結合類神經網路及文件概念圖之文件檢索研究

Document Retrieval based on Neural Network and Document Concept Graph

倘若搜尋結果能考慮主題或情境相近的內容，便能搜尋到更符合使用者期待的結果。因此，本研究使用類神經網路及文件概念圖，以探討主題或語意相近之檢索內容，實驗結果顯示，在經由類神經網路所訓練的分類器中，最佳macro-F1為70.0％，而結合文件概念圖的計算後，以查詢內容與結果的相似度而言nDCG分數可達0.959，由此可驗證，基於類神經網路及文件概念圖的結果可以補充和加強資訊檢索的表現。

If the search results can consider topics or similar situations, we can find results that are more in line with user’s expectations. Therefore, our research uses neural network and document concept graph to explore the topics or semantics similarity. The experimental results show that the best macro-F1 is 70.0% in the classifier trained via the neural network. Combined with the calculation of the concept graph of the document, the nDCG score can reach 0.959 in terms of the similarity between the search content and the results. This proves that the results based on the neural network and the document concept graph can be used to complement and enhance the performance of information retrieval.