Papers with corpus-based selection of dialogue responses
Which Model Should We Use for a Real-World Conversational Dialogue System? a Cross-Language Relevance Model or a Deep Neural Net? (2020.lrec-1)
Copied to clipboard
| Challenge: | End-to-end neural network models of conversational dialogue are popular for conversational tasks, but there are still questions about how well they work for real applications and how much data is needed to achieve acceptable performance. |
| Approach: | They compare two different kinds of end-to-end dialogue models based on cross-language relevance and cross-linguistic LSTM models for corpus-based selection of dialogue responses. |
| Outcome: | The proposed models perform well on a large corpus, but are dominated by a more moderate-sized corpus. |