Search

The pervasive use of distributional semantic models or word embeddings for both cognitive modeling and practical application is because of their remarkable ability to represent the meanings of words. However, relatively little effort has been made to explore what types of information are encoded in distributional word vectors. Knowing the internal knowledge embedded in word vectors is important for cognitive modeling using distributional semantic models. Therefore, in this paper, we attempt to identify the knowledge encoded in word vectors by conducting a computational experiment using Binder et al.'s (2016) featural conceptual representations based on neurobiologically motivated attributes. In an experiment, these conceptual vectors are predicted from text-based word vectors using a neural network and linear transformation, and prediction performance is compared among various types of information. The analysis demonstrates that abstract information is generally predicted more accurately by word vectors than perceptual and spatiotemporal information, and specifically, the prediction accuracy of cognitive and social information is higher. Emotional information is also found to be successfully predicted for abstract words. These results indicate that language can be a major source of knowledge about abstract attributes, and they support the recent view that emphasizes the importance of language for abstract concepts. Furthermore, we show that word vectors can capture some types of perceptual and spatiotemporal information about concrete concepts and some relevant word categories. This suggests that language statistics can encode more perceptual knowledge than often expected.

Since Harnad (1990) pointed out the symbol grounding prob-lem, cognitive science research has demonstrated that ground-ing in perceptual or sensorimotor experience is crucial to lan-guage. Recent embodied cognition theories have argued thatlanguage is more important for grounding abstract than con-crete words; abstract words are grounded via language. Dis-tributional semantics has recently addressed the embodied na-ture of language and proposed multimodal semantic models.However, these models are not cognitively plausible becausethey do not address the recent embodiment view of abstractconcepts. Therefore, we propose a novel multimodal distribu-tional semantics in which abstract words are represented indi-rectly through grounded representations of their semanticallyrelated concrete words. A simulation experiment demonstratedthat the proposed model achieved better performance in com-puting the word similarity than other multimodal or text-baseddistributional models. This finding suggests that the indirectembodiment view is plausible and contributes to the improve-ment of multimodal distributional semantics.

Previous studies have suggested that fiction book reading has a stronger positive effect on vocabulary developmentthan nonfiction. In this study, we examined this phenomenon in terms of word appearance information in fiction (story texts),nonfiction (explanation texts), and web text using latent semantic analysis (LSA). In a human experiment with Japanese under-graduates, we replicated fiction (story) text effectivity. Participants who often read story texts achieved the highest vocabularytest scores. Then, in a corpus experiment, we constructed a story text corpus, explanation text corpus, and web text corpus ofidentical size. Based on these corpora, we calculated the LSA similarities between words, and simulated answering the samevocabulary test as used in the human experiment. The corpus experiment demonstrated the nonfiction (explanation) text effec-tively, that is, the explanation corpus was the highest. The cause of discrepancy in the results and the educational implicationsof this study were also discussed.

In this paper, we introduce an interactive guide plate system by adopting a gaze-communicative stuffed-toy robot and a gaze-interactive display board. An attached stuffed-toy robot on the system naturally show anthropomorphic guidance corresponding to the user’s gaze orientation. The guidance is presented through gaze-communicative behaviors of the stuffed-toy robot using joint attention and eye-contact reactions to virtually express its own mind in conjunction with b) vocal guidance and c) projection on the guide plate. We adopted our image-based remote gaze-tracking method to detect the user’s gazing orientation. The results from both empirical studies by subjective / objective evaluations and observations of our demonstration experiments in a semipublic space show i) the total operation of the system, ii) the elicitation of user’s interest by gaze behaviors of the robot, and iii) the effectiveness of the gaze-communicative guide adopting the anthropomorphic robot.