Man-Machine interaction in natural language: Computational models of dialogue

This article presents a state-of-the-art review of studies on computational modeling of dialogue. Particular attention is given to the treatment of ill-formed input, the prevention of the hearer’s misconceptions, the inference of the speaker’s plans, the generation of language, and the recognition of dialogue focus—particularly in terms of implications for linguistics. Specifications and initial results of some recent studies carried out by the author are provided. This article is a revised version of a lecture held at the Third Scientific Meeting “Computer Processing of Linguistic Data,” Bled, 1985.     

A functional analysis of receptive language and productive speech: acquisition of the plural morpheme

Operant conditioning procedures were applied to two retardates to establish receptive auditory plurals: correct pointing to single or paired objects was reinforced after hearing singular or plural labels. This training proceeded until an errorless (generative) criterion of correct performance was achieved. Unreinforced probes measuring expressive use of singulars and plurals were interspersed in this receptive training. Neither subject generalized from this receptive training to expressive plurals, in that each used singulars when labeling pairs. Then, both subjects were directly trained in conventional expressive plurals to an errorless (generative) criterion. The previous design was then repeated, but the receptive repertoire was reversed: pointing at pairs in response to singular labels was reinforced, and vice-versa. Unreinforced probes of expressive plural usage again showed its independence of the current receptive repertoire in that conventional (unreversed) plural usage was displayed. Thus, the independence of the expressive repertoire (even when unreinforced) from the reinforced patterns of the receptive repertoire was demonstrated.     

The thematic hierarchy in sentence comprehension: A study on the interaction between verb class and word order in Spanish

Linking is the theory that captures the mapping of the semantic roles of lexical arguments to the syntactic functions of the phrases that realize them. At the sentence level, linking allows us to understand “who did what to whom” in an event. In Spanish, linking has been shown to interact with word order, verb class, and case marking. The current study aims to provide the first piece of experimental evidence about the interplay between word order and verb type in Spanish. We achieve this by adopting role and reference grammar and the extended argument dependency model. Two different types of clauses were examined in a self-paced reading task: clauses with object–experiencer psychological verbs and activity verbs. These types of verbs differ in the way that their syntactic and semantic structures are linked, and thus they provide interesting evidence on how information that belongs to the syntax–semantics interface might influence the predictive and integrative processes of sentence comprehension with alternative word orders. Results indicate that in Spanish, comprehension and processing speed is enhanced when the order of the constituents in the sentence mirrors their ranking on a semantic hierarchy that encodes a verb's lexical semantics. Moreover, results show that during online comprehension, predictive mechanisms based on argument hierarchization are used rapidly to inform the processing system. Our findings corroborate already existing cross-linguistic evidence on the issue and are briefly discussed in the light of other sentence-processing models.     

Repetition priming within and between languages in verb generation: evidence for shared verb concepts

Although translation equivalents for concrete nouns are known to have shared core conceptual representations in bilingual memory (Francis, 1999), the status of translation-equivalent verbs has not been systematically tested. Three repetition-priming experiments using a verb generation task were used to determine whether verbs have shared representations across languages and to identify the processes facilitated in repeated verb generation. In Experiment 1 fluent Spanish-English bilingual speakers exhibited repetition priming both within and between languages, but between-language priming was weaker. In Experiment 2 performance of non-bilingual English and Spanish speakers was equivalent to that of bilingual speakers responding in their dominant language. Experiment 3 used manipulations meant to isolate noun comprehension, verb concept selection, and verb production. The between-language priming in Experiments 1 and 3 indicates that verb concepts are shared across languages and that verb concept selection exhibits facilitation. Experiment 3 showed that the greater within-language priming was due primarily to facilitation of verb production processes.     

Repetition priming within and between languages in verb generation: Evidence for shared verb concepts

Although translation equivalents for concrete nouns are known to have shared core conceptual representations in bilingual memory (Francis, 1999), the status of translation-equivalent verbs has not been systematically tested. Three repetition-priming experiments using a verb generation task were used to determine whether verbs have shared representations across languages and to identify the processes facilitated in repeated verb generation. In Experiment 1 fluent Spanish–English bilingual speakers exhibited repetition priming both within and between languages, but between-language priming was weaker. In Experiment 2 performance of non-bilingual English and Spanish speakers was equivalent to that of bilingual speakers responding in their dominant language. Experiment 3 used manipulations meant to isolate noun comprehension, verb concept selection, and verb production. The between-language priming in Experiments 1 and 3 indicates that verb concepts are shared across languages and that verb concept selection exhibits facilitation. Experiment 3 showed that the greater within-language priming was due primarily to facilitation of verb production processes.     

Fear and pain: investigating the interaction between aversive states

Pain and fear often co-occur and appear to interact, although the nature and direction of their relation is not clearly delineated. The present study investigates how exposure to the experience of one of these states subsequently affects responding to the other. Pressure stimulation and carbon dioxide-enriched air (CO2) were used to induce pain and fear, respectively, in 48 healthy individuals. The order in which the stimuli were introduced was manipulated, as was the CO2 level. Measures of overt behavior, physiological responding, and self-report were obtained; analyses of stimuli effects generally supported their ability to produce pain or fear. Results indicate that the stimulus rated as the most aversive, the higher level of CO2, led to the highest levels of distressed responding across dependent measures. This pattern was replicated for the stimuli found to be less aversive (i.e., pressure stimulation and low-level CO2, respectively). The authors conclude that fear and pain, in and of themselves, do not dictate the nature of their interaction; the most important factor is how aversive they are perceived to be in relation to one another.     

Grammatical aspect,lexical aspect,and event duration constrain the availability of events in narratives

The present study investigates how readers’ representations of narratives are constrained by three sources of temporal information; grammatical aspect, lexical aspect, and the duration of intervening events. Participants read short stories in which a target event with an intrinsic endpoint or not (lexical aspect: accomplishments/activities) was described as ongoing or completed (grammatical aspect: imperfective/perfective). An intervening sentence described either a long or short duration event before the target situation was reintroduced later in the story. The electroencephalogram time-locked to the reintroduction of the target event elicited a larger N400 for perfective versus imperfective accomplishments, and this effect occurred only after short intervening events. Alternatively, the N400 to targets in the activity condition did not vary as a function of grammatical aspect or duration of intervening events. These results provide novel insight into how the temporal properties of events interact to constrain the availability of concepts in situation models.     

On the uses of language in working and idealized logic

The interpretation of discourse covers a continuum with two extremes: on the one hand, a text considered as an ideal, distant object, and on the other hand, a conversation regarded as a real, present event. On the basis of a distinction between relatively context-invariant propositions and relatively context-dependent statements, it is argued that statements in conversational discourse are easier to interpret than statements in texts, whereas only propositions in symbolic logic can be interpreted with exactitude. In the same way, the interpretation of dialogical arguments proceeds more easily than the interpretation of arguments in texts. While dialogical argumentation requires a dialectical approach, textual argumentation necessitates an imaginative reconstruction of the argument. From this it can be concluded that for different sorts of argumentative discourse diverse sorts of interpretative activities have to be used.     

On the mental representations originating during the interaction between language and vision

The interaction between vision and language processing is clearly of interest to both cognitive psychologists and psycholinguists. Recent research has begun to create understanding of the interaction between vision and language in terms of the representational issues involved. In this paper, we first review some of the theoretical and methodological issues in the current vision–language interaction debate. Later, we develop a model that attempts to account for effects of affordances and visual context on language-scene interaction as well as the role of sensorimotor simulation. The paper addresses theoretical issues related to the mental representations that arise when visual and linguistic systems interact.     

Evidence for a non-linguistic distinction between singular and plural sets in rhesus monkeys

Set representations are explicitly expressed in natural language. For example, many languages distinguish between sets and subsets (all vs. some), as well as between singular and plural sets (a cat vs. some cats). Three experiments explored the hypothesis that these representations are language specific, and thus absent from the conceptual resources of non-linguistic animals. We found that rhesus monkeys spontaneously discriminate sets based on a conceptual singular-plural distinction. Under conditions that do not elicit comparisons based on approximate magnitudes or one-to-one correspondence, rhesus monkeys distinguished between singular and plural sets (1 vs. 2 and 1 vs. 5), but not between two plural sets (2 vs. 3, 2 vs. 4, and 2 vs. 5). These results suggest that set-relational distinctions are not a privileged part of natural language, and may have evolved in non-linguistic species to support domain general quantitative computations.     

Learning about what others were doing: verb aspect and attributions of mundane and criminal intent for past actions

Scientists have long been interested in understanding how language shapes the way people relate to others, yet it remains unclear how formal aspects of language influence person perception. We tested whether the attribution of intentionality to a person is influenced by whether the person's behaviors are described as what the person was doing or as what the person did (imperfective vs. perfective aspect). In three experiments, participants who read what a person was doing showed enhanced accessibility of intention-related concepts and attributed more intentionality to the person, compared with participants who read what the person did. This effect of the imperfective aspect was mediated by a more detailed set of imagined actions from which to infer the person's intentions and was found for both mundane and criminal behaviors. Understanding the possible intentions of others is fundamental to social interaction, and our findings show that verb aspect can profoundly influence this process.     

Infant distractibility from social events mediates the relation between maternal responsiveness and infant language outcomes

Research demonstrates that contingent and appropriate maternal responsiveness to infant requests and bids for attention leads to better language outcomes. Research also indicates that infants who are less distracted by irrelevant competing stimulation and attend efficiently to audiovisual social events (e.g., faces and voices) show better language outcomes. However, few studies have assessed relations between maternal responsiveness, infant attention to faces and voices, and distractibility, and how together these factors lead to early language outcomes. A newly developed audiovisual protocol, the Multisensory Attention Assessment Protocol (MAAP; Bahrick et al., 2018), allows researchers to examine individual differences in attention to faces and voices and distractibility, and to assess relations with other variables. At 12 months, infants (n = 79) in an ongoing longitudinal study participated in the MAAP to assess intersensory matching of synchronous faces and voices and attention to an irrelevant competing visual distractor event. They also were observed in a brief play interaction to assess infant bids for attention and maternal responsiveness (accept, redirect, or ignore). At 18 months, receptive and expressive language were assessed using the Mullen Scales of Early Learning. Several noteworthy findings emerged: 1) mothers were generally responsive, accepting 74% and redirecting 14% of infant bids, 2) infants who had a greater number of their bids redirected by mothers, and who had better intersensory matching of synchronous faces and voices, showed less attention to the distractor, and 3) infants who showed less attention to the distractor had better receptive language. Findings demonstrate that maternal redirecting of infant attention by mothers who are generally responsive may promote better infant attentional control (lower distractibility) which in turn predicts better receptive language in toddlers.     

The early acquisition of verb meaning in German by normally developing and language impaired children

The research reported here focuses on the early acquisition of event structure in German. Based on longitudinal studies from 5 normally developing (ND) and 6 language-impaired (LI) children, a model of "event structural bootstrapping" is presented that spells out how ND children log into the verb lexicon. They project a target-consistent event tree, depicting the head-of-event of transitions. Young LI children, failing to employ this bootstrapping strategy, resort to radically underspecified event representations. The results from a truth-value judgment experiment with 16 ND and 16 LI children showed that ND children perform correctly on transitional verbs, while LI children perform at chance level on the same tasks. These findings are accounted for by the model of event structural bootstrapping to the extent that LI children lack an explicit representation of the head-of-event.     

The redundancy of recursion and infinity for natural language

An influential line of thought claims that natural language and arithmetic processing require recursion, a putative hallmark of human cognitive processing (Chomsky in Evolution of human language: biolinguistic perspectives. Cambridge University Press, Cambridge, pp 45–61, 2010; Fitch et al. in Cognition 97(2):179–210, 2005; Hauser et al. in Science 298(5598):1569–1579, 2002). First, we question the need for recursion in human cognitive processing by arguing that a generally simpler and less resource demanding process—iteration—is sufficient to account for human natural language and arithmetic performance. We argue that the only motivation for recursion, the infinity in natural language and arithmetic competence, is equally approachable by iteration and recursion. Second, we submit that the infinity in natural language and arithmetic competence reduces to imagining infinite embedding or concatenation, which is completely independent from the ability to implement infinite processing, and thus, independent from both recursion and iteration. Furthermore, we claim that a property of natural language is physically uncountable finity and not discrete infinity.     

Exemplars and prototypes in natural language concepts: A typicality-based evaluation

Are natural language categories represented by instances of the category or by a summary representation? We used an exemplar model and a prototype model, both derived within the framework of the generalized context model (Nosofsky, 1984, 1986), to predict typicality ratings for 12 superordinate natural language concepts. The models were fitted to typicality ratings averaged across participants and to the typicality judgments of individual participants. Both analyses yielded results in favor of the exemplar model. These results suggest that higher-level natural language concepts are represented by their subordinate members, rather than by a summary representation.