唐詩推薦系統之研究

A Study of Tang Poetry Recommendation System

楊智凱(yang chih k'ai)

指導教授 ： 曾憲雄

亞洲大學/資訊學院/資訊科學與應用學系碩士班/碩士(2007年)

摘要

因應華語文學習市場的崛起，中文越來越受到重視，而中文自古發展之精隨在於韻 文，其中最受矚目的莫過於詩詞，古人將當時的感受及心境融於詩詞當中，本研究即在 探討詩作的意境，從而提供一個數位平台以推薦適合使用者當時心境的詩作。 本論文研究範圍定義於唐代的近體詩作上。詩人在有限的字數內，將心境抒發於詩 作上，詩作流傳自古至今，要透過精鍊文字後的詩作，來瞭解創作者所要表達的意念及 當時的想法為其困難之處，然而藉由詩作來瞭解詩人的心境，甚而將詩作分類更非易事。 為推薦使用者適合當時心境的詩作，我們提出「唐詩推薦系統架構」，此包含5 個 階段：1.唐詩概念階層架構（Concept hierarchy）建立階段：修改同義詞詞林與音樂在情 緒上的模型，重新建構適合於唐詩的概念階層架構；2.詩作情境擷取階段：透過專家的 解析擷取專家對詩作的解釋，以對應出每首詩的屬性；3.詩作分群階段：分析詩作並將 相似程度高的詩作聚為一群；4.詩作風格分類階段：根據詩作屬性建立分類模型，以預 測使用者所屬的風格類別；5.詩作推薦階段：依使用者設定之心境，計算推薦度，以推 薦詩作。

關鍵字

唐詩概念階層架構(concept hierarchy),分群法,分類法,決策樹

並列摘要

Chinese has a rise in popularity due to the spring up of the Chinese language education.
Versification is the core of the Chinese from back in history to the present. However, the
poetry is of most importance. Chinese ancient scholars usually express their thoughts,
experiences and feelings by writing poetry. The research we did is to confer the meaning of
the poetry. Thus, we will provide a platform to recommend the poetry that suits with the
mood of the user.
In this thesis, we focus on the modern poetry of the Tang Dynasty. Chinese ancient
scholars express their thoughts, feelings and mood in the limited number of Chinese
characters by means of poetry. Up to now, it’s difficult to realize the physical meaning of the
poetry. Moreover, it’s also hard to classify the poetry through the mind of the Chinese ancient
scholars.
We propose a Tang Poetry Recommendation System scheme with 5-phase methodology
to recommend Tang poetry which suits the mood of the user: (1) Phase 1: Tang Poetry
Concept Hierarchy Construction, reconstruct the concept hierarchy which suits for Tang
poetry by correct “TongyichCilin” and the model of the emotion of the music; (2) Phase 2:
Poetry Physical Meaning Retrieval, retrieve the physical meaning of each poetry by the
analysis of the expert, and then find the value of each attribute; (3) Phase 3: Poetry
Clustering, analyze the poetry, and then cluster the poetry; (4) Phase 4: Poetry Style
Classification, according to the poetry attributes, construct the classified model to predict the
style of the user; (5) Phase 5: Poetry Recommendation, recommend the poetry according to
the rating of the user.

並列關鍵字

Tang poetry concept hierarchy, clustering, classification, decision tree

參考文獻

[1] C. C. Yeh, et al. , "Building a personalized music emotion prediction system , " Advances

in Multimedia Information Processing-PCM 2006, Lecture Notes in Computer Science 4261,

pp.730-739, Springer-Verlag, November 2006.

do We Need Them? IEEE Intelligent System, Vol. 14, No. 1, pp. 20-26 ,1999.

[3] F. F. Kuo, M. F. Chiang, M. K. Shan and S. Y. Lee, "Emotion-based music