| File | Description | Format |
|---|---|---|
| dial_utter_resp_train.npz | Training utterances & responses. | An npz object with 3 variables. - var1: utterance is a numpy array of size #user_item_pairs x #utters x utter_len. - var2: response is a numpy array of size (#user_item_pairs x 2) x utter_len. - var3: label is a numpy array of size (#user_item_pairs x 2). Label 1 indicates correct response while 0 is wrong response. |
| dial_utter_resp_val.npz | Validation utterances & responses. | An npz object with 3 variables. - var1: utterance is a numpy array of size #user_item_pairs x #utters x utter_len. - var2: response is a numpy array of size (#user_item_pairs x 10) x utter_len. - var3: label is a numpy array of size (#user_item_pairs x 10). Label 1 indicates correct response while 0 is wrong response. |
| dial_utter_resp_test.npz | Test utterances & responses. | An npz object with 3 variables. - var1: utterance is a numpy array of size #user_item_pairs x #utters x utter_len. - var2: response is a numpy array of size (#user_item_pairs x 10) x utter_len. - var3: label is a numpy array of size (#user_item_pairs x 10). Label 1 indicates correct response while 0 is wrong response. |
| dial_word_embed.npz | Word embeddings for dialogs. | An npz object with 1 variables. - var1: word_emb is a numpy array of size #(num of words + 1) x #200, the last word index is padding index. |
| dial_gt_context_train.npz | Groundtruth entities and KG context triples for training utterances. | An npz object with 4 variables. - var1: utter_gt is a numpy array of size #user_item_pairs x #utters x 1. - var2: utter_context is a numpy array of size #user_item_pairs x #utters x (#triples x 3). - var3: resp_gt is a numpy array of size (#user_item_pairs x 2). - var4: resp_context is a numpy array of size (#user_item_pairs x 2) x (#triples x 3). |
| dial_gt_context_val.npz | Groundtruth entities and KG context triples for validation utterances. | An npz object with 4 variables. - var1: utter_gt is a numpy array of size #user_item_pairs x #utters x 1. - var2: utter_context is a numpy array of size #user_item_pairs x #utters x (#triples x 3). - var3: resp_gt is a numpy array of size (#user_item_pairs x 10). - var4: resp_context is a numpy array of size (#user_item_pairs x 10) x (#triples x 3). |
| dial_gt_context_test.npz | Groundtruth entities and KG context triples for test utterances. | An npz object with 4 variables. - var1: utter_gt is a numpy array of size #user_item_pairs x #utters x 1. - var2: utter_context is a numpy array of size #user_item_pairs x #utters x (#triples x 3). - var3: resp_gt is a numpy array of size (#user_item_pairs x 10). - var4: resp_context is a numpy array of size (#user_item_pairs x 10) x (#triples x 3). |
Notes:
#utters=10,utter_len=50,#triples=8.#user_item_pairsdepends on train/validation/test.
| File | Description | Format |
|---|---|---|
| rec_train.txt | Training user-item pairs used as purchase history. | Each row is a user-item pair in the form of [user_id]\t[item_id]. The i-th row corresponds to the i-th dialog in the "dial_utter_resp_train.npz" file. |
| rec_val_candidate100.npz | Validation user-item pairs including 101 candidates. | An npz object with 1 variable. - var1: candidates is a numpy array of size #user_item_pairs x 102, where the first column is user id, second column is groundtruth item id and the rest 100 columns are negative item ids. |
| rec_test_candidate100.npz | Test user-item pairs including 101 candidates. | An npz object with 1 variable. - var1: candidates is a numpy array of size #user_item_pairs x 102, where the first column is user id, second column is groundtruth item id and the rest 100 columns are negative item ids. |
Notes:
[user_id]and[item_id]refer to the entity id of users and items in KG.
| File | Description | Format |
|---|---|---|
| kg_user_entities.txt | User entities. | Each row is user::[username]\t[entity_id]. |
| kg_item_entities.txt | Item entities. | Each row is product::[item_name]\t[entity_id]. |
| kg_other_entities.txt | Other entities. | Each row is [entity_type]::[entity_name]\t[entity_id]. |
| kg_relations.txt | Relations (no inverse relation included). | Each row is [relation_name]\t[relation_id]. |
| kg_train_triples.txt | Training triples regarding user-item interactions. | Each row is [head_entity_id]\t[tail_entity_id]\t[relation_id], where the [relation_id] is always 0 meaning the user-item interaction. |
| kg_other_triples.txt | Other triples excluding user-item interactions. | Each row is [head_entity_id]\t[tail_entity_id]\t[relation_id]. |
Note: user entity ids, item entity ids and other entity ids are consecutive. E.g., suppose there are 100 users, 200 items and 500 other entities. The user ids range in [0, 99], item ids range in [100, 299] and other entity ids range in [300, 799].
Follow the process here.
Comming Soon!
Comming Soon!
[1] Liu Yang, Minghui Qiu, Chen Qu, Jiafeng Guo, Yongfeng Zhang, W. Bruce Croft, Jun Huang, Haiqing Chen. "Response Ranking with Deep Matching Networks and External Knowledge in Information-seeking Conversation Systems", In Proceedings of SIGIR. 2018.
[2] Seungwhan Moon, Pararth Shah, Anuj Kumar, Rajen Subba. "OpenDialKG: Explainable Conversational Reasoning with Attention-based Walks over Knowledge Graphs", In Proceeding of ACL. 2019.
[3] Qibin Chen, Junyang Lin, Yichang Zhang, Ming Ding, Yukuo Cen, Hongxia Yang, Jie Tang. "Towards Knowledge-Based Recommender Dialog System", In Proceeding of EMNLP. 2019.