Semi-Automated Dialogue Act Classification for Situated Social Agents in Games

Abstract

Abstract. As a step toward simulating dynamic dialogue between agents and humans in virtual environments, we describe learning a model of social behavior composed of interleaved utterances and physical actions. In our model, utterances are abstracted as {speech act, propositional content, referent} triples. After training a classifier on 100 gameplay logs from The Restaurant Game annotated with dialogue act triples, we have automatically classified utterances in an additional 5,000 logs. A quantitative evaluation of statistical models learned from the gameplay logs demonstrates that semi-automatically classified dialogue acts yield significantly more predictive power than automatically clustered utterances, and serve as a better common currency for modeling interleaved actions and utterances.

Computational Psychometrics for the Measurement of Collaborative Problem Solving Skills

“…Based on this and other related research (Kerr et al, 2011; Orkin and Roy, 2011; Smith, 2011; Canossa, 2013; Li et al, 2013), it was evident that a machine learning-based, clustering methodology would be useful to explore patterns within our game dialog selection data. In particular we demonstrate an application of game-related, k-means clustering [as reported in other related research (Thurau and Bauckhage, 2010)] vs. other types reported such as Linear Discriminant Analysis (LDA) (Gow et al, 2012) or Mixture Model clustering (Teófilo and Reis, 2013).…”

Automatic dialogue act recognition with syntactic features

“…Hence, the authors of [25] propose a predictive paradigm where dialogue act models are first trained on a small-size corpus and used afterwards to predict future sentences or dialogue acts. In a related vein, unsupervised dialogue act tagging of unlabelled text has recently raised a lot of attention [26,27], but we will limit ourselves in the following on supervised approaches.…”

Using a Slim Function Word Classifier to Recognise Instruction Dialogue Acts

“…The most commonly used feature is the n-gram, a short sequence of contiguous words. N-grams are created by extracting them from a real-world corpus [8,9]. Any realistically sized corpus will generate large numbers of candidates.…”

“…Consequently the set of potential n-grams is analysed to find a subset which are good predictors for the classification task. For example, [9] used a set of 100 logs from a social interaction computer game containing 4,295 utterances with an average length of 4 words as a training set, using a mutual information measure to find the best 50 predictive unigrams, bigrams and trigrams for 9 distinct DAs.…”