Dialogue Act Sequence Labeling Using Hierarchical Encoder With CRF
Authors: Harshit Kumar, Arvind Agarwal, Riddhiman Dasgupta, Sachindra Joshi
AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We validate our approach on two different benchmark data sets, Switchboard and Meeting Recorder Dialogue Act, and show performance improvement over the state-of-the-art methods by 2.2% and 4.1% absolute points, respectively. |
| Researcher Affiliation | Industry | Harshit Kumar, Arvind Agarwal, Riddhiman Dasgupta, Sachindra Joshi IBM Research, India {harshitk, arvagarw, riddasgu, jsachind}@in.ibm.com |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide concrete access to source code for the methodology described. |
| Open Datasets | Yes | We evaluate the performance of our model on two benchmark datasets used in several prior studies for the DA identification task, viz.: Sw DA: Switchboard Dialogue Act Corpus (Jurafsky 1997) MRDA: The ICSI Meeting Recorder Dialogue Act corpus (Janin et al. 2003; Ang, Liu, and Shriberg 2005) |
| Dataset Splits | Yes | Table 2 presents different statistics for both datasets. For Sw DA, train and test sets are provided but not the validation set, so we use the standard practice of taking a part of training data set as validation set (Lee and Dernoncourt 2016). MRDA 5 10K 51(76K) 11(15K) 11(15K) Sw DA 42 19K 1003(173K) 112(22K) 19(4K) |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running its experiments. |
| Software Dependencies | No | The paper mentions software like 'Adadelta optimizer' and 'Glove embeddings' but does not provide specific version numbers for these or any other software dependencies needed to replicate the experiment. |
| Experiment Setup | Yes | Table 3 lists the range of values for each parameter that we experimented with, and the final value that was selected. The maximum batch-size allowed was 64. We used L2 regularization of 1e 4 in the form of weight decay and the Adadelta optimizer. The word vectors were initialized with the 300-dimensional Glove embeddings (Pennington, Socher, and Manning 2014), and were also updated during training. Dropout was applied to the embeddings obtained from the output of each encoder. The learning rate was initialized to 1.0 and reduced by a factor of 0.5 every 5 epochs. Early stopping is also used on the validation set with a patience of 5 epochs. (Table 3 includes: Pooling: Last, Word Embedding: 300D Glove, Dropout: 0.2, Bidirectional: True, Hidden Size: 300, Learning Rate: 1.0, LSTM Layers: 1) |