Learn More
We are applying a memory based learning (MBL) algorithm to the task of automatic dialog act (DA) tagging. This work is along the lines of a recent trend that considers MBL as being more appropriate for natural language processing. We did the experiments on the Switchboard corpus, overcome the problem of feature selection and yield results that seem to be(More)
This paper presents a deep architecture for learning a similarity metric on variablelength character sequences. The model combines a stack of character-level bidirectional LSTM’s with a Siamese architecture. It learns to project variablelength strings into a fixed-dimensional embedding space by using only information about the similarity between pairs of(More)
We study dependencies between discourse structure and speech recognition problems (SRP) in a corpus of speech-based computer tutoring dialogues. This analysis can inform us whether there are places in the discourse structure prone to more SRP. We automatically extract the discourse structure by taking advantage of how the tutoring information is encoded in(More)
Many practical information extraction systems use simple taxonomies for mapping extracted strings to client-specific concept codes. In such taxonomies, concepts are defined as groups of semantically similar words and phrases. For the mapping to be accurate, a new client-specific taxonomy – usually nothing more than a set of concept codes, each with a single(More)
In this paper, we advocate for the usage of word-level pitch features for detecting user emotional states during spoken tutoring dialogues. Prior research has primarily focused on the use of turn-level features as predictors. We compute pitch features at the word level and resolve the problem of combining multiple features per turn using a word-level(More)
At the end of twentieth century, huge data sets and large distributed environments posed a new challenge for machine learning algorithms. To describe the structure of data more compactly, people started to seek for meaningful components that would characterize the data in a significantly lower dimension. Probably the best known method for finding(More)
  • 1