https://github.com/fuzhenxin/query-bag-matching-cikm
Dataset used for paper query-bag matching in CIKM2019.
https://github.com/fuzhenxin/query-bag-matching-cikm
cikm2019 natural-language-processing question-ans text-matching
Last synced: 2 months ago
JSON representation
Dataset used for paper query-bag matching in CIKM2019.
- Host: GitHub
- URL: https://github.com/fuzhenxin/query-bag-matching-cikm
- Owner: fuzhenxin
- Created: 2019-08-09T14:01:30.000Z (over 5 years ago)
- Default Branch: master
- Last Pushed: 2019-09-18T08:44:08.000Z (over 5 years ago)
- Last Synced: 2024-12-26T21:08:50.511Z (4 months ago)
- Topics: cikm2019, natural-language-processing, question-ans, text-matching
- Homepage:
- Size: 1.39 MB
- Stars: 3
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Query-Bag-Matching-CIKM
This repo contains the composed Quora dataset for the short paper "Query-bag Matching]{Query-bag Matching with Mutual Coverage for Information-seeking Conversations in E-commerce" in CIKM 2019. It also contains the detailed experiments of this paper.
# An Example of Alime
- An example of the query-bag pair and the corresponding answer.
- User query:
- 怎么电话小二? (How to call you?)
- Bag (Pre-defined questions):
- 怎么联系你们? (How to contact you?)
- 客服热线是多少? (What's the hotline?)
- Answer:
- 电话是123456 (The phone number is 123456.)
# Composed Quora Dataset
- The dataset is in the dataset directory.
- The Quora dataset is composed from the [Quora Question Pairs Dataset](https://www.kaggle.com/quora/question-pairs-dataset)
# Experiment Setup
## Data
- max sequence length: 20
## hCNN
### Convolution2d for cross-attention matrix
- Conv1: kernel size 6 ; num output 8 ; stride 1 ; paddding "SAME" ; activation Relu ;
- Max pooling1: pool size 4,4 ; stride 4 ; padding "SAME"
- Conv2: kernel size 4 ; num output 16 ; stride 3 ; paddding "SAME" ; activation Relu ;
- Max pooling2: pool size 2,2 ; stride 2 ; padding "SAME"
### Convolution1d for query and question
- filters 100 ; kernel size 2/3/4 ; max pooling pool size 20-kernel_size + 1 ; max pooling stride 1