Prediction of Nash Bargaining Solution in Negotiation Dialogue
| Autor: | Katsuhide Fujita, Kosui Iwasa |
|---|---|
| Rok vydání: | 2018 |
| Předmět: |
Bargaining problem
business.industry Computer science media_common.quotation_subject 02 engineering and technology Negotiation Recurrent neural network 020204 information systems 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing Support system Artificial intelligence Function (engineering) business Set (psychology) Natural language media_common |
| Zdroj: | Lecture Notes in Computer Science ISBN: 9783319973036 PRICAI (1) |
| Popis: | There are several studies that focus on the support systems that can be used for human-human negotiations. However, existing automated agent-agent negotiation systems require the participants to set utility functions manually. Additionally, a method to predict the utility functions from the negotiation dialogues in natural language and to find a bargaining solution has not been proposed yet. By developing such a method, the existing research related to automated negotiations can be utilized for the negotiation dialogues in real-life situations. Therefore, we propose a method to predict the utility function of each agent and Nash bargaining solution only from the negotiation dialogues using gated recurrent units (GRUs) [4] with attention [3]. We demonstrate that the rate of Nash bargaining solution that was obtained by using our method outperforms the rate that was obtained while humans were negotiating. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|