從訊息文本中利用語意特徵學習分辨模態概念

Learning to Classify Modality Concepts from Textual Messages Given Linguistic Features

鄭秋函(Cheng, Chiu Han)

指導教授 ： 蘇豐文

國立清華大學/電機資訊學院/資訊系統與應用研究所/碩士(2015年)

https://doi.org/10.6843/NTHU.2015.00386

摘要

由於近年來網路上文字訊息的大量出現，自動分析語意的議題越來越重要且吸引了需多研究者的注意；然而，在沒有一個完善的人類意向和心智態度模型下，要達到好的自動語意分析成果是一件不容易的事情，很難只使用表面文字就可以完全的正確了解語意。模態概念是表達人類的想法或是態度，至今還沒有方法可以彈性化的進行分析。此論文致力於針對模態語句進行知識(epistemic)、義務(deontic)概念的辨別，並納入ConceptNet知識體庫做為特徵，利用支援向量機(support vector machine)機器學習方法去進行分類訓練，另外使用c4.5決策樹、感知器嘗試定義知識、義務的模型。本實驗採用模態語句當作訓練及測試集，並使用交叉驗證，訓練六個類別的分類；本初篇研究的實驗結果顯示，可以對模態類別進行分類，並在後續進行討論。

關鍵字

機器學習 ； 語意特徵 ； 模態概念

並列摘要

Due to prevalence of textual messages on internet, automated opinion analysis
becomes importance and raised much attention of researchers. However, to achieve
high performance of automated opinion analysis is not easy since without profound
models of human intentions and mental attitudes it is hard for computational methods
to infer from embedded messages. Modality concepts are usually associated with
expressing human opinions and attitudes but whose accurate inference is still not yet
computational feasible. This paper attempts to investigate how different modalities such
as deontic and epistemic concepts can be automated classified from textual messages
that are associated with modal sentences. The research adopts a machine learning
algorithms, employ ConceptNet to augment the selection of features and SVM as
supervised learning to train a classifier and uses C4.5 and simple perceptron to define
deontic and epistemic models. The cross validation learning experiments take 844
examples as training and test data set and measure the performance in classifying the
sentences into six different modal categories. We reach performance of F-score up to
71.2% at this preliminary research and subsequent discussions follow.

並列關鍵字

ConceptNet ； Epistemic ； Deontic ； Machine Learning

參考文獻

[2] Blum, A., & Langley, P. (1997). Selection of relevant features and examples in machine learning. Artificial Intelligence 97(1-2) , 245-271.

[3] Cortes, C, Vapnik, V.: Support vector networks, Mach Learn 20: pp. 273-297 (1995)

[5] Downing, A., Locke, P.: A university course in English grammar, London: Prentice-Hall (1992)

[15] Hall, M. (1999). Correlation based feature selection for machine learning. Doctoral dissertation, University of Waikato, Dept. of Computer Science.

[16] Khoo, A., Marom, Y., Albrecht D.: Experiments with sentence classification. In: Proceedings of the 2006 Australasian Language Technology Workshop (ALTW 2006), pp.18-25 (2006)