GithubHelp home page GithubHelp logo

u-need's Introduction

U-NEED介绍

我们收集了一个以用户需求为中心的电商对话式推荐数据集(U-NEED)。

We collect a user needs-centric E-commerce conversational recommendation dataset (U-NEED).

U-NEED包含了7,698个细粒度标注的售前对话,333,879个用户行为和332,148条商品知识元组。

U-NEED consists of 7,698 fine-grained annotated pre-sales dialogues, 333,879 user behaviors and 332,148 product knowledge tuples.

对于售前对话的每一条语句,我们雇佣了专业的众包平台来标注:说话人的动作,语句涉及的属性和语句中推荐的商品。

For each utterance of pre-sales dialogue, we hire a professional crowdsourcing platform to annotate the action of the speaker, the attributes involved, and the recommended products.

基线模型结果 Baseline Results

任务一 Task 1

Category Model Precision Recall F1
All Bert+BiLSTM+CRF 68.92% 68.75% 0.6884
Bert+CRF 66.88% 65.30% 0.6608
Bert 45.49% 56.52% 0.5041
Beauty Bert+BiLSTM+CRF 72.82% 74.81% 0.7380
Bert+CRF 67.31% 68.02% 0.6766
Bert 53.55% 62.84% 0.5782
Fashion Bert+BiLSTM+CRF 65.89% 70.71% 0.6822
Bert+CRF 60.16% 66.61% 0.6322
Bert 46.45% 58.39% 0.5174
Phones Bert+BiLSTM+CRF 67.01% 69.90% 0.6843
Bert+CRF 56.20% 59.23% 0.5768
Bert 42.12% 53.84% 0.4726
Electronic Bert+BiLSTM+CRF 65.48% 67.71% 66.58%
Bert+CRF 62.12% 61.55% 0.6183
Bert 39.81% 49.40% 0.4409
Shoes Bert+BiLSTM+CRF 78.70% 81.01% 0.7984
Bert+CRF 73.02% 77.03% 0.7497
Bert 58.51% 70.20% 0.6382

任务二 Task 2

Category Model Precision Recall F1
All DiaMultiClass 0.3222 0.4966 0.3662
DiaSeq 0.3555 0.2966 0.3153
Beauty DiaMultiClass 0.4037 0.7228 0.3662
DiaSeq 0.4761 0.4272 0.4424
Fashion DiaMultiClass 0.2711 0.3488 0.2918
DiaSeq 0.1525 0.1271 0.1355
Phones DiaMultiClass 0.4534 0.5212 0.4585
DiaSeq 0.4414 0.3789 0.3966
Electronic DiaMultiClass 0.2567 0.3657 0.2851
DiaSeq 0.2420 0.1736 0.1891
Shoes DiaMultiClass 0.3361 0.4131 0.3423
DiaSeq 0.3992 0.3305 0.3498

任务三 Task 3

Category Model Hits@10 Hits@50 NDCG@10 NDCG@50 MRR@10 MRR@50
All Bert 0.1593 0.331 0.0818 0.1192 0.0582 0.066
SASRec 0.14 0.2747 0.0725 0.1022 0.0522 0.0585
TGCRS 0.147 0.25 0.0809 0.1036 0.0606 0.0655
Beauty Bert 0.2985 0.3938 0.1842 0.2057 0.1484 0.1532
SASRec 0.1631 0.3169 0.0726 0.1067 0.0449 0.0523
TGCRS 0.2831 0.3723 0.1722 0.1914 0.1374 0.1413
Fashion Bert 0.1348 0.1489 0.0854 0.0885 0.0697 0.0703
SASRec 0.0496 0.0816 0.026 0.0333 0.0188 0.0204
TGCRS 0.1099 0.1241 0.0755 0.0786 0.065 0.0657
Phones Bert 0.4275 0.7174 0.2488 0.3138 0.1943 0.2086
SASRec 0.4094 0.7065 0.2134 0.2808 0.1546 0.1698
TGCRS 0.5942 0.7681 0.3392 0.3791 0.2609 0.2702
Electronic Bert 0.2576 0.4333 0.1484 0.1864 0.1145 0.1222
SASRec 0.1758 0.2788 0.1067 0.1283 0.0849 0.089
TGCRS 0.2818 0.397 0.169 0.1949 0.1344 0.1402
Shoes Bert 0.1014 0.255 0.0473 0.0813 0.0312 0.0384
SASRec 0.0691 0.1674 0.0388 0.0602 0.0296 0.0341
TGCRS 0.1521 0.255 0.083 0.1058 0.0618 0.0668

任务四 Task 4

Category Model dist@1 dist@2 dist@3 dist@4 bleu@1 bleu@2 bleu@3 bleu@4 Info Rel
All GPT-2 0.0284 0.0624 0.1780 0.2905 0.0688 0.0276 0.0166 0.0136 0.5700 0.4267
Transformer 0.01462 0.05366 0.1563 0.2806 0.1138 0.03715 0.02037 0.01359 1.1567 0.8800
KBRD 0.01173 0.04061 0.1259 0.2233 0.1253 0.04067 0.02528 0.01879 1.1367 0.9167
NTRD 0.1485 0.1942 0.2277 0.2489 0.0443 0.0082 0.0028 0.0016 1.0033 0.9900
Beauty GPT-2 0.0581 0.1250 0.2555 0.3811 0.0610 0.0176 0.0054 0.0025 0.6433 0.2767
Transformer 0.03654 0.09977 0.2524 0.3714 0.09455 0.02516 0.01466 0.009815 1.2767 0.5267
KBRD 0.03466 0.08759 0.1977 0.2808 0.1097 0.0325 0.01999 0.01389 1.2233 0.6133
NTRD 0.1259 0.2277 0.2963 0.3275 0.0439 0.0083 0.0042 0.0030 1.1500 0.6867
Phones GPT-2 0.0497 0.1099 0.2204 0.3266 0.1110 0.0460 0.0248 0.0173 0.7633 0.4700
Transformer 0.04254 0.1209 0.2853 0.4037 0.1279 0.04393 0.02546 0.01471 1.2267 0.9467
KBRD 0.0506 0.1359 0.3017 0.4157 0.1418 0.04371 0.0204 0.008948 1.0900 0.9867
NTRD 0.1614 0.2666 0.3190 0.3775 0.0560 0.0163 0.0102 0.0072 1.0400 1.0567
Shoes GPT-2 0.0522 0.1171 0.228 0.3357 0.0803 0.0405 0.0270 0.0202 0.5000 0.3767
Transformer 0.0319 0.07772 0.1975 0.3718 0.117 0.05629 0.04072 0.03035 1.0400 0.9600
KBRD 0.03508 0.08901 0.2224 0.4051 0.1358 0.0685 0.04701 0.03219 1.0933 1.0467
NTRD 0.2220 0.4036 0.4499 0.4635 0.0432 0.0135 0.0069 0.0041 1.0100 0.9600

任务五 Task 5

Category Model PCC SCC Cos
All DEB 0.1617 0.1864 0.9212
P-value <6e-06 <1e-07 -
Bert-RUBER 0.0742 0.1092 0.9214
P-value <0.0398 <0.0024 -
Beauty DEB 0.1642 0.1628 0.9327
P-value <0.0299 <0.0313 -
Bert-RUBER 0.0901 0.1133 0.9218
P-value -0.0126 <0.0017 -
Phones DEB 0.2678 0.2815 0.9366
P-value <0.0015 <0.0008 -
Bert-RUBER 0.0900 0.1141 0.9218
P-value <0.0126 <0.0015 -
Shoes DEB 0.1504 0.1963 0.9097
P-value <0.0416 <0.0076 -
Bert-RUBER 0.0916 0.1157 0.9219
P-value <0.0111 <0.0013 -

u-need's People

Contributors

1191000606 avatar leeeeoliu avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.