All Projects → DouYishun → Kb Qa

DouYishun / Kb Qa

基于知识库的中文问答系统(biLSTM)

Programming Languages

python
139335 projects - #7 most used programming language

Projects that are alternatives of or similar to Kb Qa

Research Papers
Awesome Research Papers
Stars: ✭ 165 (-15.38%)
Mutual labels:  knowledge
Yolov3 Tf2
YoloV3 Implemented in Tensorflow 2.0
Stars: ✭ 2,327 (+1093.33%)
Mutual labels:  tensorflow-examples
Awesome Kgqa
A collection of some materials of knowledge graph question answering
Stars: ✭ 188 (-3.59%)
Mutual labels:  question-answering
Rat Sql
A relation-aware semantic parsing model from English to SQL
Stars: ✭ 169 (-13.33%)
Mutual labels:  question-answering
Tensorflow In Practice Code
源码实现:《TensorFlow实战》黄文坚,唐源 著
Stars: ✭ 176 (-9.74%)
Mutual labels:  tensorflow-examples
Interview Comment
Stars: ✭ 182 (-6.67%)
Mutual labels:  knowledge
Awesomemrc
This repo is our research summary and playground for MRC. More features are coming.
Stars: ✭ 162 (-16.92%)
Mutual labels:  question-answering
Anyq
FAQ-based Question Answering System
Stars: ✭ 2,336 (+1097.95%)
Mutual labels:  question-answering
Mspars
Stars: ✭ 177 (-9.23%)
Mutual labels:  question-answering
Openqa
The source code of ACL 2018 paper "Denoising Distantly Supervised Open-Domain Question Answering".
Stars: ✭ 188 (-3.59%)
Mutual labels:  question-answering
Questgen.ai
Question generation using state-of-the-art Natural Language Processing algorithms
Stars: ✭ 169 (-13.33%)
Mutual labels:  question-answering
Knowledge
💡 document everything
Stars: ✭ 174 (-10.77%)
Mutual labels:  knowledge
Triviaqa
Code for the TriviaQA reading comprehension dataset
Stars: ✭ 184 (-5.64%)
Mutual labels:  question-answering
Hq bot
📲 Bot to help solve HQ trivia
Stars: ✭ 167 (-14.36%)
Mutual labels:  question-answering
Sec interview know list
信息安全方面面试清单
Stars: ✭ 191 (-2.05%)
Mutual labels:  knowledge
Improved Dynamic Memory Networks Dmn Plus
Theano Implementation of DMN+ (Improved Dynamic Memory Networks) from the paper by Xiong, Merity, & Socher at MetaMind, http://arxiv.org/abs/1603.01417 (Dynamic Memory Networks for Visual and Textual Question Answering)
Stars: ✭ 165 (-15.38%)
Mutual labels:  question-answering
Fed Note
我是Mokou, 📘 这里是写前端博客和备忘学习的地方。Vue3 源码解析连载中。喜欢请Star。
Stars: ✭ 180 (-7.69%)
Mutual labels:  knowledge
Flowqa
Implementation of conversational QA model: FlowQA (with slight improvement)
Stars: ✭ 194 (-0.51%)
Mutual labels:  question-answering
Simpletransformers
Transformers for Classification, NER, QA, Language Modelling, Language Generation, T5, Multi-Modal, and Conversational AI
Stars: ✭ 2,881 (+1377.44%)
Mutual labels:  question-answering
Learn Anything
Organize world's knowledge, explore connections and curate learning paths
Stars: ✭ 13,532 (+6839.49%)
Mutual labels:  knowledge

Knowledge-Based Question Answering

基于知识库的中文问答系统. 整体流程如下:

  1. 根据Background和Question寻找到最相关的K个Knowledge,K Knowledge+Background+Question构成一个大问题.

  2. 正确选项分别与该问题中所有错误选项组合,构成3个答案组合,分别与大问题组合构成3个样例,采用余弦距离计算大问题正确选项错误选项的相似度.

    正确选项相似度为t_sim, 错误选项相似度为f_sim,损失函数为

     loss = max(0, margin - t_sim + f_sim)
    

Model

  • 寻找相关Knowledge: LSI
  • 训练: biLSTM

Requirement

  • python3, tensorflow
  • stop_words, 中文word2vec(a2u6)

Data Format

  • knowledge

      地球是宇宙中的一颗行星,有自己的运动规律。
      地球上的许多自然现象都与地球的运动密切相关。
      地球具有适合生命演化和人类发展的条件,因此,它成为人类在宇宙中的唯一家园。
      ...
    
  • train&test

    问题为选择题,每个问题的格式为 Background, Question, Right, Wrong, Wrong, Wrong.

      B:近年来,我国有些农村出现了“有院无人住,有地无人种”的空心化现象。
      Q:“有院无人住,有地无人种”带来
      R:土地资源浪费
      W:农业发展水平提高
      W:城乡协调发展
      W:农村老龄化程度降低
    
      B:广东省佛山市三水区被称为“中国饮料之都”。除青岛啤酒、伊利等国内著名饮料企业抢先布局外,百威、红牛、可口可乐、杨协成等国际巨头也先后落户于此,作为其在中国布局中的重要一环。
      Q:众多国际饮料企业选址三水的主导区位因素是
      R:市场
      W:技术
      W:劳动力
      W:原料
    
      B:凡是大气中因悬浮的水汽凝结,能见度低于1千米时,气象学称这种天气现象为雾
      Q:深秋到第二年初春,晴朗的夜晚容易形成雾,这主要是因为
      R:晴朗的夜晚大气逆辐射弱,近地面降温快
      W:晴天大气中的水汽含量多
      W:晴朗的夜晚地面水汽蒸发强
      W:晴天大气中的凝结核物质较少
    
      ...
    

Usage

python3 train.py

该数据集下最佳参数为

  • dropout:0.45
  • k:0.5
Note that the project description data, including the texts, logos, images, and/or trademarks, for each open source project belongs to its rightful owner. If you wish to add or remove any projects, please contact us at [email protected].