A grammar of natural language is a result of statistical analyses, and is different from that of formal language. Thus, the grammar should be represented in a flexible, changeable formalism. In this paper, we propose a model of common language acquisition in multi-agents. Agents are divided into two groups, each of which has its own language. They try to communicate each other, especially with those which are in the different language group. Becuase the grammars of the languages are represented in the mutable form, the agents change their own grammars and temporarily build up a common grammar, that is of an artificial ‘pidgin,’ that enables the communication. We applied Genetic Algorithm to the statistical mutation of grammar, and utilize Lexicalized Tree Adjoining Grammar that tolerates the change of word-order. In this formalism, we could observe the phenomenon that the majority group assimilates the minority group, preserving grammatical features of each language.
Whereas many studies of implicit memory concentrated on whether conscious/intention to be remembering of retrieve or not, they seldom referred on the intention of encoding. In this study, we focus on the effect of intentional and incidental encoding on implicit and explicit memory. In experiment 1, a forced-forgetting task was set as incidental learning situation. 40 participants were divided into three groups and were tested for word fragment, cued recall and recognition task. In the word fragment task, the forgetting curve indicated that incidental implicit memory was retained longitudinally, but showed partial inhibit effect that first took a down trend and went up again. Also, cued recall task and recognition task showed long-term retention of memory by forced-forgetting. In experiment 2, 31 participants were divided into three group tested for word fragment task, cued recall task and recognition task where foveal vision was set as intentional learning and peripheral vision as incidental learning. As a result of word fragment task, it was suggested that the forgetting curve of implicit memory by incidental learning drew an U-shape curve. Also, the record of cued recall task and recognition task, as well as experiment 1, showed long-range retain of memory by incidental learning. In experiment 3, 41 participants were tested for word fragment task in which figure-treat task was set as incidental-learning situation. The result showed, like experiment 1, 2, an U-shape forgetting curve in incidental learning condition. The findings suggest that the U-shaped forgetting curve are robust in various incidental-learning situations.
In this paper, we investigated the effects of hint-presentation and repetitive production on generation of novel ideas through three experiments. We asked subjects to generate nonwords using Hiragana characters as many as possible for 5 minutes. The generated nonwords dominantly contained characters of Seion category, but its frequency decreased when subjects were presented with characters of other categories (Dakuon category in Experiment 1 and Choon and Sokuon categories in Experiment 2) before the task. Experiment 3 revealed that, in order to decrease the Seion, it was more effective to show a variety of characters as exemplars than showing all the available characters as a chart. The Seion tended to decrease also when subjects repeatedly produced the nonwords. Further analyses showed that repetitive use of same characters linearly increased while the variety of used characters reached ceiling as subjects repeatedly generated nonwords. It was also revealed that the transition of characters within produced nonwords was strongly influenced by normative order of Hiragana characters. We discussed these findings in terms of strategy to enhance creativity.
We investigated quantitatively and systematically the integrative spatial representation of extrapersonal space in walking distance with reference to both body- and environment-centered coordinate systems. A target was presented at one of the five different locations in 2m ahead from subjects for 2sec. Subjects were required to remember the location of presented target, and after a delay of 3sec, to point to the remembered target location. Under the complete dark, pointing errors lay consistently in displaced locations toward the body position, the magnitude of which was shown as a linear function of the distance from subjects' body position to target. Given a specific allocentric visual cue, the frames of the target area, in both target presentation and recall, the errors decreased for any targets. Our findings of pointing error could be explained by the weighted sum of two types of pointing error on the assumption of probabilistic independency. One type of error was would occur in the transformation process solely from the egocentric spatial representation of target with reference to subjects' body and head directing to the front. The other was shown as displacement toward the nearer frame of target presentation area. This type of error would occur in the transformation solely of the allocentric representation with reference to the frames. These results suggested (1) that the quantitative characteristic of pointing solely on the allocentric spatial representation of targets which would not be isolate experimentally, could be estimated on the assumption of probabilistic independency of the behavioral characteristic solely on the egocentric spatial representation, and (2) that for both egocentric and the allocentiric spatial representation, the basic characteristic of pointing in walking distance was shown as the errors displaced toward the center of the spatial coordinate systems.
We investigated cognitive processes of the information seeking on the WWW and the effects of subjects' knowledge and experience on their processes and performance through the protocol analysis. In our experiments, the subjects were assigned to an expert or a novice condition, and required to solve two search tasks. We analyzed the protocol data on the basis of the two space model of scientific discovery suggested by Simon & Lea (1981). We applied this model to the information seeking on the WWW. The information seeking processes on the WWW were considered as the processes of searching the Keyword space and the Web space. The experimental results showed that (1) due to the effects of knowledge and experience, the experts' performance of finding a target was higher than that of the novices, (2) when searching each space, the experts searched each space in detail but the novices did not, and (3) when shifting from one space to the other, the experts searched the two spaces alternately, but the novices tended to cling to a search of one of the two spaces.
Object constancy is essential to the manual interaction of an observer with various objects in the three-dimensional environment. In order to achieve object constancy, the employment of global structure is useful for natural objects, especially artifacts. Estimating the three-dimensional global structure of an object from two-dimensional retinal image depends on how an object-centered reference frame is assigned to the object. In the present study, we studied the effects of object discriminability based on global structure on the object identity matching of three-dimensional familiar objects from two different views: canonical view and non-canonical view. Three experiments were conducted: sequential matching (Experiment 1), simultaneous matching (Experiment 2), and sequential matching using silhouettes (Experiment 3). Findings are summarized as follows: 1) Global structure affected the matching performance differentially based on view canonicality; 2) The discrimination process based on global structure operates at the perceptual stage of recognition; and 3) The effectiveness of outline shape on the access to semantic knowledge depended on view canonicality. These results suggest that the global structure of an object is stored in semantic knowledge, and serves as the basis of recognition of familiar objects. Since the global structure of an object in the natural environment is easily obtained, object recognition based on global structure underlies rapid performance in everyday life.
In this paper, I claim that utterance-final elements in Japanese such as desu and masu indicate that a possible unit of the current speaker's turn has already completed; the hearer can start the next turn as soon as he recognizes these utterance-final elements. The subjects in the experiments rate overlapping utterance pairs according to the degree of their interruptiveness. The utterance pairs are constructed so as to involve various overlap timings by manipulating the starting points of the second utterances in the pairs. The experiment 1 shows that overlaps at utterance-final elements are less interruptive than those before utterance-final elements. The experiment 2 shows that the degree of the interruptiveness decreases as the starting point of the overlap become later in an utterance-final element, but the declination becomes moderate when the starting point come to the middle of the utterance-final element. These results sugest that the hearer can start his turn without interrupting the current speaker soon after he recognizes utterance-final elements, which serve as ‘retrospective’ indication of a possible completion of the current speaker's turn.
Longitudinal analysis of word and sentence acquisition in early language development is critically important. In this paper, we examine two issues using quantitative methodology: (1) the common features of language learning in two children, whose learning speeds were remarkably different, (2) the relationship between word and sentence acquisition. From an analysis of digital audio recordings of conversations between children and their parents, we identified the following common developmental features: (1) the ratio of lexical category used by infants was constant up to 48 months, although the children were quite different in maximal speed of lexical acquisition (56 and 30 words/month), (2) from quantitative analysis of the data, including data in several previous reports, vocabulary spurt begins during the period from fifteen to twenty months and continues for about one year, and the midpoint of the vocabulary spurt period is negatively correlated with maximal speed of lexical acquisition. (3) the use of 100 and 300 words corresponded to the appearance of two and three word sentences, respectively, (4) Mean Length of Utterance (MLU) began to increase around at the month of the vocabulary spurt, (5) both maximal speed of lexical acquisition and increasing rate of MLU of S. A. are as twice as those of S. K. It strongly suggests from these quantitative results that word acquisition process closely interacts with sentence acquisition process.
We investigated the real-time processing of two types of empty subject sentences. In a “subject-oriented sentence” such as “Taroo-ga Hanako-ni Tokyo iki-o hakuzyoosita (Taroo confessed to Hanako that he would go to Tokyo),” the person who will go to Tokyo is the subject Taroo. On the other hand, in an “object-oriented sentence” such as “Taroo-ga Hanako-ni Tokyo iki-o meireisita (Taroo ordered Hanako, (saying) she would go to Tokyo),” the object Hanako will go to Tokyo. Oda et al. (1997) found a “subject preference” phenomenon, since the main clause subject was preferred as the default antecedent for the empty subject. Furthermore, it was suggested that the human parser made top-down prediction at the beginning, and then changed to bottom-up processing by 900 msec after the end of the sentence. The stimuli were given after the end of each sentence. Consequently, it remained open whether the participants make any prediction while listening a sentence. By presenting the stimuli at six checkpoints prior to the end of the sentence, thus, the present study examined the transition of processing before the final verb has been shown. The overall results of the experiment revealed the following two findings: (i) “Yes” responses were faster than “No” responses in the subject-oriented sentences. Thus, it was assumed that the parser adopted a “subject preference” as a default strategy. (ii) The response times drastically increased at the checkpoint preceding the final verb. Therefore, it was suggested that the parser started to process a sentence before the final verb appeared.
This paper suggests that the induction field in vision can predict goodness and strength of impressions which we evaluate for images. Recently, the induction field has been focused on as a method to evaluate sensibility. Sensibility means feelings with mind for image or sound. Many research activities using induction field make efforts to evaluate sensibility of images. However, these previous methods strongly depend on type of image. There is no basic evaluating method using induction field for sensibility. This paper proposes complexity of induction field as an indexe of goodness, and potential energy as strength of impressions. To confirm these indexes psychological experiments were done. It is indicated that these measurements using induction field can evaluate goodness and strength of impressions of images.