Search

Reinforcement learning has been implemented to model a task by doing the task repeatedly to get the maximum results based on the reward and punishment policy. It has been implemented in the game and agent-based modelling. In the game, the game agent or Non-Players character can be modelled using several techniques to achieve the goal (e.g. reinforcement learning, deep neural network and Monte Carlo Tree Search). Deep neural networks and Monte Carlo Tree Search, two more sophisticated techniques in reinforcement learning algorithms, assisted the present reinforcement learning in resolving more challenging issues. However, this area has two challenges: the minimum number of data to model and generalization to different environments. Determining a minimum number of data required by the architecture to train the model is quite a cumbersome task to be applied to real- world jobs and situations since it demands substantial data to be explicitly provided and trial-and-error re-configuration. This work proposes a Data-Efficient Reinforcement Learning model by augmenting the data and implementing episodic memory. To illustrate the effectiveness of the proposed model, this research compares it to several models, such as the Deep Q-Network (DQN) with episodic memory to the same model with data augmentation and episodic memory. The model adds to the observations before being stored in the agent's memory, causing the agent to use the same logic and take the same action in comparable situations. The outcome demonstrates that the augmented model can surpass the fundamental model in speed (with an improvement of 50% quicker). [ABSTRACT FROM AUTHOR]

Recognising emotions from a conversation is a paramount task for a machine to understand the full context of the conversation. However, recognising emotions is not an easy task for a machine that cannot understand social context. This research aims to explore several conventional machine learning methods in emotion recognition modelling, such as Na¨ıve Bayes, K-Nearest Neigh- bour, Decision Trees and Random Forest. Moreover, this research also presents the exploration of deep learning architecture such as Vanilla Deep Neural Networks, Convolutional Neural Networks and Long-Short Term Memory. Moreover, five Mel-Frequency Cepstral Coefficients numbers are also explored (32, 40, 48, 64, 128) with two sampling techniques (Fast Fourier Transform and Kaiser Fast). The results demonstrate that the best training accuracy is achieved by several algorithms such as KNN, ANN and LSTM (100%). Moreover, the model trained with ANN (73%) achieves the best training accuracy. [ABSTRACT FROM AUTHOR]

Emotions hold a paramount role in the conversation, as it expresses context to the conversation. Text/word in conversation consists of lexical and contextual meanings. Extracting emotions from text has been an interesting work recent thees years. With the advancement of machine learning techniques and hardware to support the machine learning process, recognising emotions from a text with machine learning provides promising and significant results. This research aims to explore several popular machine learning to recognise emotions from a conversation in social media. The algorithms proposed in this research are ranged from traditional machine learning to deep learning techniques. The dataset used in this paper is provided by AffectiveTweets, with a baseline of F1Score of 0.71 with word N-grams and SentiStrength. The research contributes extensive explorations in a number of machine learning algorithms, resulting in a total of 2302 features sets were explored, where each features sets has 100-1000 features extracted from the text. The results demonstrate Generalised Linear Model provides the best Accuracy score (0.92), Recall (0.902), Precision (0.902), F1 score (0.901) with standard deviation of accuracy of ±1,2%. [ABSTRACT FROM AUTHOR]

This research contributes to data collection for Indonesian Language conversation corpus or dataset collected from conversation in the movies through 1961 subtitles with 1678320 unique words. The dataset collected thus trained with deep learning algorithm, a dual encoder Recurrent Neural Network (RNN), Long-Short Term Memory (LSTM) with Attention mechanism with the best result was 2.37% unknown words, loss rate was 1.66, and perplexity of 4.96, trained with vocabulary size of 24000. The dataset could be used for many machine (deep) learning for Natural Language Processing Problem in Indonesian Language, while the pre-trained model can be integrated into several system such as agent-based system, be they a chat-bot or an Embodied Conversational Agent (ECA). For future work, more data will be collected not only from the movie conversation but also natural human-human conversation in Indonesian Language. [ABSTRACT FROM AUTHOR]

Computer has been assisting human in almost all aspects of daily life. Despite the computer outsmart human in some tasks, it is a social ignorant tool. It does not understand and capable of doing natural conversation with human. To build a system that can naturally understand and communicate with human, it is essential to train the system with natural conversation data. This paper proposes a dataset which consists of natural dyadic conversation in Indonesian language. Where literature suggested that there are exists inadequate number of conversation dataset in Indonesian Language. There are 3164 words (formal and slang (informal nonstandard words) annotated from the recording of five groups, with the largest number of words belong to Food topic (Group 3, 826 words), and the lowest one belongs to Travelling topic (Group 2, 372 words). The dataset contributes a pre-trained conversation model with deep learning (LSTM). The model is trained in 10000 iterations, 128 batches, and 4 hidden layers, resulted in a perplexity of 2.01. [ABSTRACT FROM AUTHOR]