Analysis system of speech acts and discourse structures using maximum entropy model

Abstract

We propose a statistical dialogue analysis model to determine discourse structures as well as speech acts using maximum entropy model. The model can automatically acquire probabilistic discourse knowledge from a discourse tagged corpus to resolve ambiguities. We propose the idea of tagging discourse segment boundaries to represent the structural information of discourse. Using this representation we can effectively combine speech act analysis and discourse structure analysis in one framework.

Improving Korean Speech Acts Analysis by Using Shrinkage and Discourse Stack

“…5. The comparison of the performances for the shrinkage technique according to the distribution of speech acts Table 7 shows results from the proposed model and previous speech acts analysis models: the maximum entropy model (MEM) [1], the decision tree model (DTM) [8], and the neural network model (NNM) [5]. We report the performance of each system when using the same test data set as that of this paper.…”

“…Since discourse structure information represents the relationship between two consecutive utterances, it is efficient to use discourse structure Use speech acts that pop in discourse stack and Sub-dialogue End (SE) else Use speech acts of previous utterance and Dialogue Continue (DC) End information for speech acts analysis [1]. Especially, the speech act of seventh utterance in Table 1 (UID: 7) is tied with that of second utterance (UID: 2).…”

“…We used the Korean dialogue corpus which has used in previous research [1][5] [8]. This corpus was transcribed from recordings in real fields such as hotel reservation, airline reservation and tour reservation and consists of 528 dialogues, 10,285 utterances (19.48 utterances per dialogue).…”

“…In addition, discourse structure information can be used to identify the speech acts of utterances [1]. But most previous research has used only speech acts of previous utterances without considering discourse structure information to determine the speech act of current utterance.…”

“…Various machine learning models have been used to efficiently classify speech acts such as MEM (Maximum Entropy Model) [1], HMM (Hidden Markov Model) with Decision Tree [8] [11], Neural Network Model [5]. And there are also studies on methods of automatically selecting efficient features with useful information for speech acts analysis [5] [10].…”