implement, however, the main disadvantage lies in that important local information is lost when compressing such a complicated sentence into a single vector. Patients seeking online information about their health, connecting patients with doctors worldwide to know about their health via question and answering. Doctors able to interact with many patients about particular issue and provides instant trusted answers for complex and sophisticated problems. Previously external dictionary is used to relate medical data which was not that much sufficient enough. Here we incorporate corpus aware terminology which is used to relate the natural language medical data with medical terminology this narrow down the path between health seekers and health providers. For example: heart attack can also be said as myocardial disorder. A tri-stage framework is used to accomplish the task. A. Noun phase extraction B. Medical concept detection C. Medical concept normalization Due to loss of information global learning approach is used to complement local mining approach. A central topic in developing intelligent search systems is to provide answers in finer-grained text units, rather than to simply rank lists of documents in response to Web queries. This can not only save the users' efforts in fulfilling their information needs, but also will improve the user experience in applications where the output bandwidth is limited, such as mobile Web search and spoken search. Significant progress has been made at answering factoid queries , such as "how many people live in Australia?", as defined in the TREC QA track.
doi:10.22214/ijraset.2018.2001 fatcat:xowfhw5ojzb2xjlukqjetvwvry