شماره ركورد كنفرانس :
4650
عنوان مقاله :
Automatic query-based keyword and keyphrase extraction
پديدآورندگان :
Bayatmakou Farnoush Amirkabir University of Technology , Ahmadi Abbas Amirkabir University of Technology , Mohebi Azadeh Iranian Research Institute for Information Science and Technology
كليدواژه :
Query , based keyword extraction , Automatic keyword and keyphrase extraction , RAKE algorithm
عنوان كنفرانس :
نوزدهمين كنفرانس بين المللي هوش مصنوعي و پردازش سيگنال
چكيده فارسي :
Extracting keywords and keyphrases mainly for identifying content of a document, has an importance role in text processing tasks such as text summarization, information retrieval, and query expansion. In this research, we introduce a new keyword/keyphrase extraction approach in which both single and multi-document keyword/keyphrase extraction techniques are considered. The proposed approach is specifically practical when a user is interested in additional data such as keywords/keyphrases related to a topic or query. In the proposed approach, first a set of documents are retrieved based on user’s query, then a single document keyword extraction method is applied to extract candidate keyword/keyphrases from each retrieved document. Finally, a new re-scoring scheme is introduced to extract final keywords/keyphrases. We have evaluated the proposed method based on the relationship between the final keyword/keyphrases with the initial user query, and based user’s satisfaction. Our experimental results show how much the extracted keywords/keyphrases are relevant and wellmatched with user’s need.
چكيده لاتين :
Extracting keywords and keyphrases mainly for identifying content of a document, has an importance role in text processing tasks such as text summarization, information retrieval, and query expansion. In this research, we introduce a new keyword/keyphrase extraction approach in which both single and multi-document keyword/keyphrase extraction techniques are considered. The proposed approach is specifically practical when a user is interested in additional data such as keywords/keyphrases related to a topic or query. In the proposed approach, first a set of documents are retrieved based on user’s query, then a single document keyword extraction method is applied to extract candidate keyword/keyphrases from each retrieved document. Finally, a new re-scoring scheme is introduced to extract final keywords/keyphrases. We have evaluated the proposed method based on the relationship between the final keyword/keyphrases with the initial user query, and based user’s satisfaction. Our experimental results show how much the extracted keywords/keyphrases are relevant and wellmatched with user’s need.
