Dialogue Corpus Research Articles

외국인 한국어 학습자를 위한 한국 역사 영상 콘텐츠 문법 연구

This study proposes an approach to engaging foreign learners of Korean with Korean history, which can be somewhat heavy and challenging, by treating it as 'history consumption.' To provide learning materials that aid foreign learners of Korean in watching Korean historical video content, the study compiled a corpus of dialogues from a total of 44 Korean historical video works available on Netflix. The study pointed out that existing history textbooks for foreigners are not suitable for foreign learners of Korean and highlighted the issue that it increases the burden on instructors to thoroughly teach both Korean culture and history in actual Korean language learning settings. As an alternative, the study proposed the use of Korean historical video content and, based on previous research showing that watching videos with subtitles in the target language enhances learning effectiveness, organized frequently occurring grammar items in Korean historical video content by frequency, usage by social status, and usage by gender to assist learning.

Negative politeness and no worries in Australian English

AbstractThis paper explores how a distinctive politeness token can establish itself in a variety of world English by virtue of its pragmatics. It examines the usage of no worries as a negative politeness thanking response in a large multigenerational corpus of informal dialogues by men and women recorded at the University of Western Australia (UWA). Data from this UWA corpus reflect the steady evolution of no worries, from its indexical association with Australian men's talk to be taken up by younger Australian women. Though rooted in laconic spoken exchanges, the data also show extended uses of no worries in quoted speech as well as narrative reconstructions of spoken encounters to highlight their outcomes. In its further linguistic evolution, it works as a stance adverbial in dramatized interior monologues. It thus acquires new functions in various discourse contexts where a modicum of negative politeness adds value to the narrative or argument.

Travel Agency Task Dialogue Corpus: A Multimodal Dataset with Age-Diverse Speakers

When individuals communicate, they use different vocabularies, speaking speeds, facial expressions, and gestural languages, depending on those with whom they are speaking. This study focuses on the age of the speaker as a factor that affects the style of communication. We collected a multimodal dialogue corpus with various speaker ages. We used travel as the topic, as it interests people of all ages, and we set up a task based on a tourism consultation between an operator and a customer at a travel agency. This paper presents the details of the dialogue task, collection procedures and annotations, and analysis of the characteristics of the dialogues and facial expressions, focusing on the age of the speakers. The results of the analysis suggest that the adult speakers have more independent opinions, the older speakers express their opinions more frequently than other age groups, and those in the operator role smiled more frequently at minors.

Yes, I am afraid of the sharks and also wild lions!: A multitask framework for enhancing dialogue generation via knowledge and emotion grounding

Current end-to-end neural conversation models inherently lack the capability to generate coherently engaging responses. Efforts to boost informativeness have an adversarial effect on emotional and factual accuracy, as validated by several sequence-based models. While these issues can be alleviated by access to emotion labels and background knowledge, there is no guarantee of relevance and informativeness in the generated responses. In real dialogue corpus, informative words like named entities, and words that carry specific emotions can often be infrequent and hard to model, and one primary challenge of the dialogue system is how to promote the model’s capability of generating high-quality responses with those informative words. Furthermore, earlier approaches depended on straightforward concatenation techniques that lacked robust representation capabilities in order to account for human emotions. To address this problem, we propose a novel multitask hierarchical encoder–decoder model, which can enhance the multi-turn dialogue response generation by incorporating external textual knowledge and relevant emotions. Experimental results on a benchmark dataset indicate that our model is superior over competitive baselines concerning both automatic and human evaluation.

The Video Game Dialogue Corpus

This paper presents the Video Game Dialogue Corpus, the first large-scale, consistently coded, open source corpus of dialogue from video games. It contains over 6.2 million words of English dialogue from fifty games in the Role Playing Game (rpg) genre. This includes games produced between 1985 and 2020, rated for children, teenagers and adults, and in both ‘Western’ and ‘Japanese’ sub-genres. The corpus design is described, including custom data formats for representing branching dialogue. We demonstrate the use of the corpus by comparing the dialogue of female and male characters, where we find reflections of gendered language in other media as well as patterns that seem specific to video games. We provide the source code for a ‘self-inflating corpus’ – a pipeline that obtains the data then processes and parses it into a standard format. This makes the corpus available for teaching and research purposes, providing the first such resource for empirical analysis of video game dialogue.

A Goal Interaction Graph Planning Framework for Conversational Recommendation

Multi-goal conversational recommender system (MG-CRS) which is more in line with realistic scenarios has attracted a lot of attention. MG-CRS can dynamically capture the demands of users in conversation, continuously engage their interests, and make recommendations. The key of accomplishing these tasks is to plan a reasonable goal sequence which can naturally guide the user to accept the recommended goal. Previous works have demonstrated that mining the correlations of goals from the goal sequences in the dialogue corpus is helpful for recommending the goal that the user is interested in. However, they independently model correlations for each level of goal (i.e., goal type or entity) and neglect the order of goals appear in the dialogue. In this paper, we propose a goal interaction graph planning framework which constructs a directed heterogeneous graph to flexibly model the correlations between any level of goals and retain the order of goals. We design a goal interaction graph learning module to model the goal correlations and propagate goal representations via directed edges, then use an encoder and a dual-way fusion decoder to extract the most relevant information with the current goal from the conversation and domain knowledge, making the next-goal prediction fully exploit the prior goal correlations and user feedback. Finally we generate engaging responses based on the predicted goal sequence to complete the recommendation task. Experiments on two benchmark datasets show that our method achieves significant improvements in both the goal planning and response generation tasks.

Research on discourse role recognition in task-oriented collaborative dialogue

Task-oriented collaborative dialogues have become an indispensable form of communication in our daily work and learning, in which participants exchange ideas and share information to advance goals. It is crucial to automatically analyze participants’ contributions and understand these dialogues relative to individuals with limited attention spans. In this paper, seven Discourse Role (DR) labels are designed to describe discourse’s different roles in collaborative dialogues for goal achievement. We collected about 11K discourses from a publicly available dialogue corpus and annotated them with DR tags to construct a dataset named MRDR (Meeting Recorder Discourse Role). In addition, this paper proposes a novel hierarchical model, STTAHM (Speaker Turn and Topic-Aware Hierarchical Model), for Discourse Role classification. The model is equipped to perceive speaker turn and dialogue topic and can effectively capture the discourse’s local and global semantic information. Experimental results show that our proposed method is effective on the constructed dataset, and the accuracy of Discourse Role classification reaches 86.99%.

The Representation of Earlier African American Vernacular English by Charles W. Chesnutt

This article examines the validity of regional speech represented in the writings of an African American short story writer, essayist, lawyer, and stenographer Charles W. Chesnutt (1858–1932) with a corpus of dialogues extracted from his literary works (approximately 40,000 words). As a touchstone for exploring its accuracy, the Atlas data from Fayetteville, North Carolina, and the surrounding regions in Cape Fear and Pee Dee Valley− a region where the author spent his youth and his stories are set−are compared with the speech of his African American characters. The result reveals that the Atlas records and Chesnutt’s dialogues share many expressions in their regional lexicons ranging from single words to phrases. Considering the difficulty of recording all the synonyms and semantically related words from the same informant during the interview, the overall agreement between the two provides further evidence that as an African American writer who spent his formative years in a rural community in North Carolina, Chesnutt’s literary dialect offers a glimpse of earlier AAVE spoken in the region in the mid-nineteenth century. Furthermore, this article touches on Chesnutt’s racial identity and what led him to enter the literary world as it is pertinent to his linguistic background and the motive behind the use of literary dialect.

Touch-text answer for human-robot interaction via supervised adversarial learning

In daily life, touch modality plays an important role in conveying human intentions and emotions. To further improve touch-based human-robot interaction, robots need to infer human emotions from touch signals and respond accordingly. Therefore, it is a major challenge to correlate the emotional state of touch gestures with text responses. At present, there are few researches on touch-text dialogue, and robots cannot respond to human tactile gestures with appropriate text, so touch-text-based human-robot interaction is not yet possible. To solve these problems, we first built a touch-text dialogue (TTD) corpus based on six basic emotions through experiments, which contains 1109 touch-text sample pairs. And then, we designed a supervised adversarial learning for touch-text answer (SATTA) model to realize the touch-text based human-robot interaction. The SATTA model correlates the data of text mode with that of touch mode by reducing the emotion discrimination loss in the public space and the feature difference between the sample pairs of two modes. At the same time, the feature representation is mapped into the label space to reduce the classification loss of samples. The experiment in the TTD corpus validates the proposed method.

Character expression of a conversational robot for adapting to user personality

User adaptation is important in conversational robots to enhance the user experience and engagement. One way of user adaptation is to change the dialogue behavior. A character (e.g. extrovert or agreeable) can be defined to achieve human-like behaviors and user adaptation, and the appropriate character differs depending on each user. In this study, we investigate user adaptation by character expression of conversational robots. Our character expression model adopts the Big Five traits for controlling four dialogue behaviors: amount of utterance, backchannel frequency, filler frequency, and switching pause length. We cluster the system character and user personality into four classes based on an analysis of a speed dating dialogue corpus, and we found the best combinations between the system character and user personality. We implemented the character expression model into a laboratory guide and a chit-chat robot and conducted subjective experiments, where the subjects talked with robots with four different characters and evaluated their impressions of each character. The results shows that the Role model character was preferred for the task-oriented dialogue of the laboratory guide, but a robot character complementary to the subject personality was preferred for the non-task-oriented dialogue of the speed dating and chit-chat.

Yes(Rom.Da). Usages and functions in L2 proficiency examinations

AbstractThe article analyses a Romanian corpus of intercultural dialogues recorded in natural settings during language proficiency examinations and discusses the usages of the frequently occurring wordyes(Rom.da). The framework is provided by the Mixed Game Model (Weigand 2009,2010,2017,2018, among others) intersected with the socio-cognitive approach to intercultural communication (Kecskes 2007,2008,2010,2013, among others). Beyond the analysis of a single lexical unit in a limited collection of texts, the author aims to demonstrate how a particular word of the lexicon becomes a culturally-socially-cognitively-discursively-rhetorically situated utterance that acquires specific functions when it is used by interlocutors who negotiate their communicative agendas to come not only to mutual understanding but also to achieve their convergent-divergent institutional purposes.

Toward enriched decoding of mandarin spontaneous speech

A deep neural network (DNN)-based automatic speech recognition (ASR) method for enriched decoding of Mandarin spontaneous speech is proposed. It adopts an enhanced approach over the baseline model built with factored time delay neural networks (TDNN-f) and rescored with RNNLM to first building a baseline system composed of a TDNN-f acoustic model (AM), a trigram language model (LM), and a recurrent neural network language model (RNNLM) to generate a word lattice. It then sequentially incorporates a multi-task Part-of-Speech-RNNLM (POS-RNNLM), a hierarchical prosodic model (HPM), and a reduplication-word LM (RLM) into the decoding process by expanding the word lattice and performing rescoring to improve recognition performance and enrich the decoding output with syntactic parameters of POS and punctuation (PM), prosodic tags of word-juncture break types and syllable prosodic states, and an edited recognition text with reduplication words being eliminated. Experimental results on the Mandarin conversational dialogue corpus (MCDC) showed that SER, CER, and WER of 13.2 %, 13.9 %, and 19.1 % were achieved when incorporating the POS-RNNLM and HPM into the baseline system. They represented relative SER, CER, and WER reductions of 7.7 %, 7.9 % and 5.0 % as comparing with those of the baseline system. Futhermore, the use of the RLM resulted in additional 3 %, 4.6 %, and 4.5 % relative SER, CER, and WER reductions through eliminating reduplication words.

Multiheaded deep learning chatbot for increasing production and marketing

Some businesses on product development prefer to use a chatbot for judging the customer's view. Today, the ability of a chatbot to consider the context is challenging due to its technical nature. Sometimes, it may misjudge the context, making the wrong decision in predicting the product's originality in the market. This task of chatbot helps the enterprise make huge profits from accurate predictions. However, chatbots may commit errors in dialogs and bring inappropriate responses to users, reducing the confidentiality of product and marketing information. This, in turn, reduces the enterprise gain and imposes cost complications on businesses. To improve the performance of chatbots, AI models are used based on deep learning concepts. This research proposes a multi-headed deep neural network (MH-DNN) model for addressing the logical and fuzzy errors caused by retrieval chatbot models. This model cuts down on the error raised from the information loss. Our experiments extensively trained the model on a large Ubuntu dialog corpus. The recall evaluation scores showed that the MH-DNN approach slightly outperformed selected state-of-the-art retrieval-based chatbot approaches. The results obtained from the MHDNN augmentation approach were pretty impressive. In our proposed work, the MHDNN algorithm exhibited accuracy rates of 94% and 92%, respectively, with and without the help of the Seq2Seq technique.

Transition-Relevance Places Machine Learning-Based Detection in Dialogue Interactions

A transition-relevance place (TRP) represents a place in a conversation where a change of speaker can occur. The appearance and use of these points in the dialogue ensures a correct and smooth alternation between the speakers. In the presented article, we focused on the study of prosodic speech parameters in the Slovak language, and we tried to experimentally verify the potential of these parameters to detect TRP. To study turn-taking issues in dyadic conversations, the Slovak dialogue corpus was collected and annotated. TRP places were identified by the human annotator in the manual labelling process. The data were then divided into chunks that reflect the length of the interpausal dialogue units and the prosodic features were computed. In the Matlab environment, we compared different types of classifiers based on machine learning in the role of an automatic TRP detector based on pitch and intensity parameters. The achieved results indicate that prosodic parameters can be useful in detecting TRP after splitting the dialogue into interpausal units. The designed approach can serve as a tool for automatic conversational analysis or can be used to label large databases for training predictive models, which can help machines to enhance human-machine spoken dialogue applications.

Aspects of Dynamics in Dialogue Collaboration

Collaborative dialogue is an important category of human interaction and is widely studied in the literature, especially in fields that attempt to develop new technologies that enable wider varieties of collaborative dialogues. The ingredients of collaboration in dialogue are less thoroughly addressed. We describe the theoretical framework within which we are working and our approach to the construction of a theory of what may make dialogue collaborative. We study a multimodal dialogue corpus (MULTISIMO) testing for positive and negative correlations between dialogue content features and interaction features that one might reasonably imagine are related to assessments of degrees of collaboration. The duration before the second speaker’s first turn and degree of imbalance in the number of words produced by speakers negatively correlate with collaboration assessments (that is, imbalances of content and a delay in the first speaker yielding the floor lead to diminished perceptions of collaboration), while a monotonically increasing duration of focus in successive dialogue sections (rather than overall dialogue duration) correlates positively (that is, when participants are deemed to be extending the duration of the task rather than increasing speed with experience, this is perceived as collaborative).

Gender bias in video game dialogue.

Gender biases in fictional dialogue are well documented in many media. In film, television and books, female characters tend to talk less than male characters, talk to each other less than male characters talk to each other, and have a more limited range of things to say. Identifying these biases is an important step towards addressing them. However, there is a lack of solid data for video games, now one of the major mass media which has the ability to shape conceptions of gender and gender roles. We present the Video Game Dialogue Corpus, the first large-scale, consistently coded corpus of video game dialogue, which makes it possible for the first time to measure and monitor gender representation in video game dialogue. It demonstrates that there is half as much dialogue from female characters as from male characters. Some of this is due to a lack of female characters, but there are also biases in who female characters speak to, and what they say. We make suggestions for how game developers can avoid these biases to make more inclusive games.

Quirky conversations: how people with a diagnosis of schizophrenia do dialogue differently.

People with a diagnosis of schizophrenia (PSz) have difficulty engaging in social interaction, but little research has focused on dialogues involving PSz interacting with partners who are unaware of their diagnosis. Using quantitative and qualitative methods on a unique corpus of triadic dialogues of PSz first social encounters, we show that turn-taking is disrupted in dialogues involving a PSz. Specifically, there are on average longer gaps between turns in groups which contain a PSz compared to those which do not, particularly when the speaker switch occurs from one control (C) participant to the other. Furthermore, the expected link between gesture and repair is not present in dialogues with a PSz, particularly for C participants interacting with a PSz. As well as offering some insights into how the presence of a PSz affects an interaction, our results also demonstrate the flexibility of our mechanisms for interaction. This article is part of a discussion meeting issue 'Face2face: advancing the science of social interaction'.

SAM: Multi-turn Response Selection Based on Semantic Awareness Matching

Multi-turn response selection is a key issue in retrieval-based chatbots and has attracted considerable attention in the NLP (Natural Language processing) field. So far, researchers have developed many solutions that can select appropriate responses for multi-turn conversations. However, these works are still suffering from the semantic mismatch problem when responses and context share similar words with different meanings. In this article, we propose a novel chatbot model based on Semantic Awareness Matching, called SAM. SAM can capture both similarity and semantic features in the context by a two-layer matching network. Appropriate responses are selected according to the matching probability made through the aggregation of the two feature types. In the evaluation, we pick 4 widely used datasets and compare SAM’s performance to that of 12 other models. Experiment results show that SAM achieves substantial improvements, with up to 1.5% R 10 @1 on Ubuntu Dialogue Corpus V2, 0.5% R 10 @1 on Douban Conversation Corpus, and 1.3% R 10 @1 on E-commerce Corpus.

Defining Filler Particles: A Phonetic Account of the Terminology, Form, and Grammatical Classification of “Filled Pauses”

The terms hesitation, planner, filler, and filled pause do not always refer to the same phonetic entities. This terminological conundrum is approached by investigating the observational, explanatory, and descriptive inadequacies of the terms in use. Concomitantly, the term filler particle is motivated and a definition is proposed that identifies its phonetic exponents and describes them within the linguistic category of particles. The definition of filler particles proposed here is grounded both theoretically and empirically and then applied to a corpus of spontaneous dialogues with 32 speakers of German, showing that in addition to the prototypical phonetic forms, there is a substantial amount of non-prototypical forms, i.e., 9.5%, comprising both glottal (e.g., [Ɂ]) and vocal forms (e.g., [ɛɸ], [j~ɛvə]). The grammatical classification and the results regarding the phonetic forms are discussed with respect to their theoretical relevance in filler particle research and corpus studies. The phonetic approach taken here further suggests a continuum of phonetic forms of filler particles, ranging from singleton segments to multi-syllabic entities.

한국어의 우호적 담화 형성의 언어적 기제에 대한 코퍼스기반 연구

The purpose of this study is to investigate the linguistic mechanisms for the formation of amicable discourses in random oral dialogue corpus in Korean. The four categories of linguistic mechanisms are explored in this paper: (i) interrogative mood markers and honorific Si- markers, (ii) adverbial and predicative modal hedges, (iii) first person pronoun markers Ce(I) and Wuri(we), and (iv) positive response markers Ye(yes) and Ney(yes). The following claims are made in this paper. First, the interrogative mood markers and honorific markers are used for the amicable discourse with the percentage of 0.8% and 0.9%, respectively. Second, the adverbial and predicative modal hedges are most frequently used with the percentage of 5.9%. Third, the 1SUPst/SUP person pronominal markers are used with the percentage of 1.5%. Fourth, the positive response markers are least frequently used mechanism with the percentage of 0.2%. Hence the linguistic mechanisms for amicable discourse are ranked as follows; adverbial and predicative modal hedges the 1SUPst/SUP person pronominal markers interrogative mood markers and honorific Si- markers the positive response markers.