Filip Jurčíček

Spoken language understanding from unaligned data using discriminative classification models

While data-driven methods for spoken language understanding reduce maintenance and portability costs compared with handcrafted parsers, the collection of word-level semantic annotations for training remains a time-consuming task. A recent line of research has focused on building generative models from unaligned semantic representations, using expectation-maximisation techniques to align semantic concepts. This paper presents an efficient, simple technique that parses a semantic tree by recursively calling discriminative semantic classification models. Results show that it outperforms methods based on the Hidden Vector State model and Markov Logic Networks, while performance is close to more complex grammar induction techniques. We also show that our method is robust to speech recognition errors, by improving over a handcrafted parser previously used for dialogue data collection.

On-line policy optimisation of spoken dialogue systems via live interaction with human subjects

Statistical dialogue models have required a large number of dialogues to optimise the dialogue policy, relying on the use of a simulated user. This results in a mismatch between training and live conditions, and significant development costs for the simulator thereby mitigating many of the claimed benefits of such models. Recent work on Gaussian process reinforcement learning, has shown that learning can be substantially accelerated. This paper reports on an experiment to learn a policy for a real-world task directly from human interaction using rewards provided by users. It shows that a usable policy can be learnt in just a few hundred dialogues without needing a user simulator and, using a learning strategy that reduces the risk of taking bad actions. The paper also investigates adaptation behaviour when the system continues learning for several thousand dialogues and highlights the need for robustness to noisy rewards.

Reinforcement learning for parameter estimation in statistical spoken dialogue systems

Reinforcement techniques have been successfully used to maximise the expected cumulative reward of statistical dialogue systems. Typically, reinforcement learning is used to estimate the parameters of a dialogue policy which selects the systems responses based on the inferred dialogue state. However, the inference of the dialogue state itself depends on a dialogue model which describes the expected behaviour of a user when interacting with the system. Ideally the parameters of this dialogue model should be also optimised to maximise the expected cumulative reward. This article presents two novel reinforcement algorithms for learning the parameters of a dialogue model. First, the Natural Belief Critic algorithm is designed to optimise the model parameters while the policy is kept fixed. This algorithm is suitable, for example, in systems using a handcrafted policy, perhaps prescribed by other design considerations. Second, the Natural Actor and Belief Critic algorithm jointly optimises both the model and the policy parameters. The algorithms are evaluated on a statistical dialogue system modelled as a Partially Observable Markov Decision Process in a tourist information domain. The evaluation is performed with a user simulator and with real users. The experiments indicate that model parameters estimated to maximise the expected reward function provide improved performance compared to the baseline handcrafted parameters.

Natural actor and belief critic: Reinforcement algorithm for learning parameters of dialogue systems modelled as POMDPs

This article presents a novel algorithm for learning parameters in statistical dialogue systems which are modeled as Partially Observable Markov Decision Processes (POMDPs). The three main components of a POMDP dialogue manager are a dialogue model representing dialogue state information; a policy that selects the systems responses based on the inferred state; and a reward function that specifies the desired behavior of the system. Ideally both the model parameters and the policy would be designed to maximize the cumulative reward. However, while there are many techniques available for learning the optimal policy, no good ways of learning the optimal model parameters that scale to real-world dialogue systems have been found yet. The presented algorithm, called the Natural Actor and Belief Critic (NABC), is a policy gradient method that offers a solution to this problem. Based on observed rewards, the algorithm estimates the natural gradient of the expected cumulative reward. The resulting gradient is then used to adapt both the prior distribution of the dialogue model parameters and the policy parameters. In addition, the article presents a variant of the NABC algorithm, called the Natural Belief Critic (NBC), which assumes that the policy is fixed and only the model parameters need to be estimated. The algorithms are evaluated on a spoken dialogue system in the tourist information domain. The experiments show that model parameters estimated to maximize the expected cumulative reward result in significantly improved performance compared to the baseline hand-crafted model parameters. The algorithms are also compared to optimization techniques using plain gradients and state-of-the-art random search algorithms. In all cases, the algorithms based on the natural gradient work significantly better.

Parameter learning for POMDP spoken dialogue models

The partially observable Markov decision process (POMDP) provides a popular framework for modelling spoken dialogue. This paper describes how the expectation propagation algorithm (EP) can be used to learn the parameters of the POMDP user model. Various special probability factors applicable to this task are presented, which allow the parameters be to learned when the structure of the dialogue is complex. No annotations, neither the true dialogue state nor the true semantics of user utterances, are required. Parameters optimised using the proposed techniques are shown to improve the performance of both offline transcription experiments as well as simulated dialogue management performance.

Sequence-to-Sequence Generation for Spoken Dialogue via Deep Syntax Trees and Strings.

We present a natural language generator based on the sequence-to-sequence approach that can be trained to produce natural language strings as well as deep syntax dependency trees from input dialogue acts, and we use it to directly compare two-step generation with separate sentence planning and surface realization stages to a joint, one-step approach. We were able to train both setups successfully using very little training data. The joint setup offers better performance, surpassing state-of-the-art with regards to n-gram-based scores while providing more relevant outputs.

Bayesian dialogue system for the Let's Go Spoken Dialogue Challenge

This paper describes how Bayesian updates of dialogue state can be used to build a bus information spoken dialogue system. The resulting system was deployed as part of the 2010 Spoken Dialogue Challenge. The purpose of this paper is to describe the system, and provide both simulated and human evaluations of its performance. In control tests by human users, the success rate of the system was 24.5% higher than the baseline Lets Go! system.

Back-off action selection in summary space-based POMDP dialogue systems

This paper deals with the issue of invalid state-action pairs in the Partially Observable Markov Decision Process (POMDP) framework, with a focus on real-world tasks where the need for approximate solutions exacerbates this problem. In particular, when modelling dialogue as a POMDP, both the state and the action space must be reduced to smaller scale summary spaces in order to make learning tractable. However, since not all actions are valid in all states, the action proposed by the policy in summary space sometimes leads to an invalid action when mapped back to master space. Some form of back-off scheme must then be used to generate an alternative action. This paper demonstrates how the value function derived during reinforcement learning can be used to order back-off actions in an N-best list. Compared to a simple baseline back-off strategy and to a strategy that extends the summary space to minimise the occurrence of invalid actions, the proposed N-best action selection scheme is shown to be significantly more robust.

Training a Natural Language Generator From Unaligned Data

We present a novel syntax-based natural language generation system that is trainable from unaligned pairs of input meaning representations and output sentences. It is divided into sentence planning, which incrementally builds deep-syntactic dependency trees, and surface realization. Sentence planner is based on A* search with a perceptron ranker that uses novel differing subtree updates and a simple future promise estimation; surface realization uses a rule-based pipeline from the Treex NLP toolkit. Our first results show that training from unaligned data is feasible, the outputs of our generator are mostly fluent and relevant.

Incremental LSTM-based dialog state tracker

A dialog state tracker is an important component in modern spoken dialog systems. We present an incremental dialog state tracker, based on LSTM networks. It directly uses automatic speech recognition hypotheses to track the state. We also present the key non-standard aspects of the model that bring its performance close to the state-of-the-art and experimentally analyze their contribution: including the ASR confidence scores, abstracting scarcely represented values, including transcriptions in the training data, and model averaging.