NTU NLP Lab System at SemEval-2018 Task 10: Verifying Semantic Differences by Integrating Distributional Information and Expert Knowledge

Proceedings of the 12th International Workshop on Semantic Evaluation (SemEval-2018), pages 1027–1033 New Orleans, Louisiana, June 5–6, 2018. ©2018 Association for Computational Linguistics NTU NLP Lab System at SemEval-2018 Task 10: Verifying Semantic Differences by Integrating Distributional Information and Expert Knowledge Yow-Ting Shiue1, Hen-Hsen Huang1, and Hsin-Hsi Chen1,2 1Department of Computer Science and Information Engineering, National Taiwan University, Taipei, Taiwan 2MOST Joint Research Center for AI Technology and All Vista Healthcare, Taiwan orina1123@gmail.com,hhhuang@nlg.csie.ntu.edu.tw,hhchen@ntu.edu.tw Abstract This paper presents the NTU NLP Lab system for the SemEval-2018 Capturing Discrimina- tive Attributes task. Word embeddings, point- wise mutual information (PMI), ConceptNet edges and shortest path lengths are utilized as input features to build binary classifiers to tell whether an attribute is discriminative for a pair of concepts. Our neural network model reaches about 73% F1 score on the test set and ranks the 3rd in the task. Though the attributes to deal with in this task are all visual, our mod- els are not provided with any image data. The results indicate that visual information can be derived from textual data. 1 Introduction Modern semantic models are good at captur- ing semantic similarity and relatedness. The widely-used distributional word representations, or word embeddings, have achieved promising performance on various semantic tasks. The word pair similarities calculated with these models are to some extent consistent with human judgments, and many downstream applications such as senti- ment analysis and machine translation have bene- fited from word embeddings’ ability to aggregate the information of lexical items with similar mean- ing but different surface forms. However, the ability to distinguish one concept from another similar concept is also core to lin- guistic competence. Our knowledge about what is a “subway”, for example, may contain “it is a kind of train that runs underground”. Also, discriminat- ing things is an important mechanism for teaching and learning. For example, if we would like to ex- plain how a “plate” is different from a “bowl”, we may use expressions like “a plate is flatter” or “a bowl is deeper”. All these examples show that one form of semantic difference is a discriminative at- tribute which applies to one of the two concepts being compared but does not apply to the other. In the SemEval-2018 Capturing Discriminative Attributes task (Krebs et al., 2018), participants need to put forward semantic models that are aware of semantic differences. A data instance consists of a triple and a label. In this paper, we denote a triple with < w1, w2, a >, in which w1 and w2 are the two words (concepts) to be com- pared, and a is an attribute. The label is either positive (1) or negative (0). In a positive exam- ple, a is an attribute of w1 but not an attribute of w2. For negative examples, there are two cases: 1) both w1 and w2 have attribute a ; 2) neither w1 nor w2 has attribute a. In this task, a is limited to visual ones such as color and …

Content truncated to 500 words