Sensicon: An Automatically Constructed Sensorial Lexicon

Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1511–1521, October 25-29, 2014, Doha, Qatar. c⃝2014 Association for Computational Linguistics Sensicon: An Automatically Constructed Sensorial Lexicon Serra Sinem Tekiroğlu University of Trento Fondazione Bruno Kessler Trento, Italy tekiroglu@fbk.eu Gözde Özbal Trento RISE Trento, Italy gozbalde@gmail.com Carlo Strapparava Fondazione Bruno Kessler Trento, Italy strappa@fbk.eu Abstract Connecting words with senses, namely, sight, hearing, taste, smell and touch, to comprehend the sensorial information in language is a straightforward task for hu- mans by using commonsense knowledge. With this in mind, a lexicon associating words with senses would be crucial for the computational tasks aiming at interpreta- tion of language. However, to the best of our knowledge, there is no systematic at- tempt in the literature to build such a re- source. In this paper, we present a senso- rial lexicon that associates English words with senses. To obtain this resource, we apply a computational method based on bootstrapping and corpus statistics. The quality of the resulting lexicon is evaluated with a gold standard created via crowd- sourcing. The results show that a sim- ple classifier relying on the lexicon out- performs two baselines on a sensory clas- sification task, both at word and sentence level, and confirm the soundness of the proposed approach for the construction of the lexicon and the usefulness of the re- source for computational applications. 1 Introduction Sensorial information interpenetrates languages with various semantic roles in different levels since the main interaction instrument of humans with the outside world is the sensory organs. The trans- formation of the raw sensations that we receive through the sensory organs into our understand- ing of the world has been an important philo- sophical topic for centuries. According to a clas- sification that dates back to Aristotle (Johansen, 1997), senses can be categorized into five modali- ties, namely, sight, hearing, taste, smell and touch. With the help of perception, we can process the data coming from our sensory receptors and be- come aware of our environment. While interpret- ing sensory data, we unconsciously use our exist- ing knowledge and experience about the world to create a private experience (Bernstein, 2010). Language has a significant role as our main communication device to convert our private ex- periences to shared representations of the environ- ment that we perceive (Majid and Levinson, 2011). As a basic example, onomatopoeic words, such as knock or woof, are acquired by direct imitation of the sounds allowing us to share the experience of what we hear. As another example, where an im- itation is not possible, is that giving a name to a color, such as blue, provides a tool to describe a visual feature of an object. In addition to the words that describe the direct sensorial features of ob- jects, languages include many other lexical items that are connected to sensory modalities in various semantic roles. For instance, while some words can be used to describe a perception activity (e.g., to sniff, to …

Content truncated to 500 words