Dilek Z. Hakkani-Tur

End-to-end joint learning of natural language understanding and dialogue manager

Natural language understanding and dialogue policy learning are both essential in conversational systems that predict the next system actions in response to a current user utterance. Conventional approaches aggregate separate models of natural language understanding (NLU) and system action prediction (SAP) as a pipeline that is sensitive to noisy outputs of error-prone NLU. To address the issues, we propose an end-to-end deep recurrent neural network with limited contextual dialogue memory by jointly training NLU and SAP on DSTC4 multi-domain human-human dialogues. Experiments show that our proposed model significantly outperforms the state-of-the-art pipeline models for both NLU and SAP, which indicates that our joint model is capable of mitigating the affects of noisy NLU outputs, and NLU model can be refined by error flows backpropagating from the extra supervised signals of system actions.

Spoken language understanding and interaction: machine learning for human-like conversational systems

Abstract In recent years, the interest in research in speech understanding and spoken interaction has soared due to the emergence of virtual personal assistants. However, while the ability of these agents to recognise conversational speech is maturing rapidly, their ability to understand and interact is still limited. At the same time we have witnessed the development of the number of models based on machine learning that made a huge impact on spoken language understanding accuracies and the interaction quality overall. This special issue brings together a number of articles that tackle different aspects of spoken language understanding and interaction: clarifications in dialogues, adaptation to different domains, semantic tagging and error handling. These studies all have a common purpose of building human-like conversational systems.

Learning concepts through conversations in spoken dialogue systems

Spoken dialogue systems must be able to recover gracefully from unexpected user inputs. In many cases, these unexpected utterances may be within the scope of the system, but include previously unseen phrases that the system cannot interpret. In this work, we augment a spoken dialogue system with the ability to learn about new concepts by conversing with the user in natural language. We present a novel model that detects phrases corresponding to such concepts, using information from a neural slotfiller as well as syntactic cues. The system then prompts the user for a definition of the detected phrases, and uses these definitions to re-parse the original utterance. We demonstrate significant gains by learning from the user, compared to a baseline system.

Deep Learning in Conversational Language Understanding

Recent advancements in AI resulted in increased availability of conversational assistants that can help with tasks such as seeking times to schedule an event and creating a calendar entry at that time, finding a restaurant and booking a table there at a certain time. However, creating automated agents with human-level intelligence still remains one of the most challenging problems of AI. One key component of such systems is conversational language understanding, which is a holy grail area of research for decades, as it is not a clearly defined task but relies heavily on the AI application it is used for. Nevertheless, this chapter attempts to compile the recent deep learning based literature on such goal-oriented conversational language understanding studies, starting with a historical perspective, pre-deep learning era work, moving toward most recent advances in this field.

Morphological Disambiguation for Turkish

Morphological disambiguation is the task of determining the contextually correct morphological parses of tokens in a sentence. A morphological disambiguator takes in a set of morphological parses for each token, generated by a morphological analyzer, and then selects a morphological parse for each, considering statistical and/or linguistic contextual information. This task can be seen as a generalization of the part-of-speech (POS) tagging problem, for morphologically rich languages. The disambiguated morphological analysis is usually crucial for further processing steps such as dependency parsing. In this chapter, we review the morphological disambiguation problem for Turkish and discuss approaches for solving this problem as they have evolved from manually crafted constraint-based rule systems to systems employing machine learning.

Deep Learning in Spoken and Text-Based Dialog Systems

Last few decades have witnessed substantial breakthroughs on several areas of speech and language understanding research, specifically for building human to machine conversational dialog systems. Dialog systems, also known as interactive conversational agents, virtual agents or sometimes chatbots, are useful in a wide range of applications ranging from technical support services to language learning tools and entertainment. Recent success in deep neural networks has spurred the research in building data-driven dialog models. In this chapter, we present state-of-the-art neural network architectures and details on each of the components of building a successful dialog system using deep learning. Task-oriented dialog systems would be the focus of this chapter, and later different networks are provided for building open-ended non-task-oriented dialog systems. Furthermore, to facilitate research in this area, we have a survey of publicly available datasets and software tools suitable for data-driven learning of dialog systems. Finally, appropriate choice of evaluation metrics are discussed for the learning objective.