Modeling human dialogue with computers

Conversation, talk, the communicative process, can be compared to the symphonic play of a piece of music. There is an orchestra, the musicians, whose tones and notes must flow, complement and harmonize with one another. There is a main theme. As the music builds variations on the theme are played, and new themes and subthemes are introduced. The basinett responds to the strings, the bass emphasizes the mood of the violin, while the french horn adds a new melody. The subtlety and evenness of flow, the seeming simplicity of the piece, the rich interleaving of subdevelopments, variations, tangents, and recapitulations, are what make a piece of music great.Coherent communication is much the same. It looks easy. It flows. It seems effortless. In reality, of course, it is rich and complex, in both form and content. As in music, conversants may begin a conversation slowly, tentatively, with some simple (or complex) topic in mind. Slowly the momentum builds, the connections surface, and the themes become defined. As in music, these themes are then elaborated upon in various ways. Topic elaborations often spawn their own variations. New themes and subthemes are introduced. All the while, as in music, original themes reappear and are interwoven with the new ones. The twists and turns of the conversation seem to occur effortlessly.In music, there is usually one composer in control of which instrument plays next; one composer who is in charge of what the instrument will say. There is a single coordinator of theme introductions, developments, variations, and interweavings. The composition is done off-line and usually involves editing and re-editing for flaws. In discourse, however, we have on-line development with multiple parties responsible for flow coordination. All parties have to organize their own turns, themes, and variations, in addition to their being able to integrate these items into co-conversants'.In their book, Understanding Computers and Cognition, Winograd and Flores (1987) cite the following from Gadamer (Winograd and Flores, 1987, p. 32):It is not so much our judgments as it is our prejudices that constitute our being ... the historicity of our existence entails that prejudices, in the literal sense of the word, constitute the initial directedness of our whole ability to experience. Prejudices are biases of our openness to the world. They are simply conditions whereby we experience something — whereby what we encounter says something to us. (Gadamer, Philosophical Hermeneutics, 1976, p. 9)     

Rationale for a pragma-dialectical perspective

Starting from a concept of reasonableness as well-consideredness, it is discussed in what way science could serve as a model for reasonable argumentation. It turns out that in order to be reasonable two requirements have to be fulfilled. The argumentation should comply with rules which are both problem-valid and intersubjectively valid. Geometrical and anthropological perspectives don't meet these criteria, but a critical perspective does. It is explained that a pragma-dialectical approach to argumentation which agrees with this critical perspective is indeed problem-valid and that strong pragmatic and utilitarian arguments can be given for its intersubjective validity. Thus, conventional validity is promoted for a code of conduct for discussants who want to resolve their disputes reasonably by way of a critical discussion.     

ACQUISITION OF EXPRESSIVE SIGNING BY AUTISTIC CHILDREN: AN EVALUATION OF THE RELATIVE EFFECTS OF SIMULTANEOUS COMMUNICATION AND SIGN-ALONE TRAINING

Two methods of training autistic children to use manual signs were compared. Two children, one mute and one capable of some verbal imitation, were taught to use signs as expressive labels for pictures of objects. Using an alternating treatments design, speed of sign acquisition was compared across two training conditions in which signs were presented either accompanied by, or without, the corresponding verbal label. In both conditions, the training procedure incorporated reinforcement, modeling, prompting, fading, and stimulus rotation. The efficacy of training in both treatment conditions was demonstrated by the use of a multiple baseline control across signs, but no clear differences in acquisition speed across conditions were apparent. Posttests conducted to assess stimulus control of signing, and learning of verbal labels when these were present in training, showed that the behavior of the imitative, but not the mute, child was controlled by the verbal stimuli. The implications of the results both for understanding deficits characteristic of autistic children and for developing appropriate language training procedures are discussed.     

Stressed and word-final syllables in infant-directed speech

Comparisons between infant-directed and adult-directed speech were conducted to determine whether word-final syllables are highlighted in infant-directed speech. Samples of adult-directed and infant-directed speech were collected from 8 mothers of 6-month-old and 8 mothers of 9- month-old infants. Mothers were asked to label seven objects both to an experimenter and to their infant. Duration, pitch, and amplitude were measured for whole words and for each of the target word syllables. As in prior research, the infant-directed targets were higher pitched and longer than adult-directed targets. The results also extend beyond previous results in showing that lengthening of final syllables in infant-directed speech is particularly exaggerated. Results of analyses comparing word-final versus nonfinal unstressed syllables in utterance-medial position in infant-directed speech showed that lengthening of unstressed word-final syllables occurs even in utterance-internal positions. These results could suggest a mechanism for proposals that word-final syllables are perceptually salient to young children.     

Presumptions and the Distribution of Argumentative Burdens in Acts of Proposing and Accusing

This paper joins the voices warning against hasty transference of legal concepts of presumption to other kinds of argumentation, especially to deliberation about future acts and policies. Comparison of the pragmatics which respectively constitute the illocutionary acts of accusing and proposing reveals important differences in the ways presumptions prompt accusers and proposers to undertake probative responsibilities and, also, points to corresponding differences in their probative duties. This comparison has theoretically important implication regarding the norms governing persuasive argumentation. The paper is based on a broadly Gricean account of speech acts.     

Discrimination of pitch contours by neonates

With the high-amplitude sucking procedure, newborns were presented with two lists of phonetically varied Japanese words differing in pitch contour. Discrimination of the lists was found, thus indicating that newborns are able to extract pitch contour information at the word level.     

INCREASING AUTISTIC CHILDREN'S DAILY SPONTANEOUS SPEECH

We investigated the effectiveness of teaching parents of 3 autistic boys to use a time delay procedure to increase their children's appropriate spontaneous speech in several naturally occurring daily settings (e.g., saying “good morning” in the morning). Generalization across settings and within settings across persons and locations was assessed. Variation in the children's spontaneous speech was also assessed. In addition procedural errors in the parents' use of time delay were calculated. Results indicated that all children increased their daily spontaneous speech and generalized their speech to other locations and persons.     

言语产生中双词素词的语音编码

采用同音判断和音节监控方法 ,考察言语产生中双词素词语音激活的特点。选用以偏正结构的双词素词为名称的图片和与双词素词中首尾两个词素同音的两组探测字作为实验材料 ,探测字在图片呈现后 5 0毫秒 (实验一 ) ,或图片呈现前 130 0毫秒 (实验二 )出现 ,要求被试判断出现的字是否与图片名称中任何一个词素同音。实验结果与绝大部分言语产生理论的预期相反 ,对图片名称第一词素的反应慢于第二词素。考虑到首尾两个词素对整词意义贡献的差别 ,作者把实验结果解释为词素意义对语音激活的作用 :第二词素的语义重要性决定了对应音节激活的快速性 ,双词素词中词素音位激活的速度和时间性受词义和词素意义激活程度的影响 ,而不完全取决于词素发音的序列性。实验不支持音位编码从左到右、序列进行的观点     

言语产生中的词汇通达理论

首先简单介绍了言语产生中词汇产生研究方法,然后重点介绍了言语产生中的2类重要的词汇通达理论：交互激活理论和独立阶段理论,围绕词汇通达中的3个重要问题：词汇通达的阶段、词汇通达的时间进程、词汇通达阶段是分离独立的还是交互作用的展开论述,阐述了2类理论的相同点和相异点。     

言语生成中内隐学习现象的研究

言语生成中的结构启动和语误现象,通常分别用瞬时激活和语音的约束范围来解释它们产生的机制。20世纪60年代末发端的内隐学习的研究,为深入理解言语生成机制提供了崭新的视角。言语生成的各种现象中,语言经验和学习机制发挥着重要作用,结构启动和语误更倾向于是一种内隐的序列学习。