Code to combine conversational agent and recommender system together as defined in this paper.
Conversational agents are used widely in e-commerce websites to interact with customers and help them solve questions and find products. Recommendation is also an essential feature of e-commerce websites, which help customers discover more interesting products. In this paper, we implement a Conversational Restaurant Recommender System(CRRS), which combines the conversational agent and recommender system together. We also evaluate the performance of CRRS by a laboratory experiment. Our result shows that CRRS is an efficient tool, which always recommends suitable restaurants based on our users’ needs.
- python3.7
- torch1.2
- numpy
- tqdm
- sklearn
- pickle
python example.py 0 0 --num 3
First parameter is user type. 0: simulated users, 1:real users.
Second parameter is agent type. 0: rule-based method, 1: RL-based method.
Third parameter (optional) is the number of conversation round, default is 1.
SampleConversations.pdf is the log of sample conversations.
When user type is real users,
- All the conversations are through terminal commands, and users have to restrictly type the correct answer.
- target restaurant is randomly chosen by the system and revealed to users;
- target restaurant information is provided;
- users have to answer questions from the system based on these information.