“Hulan-2015-01-29”版本间的差异

来自cslt Wiki
跳转至: 导航搜索
(以“=Dialog system= ==Algorithm== ===Spell mistake=== :* train a new lm (or mix the two lm)('''caoli''') ===improve fuzzy match=== * add Synonyms similarity using MERT-4...”为内容创建页面)
 
Lr讨论 | 贡献
reverse module
 
(相同用户的2个中间修订版本未显示)
第2行: 第2行:
 
==Algorithm==
 
==Algorithm==
 
===Spell mistake===
 
===Spell mistake===
:* train a new lm (or mix the two lm)('''caoli''')
+
:* done
 
===improve fuzzy match===
 
===improve fuzzy match===
* add Synonyms similarity using MERT-4 method
+
* hold
 
===improve lucene search===
 
===improve lucene search===
* lucene Optimization(liurong)
+
* learning to rank [http://cslt.riit.tsinghua.edu.cn/mediawiki/index.php/Huilan-learning-to-rank]
:* rewrite the method to select the 50 standard question not same template.(hold)
+
:* result: 0.688->0.78
:* using MERT-4 method to get good value of multi-feature.like IDF,NER,baidu_weight,keyword etc.(hold)
+
* MERT-4 method ('''caoli liurong''')
+
:* feature score: standard question,standard question and answer, standard question and extracted question,baixing-dict,sogou-weight.
+
 
===Multi-Scene Recognition===
 
===Multi-Scene Recognition===
 
* demo (done)
 
* demo (done)
===Knowledge Online Learning===
+
===Online Learning===
* research the algorithm('''liurong''')
+
* begin to set.
* add the index.
+
===reverse module===
==knowledge structure==
+
* knowledge graph
* how to inference the knowledge.
+
:* process of setting knowledge graph and support reverse ask[http://cslt.riit.tsinghua.edu.cn/mediawiki/images/1/18/%E6%94%BF%E5%BA%9C%E7%BB%84%E7%BB%87%E6%9C%BA%E6%9E%84%E5%9B%BE%E8%B0%B1--%E6%B1%87%E8%81%94.pdf]
 +
:* search the opensource jena and D2R
 +
 
 
==Knowledge Management and labeling system==
 
==Knowledge Management and labeling system==
 
* continue coding.
 
* continue coding.
 
==Patent==
 
==Patent==
* the GA method to improve QA,draft done .('''liurong''')
+
* done
* Submit patent
+
 
==new word find==
 
==new word find==
* get code from duxk
+
* done
* add sensitive words
+
==source from company==
+
* baixiti-dict ('''duxk''')
+
* history test set ('''duxk''')
+

2015年1月29日 (四) 12:11的最后版本

Dialog system

Algorithm

Spell mistake

  • done

improve fuzzy match

  • hold

improve lucene search

  • learning to rank [1]
  • result: 0.688->0.78

Multi-Scene Recognition

  • demo (done)

Online Learning

  • begin to set.

reverse module

  • knowledge graph
  • process of setting knowledge graph and support reverse ask[2]
  • search the opensource jena and D2R

Knowledge Management and labeling system

  • continue coding.

Patent

  • done

new word find

  • done