cslt Wiki - 用户贡献 [zh-cn]

NLP Status Report 2017-9-11

2017-09-11T07:19:37Z

Zhangjy：以“{| class="wikitable" !Date !! People !! Last Week !! This Week |- | rowspan="6"|2017/8/14 |Jiyuan Zhang || || |- |Aodong LI || || |- |Shiyue Zhang || || |- |Sh...”为内容创建页面

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/8/14
|Jiyuan Zhang ||
||

|-
|Aodong LI ||

||

|-
|Shiyue Zhang ||

||
|-
|Shipan Ren
||
||
|-

|Jiayu Guo||

||
|-
|-

|}

NLP Status Report 2017-8-28

2017-08-28T06:35:51Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/8/14
|Jiyuan Zhang ||
*code refactoring
*wrote a document[http://cslt.riit.tsinghua.edu.cn/mediawiki/index.php/%E6%96%87%E4%BB%B6:VvPoem.docx]
||

|-
|Aodong LI ||

||

|-
|Shiyue Zhang ||

||

|-
|Shipan Ren
||
* read the released information of other toolkits for nmt
* cleaned up the code
* wrote the documents
||
* write the papers of our baseline system
* read augmented nmt code
|-

|Jiayu Guo||
* process data and run model;
* test results.
checkpoint-100000 translation model
BLEU： 11.11

*source:在秦者名错，与张仪争论,於是惠王使错将伐蜀，遂拔，因而守之。
*target:在秦国的名叫司马错，曾与张仪发生争论，秦惠王采纳了他的意见，于是司马错率军攻蜀国，攻取后，又让他做了蜀地郡守。
*trans：当时秦国的人都很欣赏他的建议，与张仪一起商议，所以吴王派使者率军攻打蜀地，一举攻，接着又下令守城。
*source:神大用则竭，形大劳则敝，形神离则死。
*target:精神过度使用就会衰竭，形体过度劳累就会疲惫，神形分离就会死亡。
*trans: 精神过度就可衰竭,身体过度劳累就会疲惫，地形也就会死。
*source:今天子接千岁之统，封泰山，而余不得从行，是命也夫，命也夫！
*target:现天子继承汉朝千年一统的大业，在泰山举行封禅典礼而我不能随行，这是命啊，是命啊！
*trans: 现在天子可以继承帝位的成就爵位，爵位至泰山，而我却未能执行先帝的命运。
*1.data used Zizhitongjian only(6,000 pairs), we can get BLEU 6 at most.
*2.data used Zizhitongjian only(12,000 pairs), we can get BLEU 7 at most.
*3.data used Shiji and Zizhitongjian(43,0000 pairs), we can get BLEU about 9.
*4.data used Shiji and Zizhitongjian(43,0000 pairs), and split the ancient language text one character by one, we can get BLEU 11.11 at most.
*The main factors now is the data(pairs of sentence/the quality——the modern language text includes context information).
||
*plan to read source code of seq2seq model and learn tensorflow;
*plan to read a paper named Automatic Long Sentence Segmentation for NMT
|-
|-
|zhangshuai
||
* learn model source code
||
* learn tensorflow and source code
|-
|}

NLP Status Report 2017-8-28

2017-08-28T06:35:31Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/8/14
|Jiyuan Zhang ||
*code refactoring
*wrote a document[[http://cslt.riit.tsinghua.edu.cn/mediawiki/index.php/%E6%96%87%E4%BB%B6:VvPoem.docx]]
||

|-
|Aodong LI ||

||

|-
|Shiyue Zhang ||

||

|-
|Shipan Ren
||
* read the released information of other toolkits for nmt
* cleaned up the code
* wrote the documents
||
* write the papers of our baseline system
* read augmented nmt code
|-

|Jiayu Guo||
* process data and run model;
* test results.
checkpoint-100000 translation model
BLEU： 11.11

*source:在秦者名错，与张仪争论,於是惠王使错将伐蜀，遂拔，因而守之。
*target:在秦国的名叫司马错，曾与张仪发生争论，秦惠王采纳了他的意见，于是司马错率军攻蜀国，攻取后，又让他做了蜀地郡守。
*trans：当时秦国的人都很欣赏他的建议，与张仪一起商议，所以吴王派使者率军攻打蜀地，一举攻，接着又下令守城。
*source:神大用则竭，形大劳则敝，形神离则死。
*target:精神过度使用就会衰竭，形体过度劳累就会疲惫，神形分离就会死亡。
*trans: 精神过度就可衰竭,身体过度劳累就会疲惫，地形也就会死。
*source:今天子接千岁之统，封泰山，而余不得从行，是命也夫，命也夫！
*target:现天子继承汉朝千年一统的大业，在泰山举行封禅典礼而我不能随行，这是命啊，是命啊！
*trans: 现在天子可以继承帝位的成就爵位，爵位至泰山，而我却未能执行先帝的命运。
*1.data used Zizhitongjian only(6,000 pairs), we can get BLEU 6 at most.
*2.data used Zizhitongjian only(12,000 pairs), we can get BLEU 7 at most.
*3.data used Shiji and Zizhitongjian(43,0000 pairs), we can get BLEU about 9.
*4.data used Shiji and Zizhitongjian(43,0000 pairs), and split the ancient language text one character by one, we can get BLEU 11.11 at most.
*The main factors now is the data(pairs of sentence/the quality——the modern language text includes context information).
||
*plan to read source code of seq2seq model and learn tensorflow;
*plan to read a paper named Automatic Long Sentence Segmentation for NMT
|-
|-
|zhangshuai
||
* learn model source code
||
* learn tensorflow and source code
|-
|}

文件:VvPoem.docx

2017-08-28T06:21:54Z

Zhangjy：

NLP Status Report 2017-8-28

2017-08-28T06:18:48Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/8/14
|Jiyuan Zhang ||
*code refactoring
*wrote a document
||

|-
|Aodong LI ||

||

|-
|Shiyue Zhang ||

||

|-
|Shipan Ren ||
* organized all the experimental results(our baseline system,Moses,THUMT)
* trained and tested translation models（Toolkit:THUMT ）
* compared with our system
||
* prepare to release the baseline system（tensorflow1.0 version）
|-

|Jiayu Guo||
* process data and run model;
* test results.
checkpoint-100000 translation model
BLEU： 11.11

*source:在秦者名错，与张仪争论,於是惠王使错将伐蜀，遂拔，因而守之。
*target:在秦国的名叫司马错，曾与张仪发生争论，秦惠王采纳了他的意见，于是司马错率军攻蜀国，攻取后，又让他做了蜀地郡守。
*trans：当时秦国的人都很欣赏他的建议，与张仪一起商议，所以吴王派使者率军攻打蜀地，一举攻，接着又下令守城。
*source:神大用则竭，形大劳则敝，形神离则死。
*target:精神过度使用就会衰竭，形体过度劳累就会疲惫，神形分离就会死亡。
*trans: 精神过度就可衰竭,身体过度劳累就会疲惫，地形也就会死。
*source:今天子接千岁之统，封泰山，而余不得从行，是命也夫，命也夫！
*target:现天子继承汉朝千年一统的大业，在泰山举行封禅典礼而我不能随行，这是命啊，是命啊！
*trans: 现在天子可以继承帝位的成就爵位，爵位至泰山，而我却未能执行先帝的命运。
*1.data used Zizhitongjian only(6,000 pairs), we can get BLEU 6 at most.
*2.data used Zizhitongjian only(12,000 pairs), we can get BLEU 7 at most.
*3.data used Shiji and Zizhitongjian(43,0000 pairs), we can get BLEU about 9.
*4.data used Shiji and Zizhitongjian(43,0000 pairs), and split the ancient language text one character by one, we can get BLEU 11.11 at most.
*The main factors now is the data(pairs of sentence/the quality——the modern language text includes context information).
||
*plan to read source code of seq2seq model and learn tensorflow;
*plan to read a paper named Automatic Long Sentence Segmentation for NMT
|-
|-
|zhangshuai
||
* learn model source code
||
* learn tensorflow and source code
|-
|}

NLP Status Report 2017-8-28

2017-08-28T06:13:55Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/8/14
|Jiyuan Zhang ||
*done some work about code refactoring for poem system
||
*plan to complete code refactoring for poem system
|-
|Aodong LI ||

||

|-
|Shiyue Zhang ||

||

|-
|Shipan Ren ||
* organized all the experimental results(our baseline system,Moses,THUMT)
* trained and tested translation models（Toolkit:THUMT ）
* compared with our system
||
* prepare to release the baseline system（tensorflow1.0 version）
|-

|Jiayu Guo||
* process data and run model;
* test results.
checkpoint-100000 translation model
BLEU： 11.11

*source:在秦者名错，与张仪争论,於是惠王使错将伐蜀，遂拔，因而守之。
*target:在秦国的名叫司马错，曾与张仪发生争论，秦惠王采纳了他的意见，于是司马错率军攻蜀国，攻取后，又让他做了蜀地郡守。
*trans：当时秦国的人都很欣赏他的建议，与张仪一起商议，所以吴王派使者率军攻打蜀地，一举攻，接着又下令守城。
*source:神大用则竭，形大劳则敝，形神离则死。
*target:精神过度使用就会衰竭，形体过度劳累就会疲惫，神形分离就会死亡。
*trans: 精神过度就可衰竭,身体过度劳累就会疲惫，地形也就会死。
*source:今天子接千岁之统，封泰山，而余不得从行，是命也夫，命也夫！
*target:现天子继承汉朝千年一统的大业，在泰山举行封禅典礼而我不能随行，这是命啊，是命啊！
*trans: 现在天子可以继承帝位的成就爵位，爵位至泰山，而我却未能执行先帝的命运。
*1.data used Zizhitongjian only(6,000 pairs), we can get BLEU 6 at most.
*2.data used Zizhitongjian only(12,000 pairs), we can get BLEU 7 at most.
*3.data used Shiji and Zizhitongjian(43,0000 pairs), we can get BLEU about 9.
*4.data used Shiji and Zizhitongjian(43,0000 pairs), and split the ancient language text one character by one, we can get BLEU 11.11 at most.
*The main factors now is the data(pairs of sentence/the quality——the modern language text includes context information).
||
*plan to read source code of seq2seq model and learn tensorflow;
*plan to read a paper named Automatic Long Sentence Segmentation for NMT
|-
|-
|zhangshuai
||
* learn model source code
||
* learn tensorflow and source code
|-
|}

NLP Status Report 2017-8-21

2017-08-21T05:13:42Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/8/14
|Jiyuan Zhang ||
*done some work about code refactoring for poem system
||
*plan to complete code refactoring for poem system
|-
|Aodong LI ||

||

|-
|Shiyue Zhang ||

||

|-
|Shipan Ren ||
* organized all the experimental results(our baseline system,Moses,THUMT)
* train translation models by using THUMT
* test the bleu of these models
* compare with our system
||
* prepare to release the baseline system（tensorflow1.0 version）
|-

|Jiayu Guo||
* process data and run model;
* test results.
checkpoint-100000 translation model
BLEU： 11.11

*source:在秦者名错，与张仪争论,於是惠王使错将伐蜀，遂拔，因而守之。
*target:在秦国的名叫司马错，曾与张仪发生争论，秦惠王采纳了他的意见，于是司马错率军攻蜀国，攻取后，又让他做了蜀地郡守。
*trans：当时秦国的人都很欣赏他的建议，与张仪一起商议，所以吴王派使者率军攻打蜀地，一举攻，接着又下令守城。
*source:神大用则竭，形大劳则敝，形神离则死。
*target:精神过度使用就会衰竭，形体过度劳累就会疲惫，神形分离就会死亡。
*trans: 精神过度就可衰竭,身体过度劳累就会疲惫，地形也就会死。
*source:今天子接千岁之统，封泰山，而余不得从行，是命也夫，命也夫！
*target:现天子继承汉朝千年一统的大业，在泰山举行封禅典礼而我不能随行，这是命啊，是命啊！
*trans: 现在天子可以继承帝位的成就爵位，爵位至泰山，而我却未能执行先帝的命运。

||
* read source code of seq2seq model;
*1.data used Zizhitongjian only(6,000 pairs), we can get BLEU 6 at most.
*2.data used Zizhitongjian only(12,000 pairs), we can get BLEU 7 at most.
*3.data used Shiji and Zizhitongjian(43,0000 pairs), we can get BLEU about 9.
*4.data used Shiji and Zizhitongjian(43,0000 pairs), and split the ancient language text one character by one, we can get BLEU 11.11 at most.
*The main factors now is the data(including pairs of sentence、the quality——cause the modern language text include context information.
|-
|}

NLP Status Report 2017-8-21

2017-08-21T05:04:48Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/8/14
|Jiyuan Zhang ||
*done work about code refactoring for poem system
||
*plan to complete code refactoring for poem system
|-
|Aodong LI ||

||

|-
|Shiyue Zhang ||

||

|-
|Shipan Ren ||
* organized the results of the experiment
* learned how to use THUMT and how did it work
||
* train translation models by using THUMT
* test the bleu of these models
* compare with our system
|-

|Jiayu Guo||
* process data and run model;
* test results.
checkpoint-100000 translation model
BLEU： 11.11

*source:在秦者名错，与张仪争论,於是惠王使错将伐蜀，遂拔，因而守之。
*target:在秦国的名叫司马错，曾与张仪发生争论，秦惠王采纳了他的意见，于是司马错率军攻蜀国，攻取后，又让他做了蜀地郡守。
*trans：当时秦国的人都很欣赏他的建议，与张仪一起商议，所以吴王派使者率军攻打蜀地，一举攻，接着又下令守城。
*source:神大用则竭，形大劳则敝，形神离则死。
*target:精神过度使用就会衰竭，形体过度劳累就会疲惫，神形分离就会死亡。
*trans: 精神过度就可衰竭,身体过度劳累就会疲惫，地形也就会死。
*source:今天子接千岁之统，封泰山，而余不得从行，是命也夫，命也夫！
*target:现天子继承汉朝千年一统的大业，在泰山举行封禅典礼而我不能随行，这是命啊，是命啊！
*trans: 现在天子可以继承帝位的成就爵位，爵位至泰山，而我却未能执行先帝的命运。

||
* read source code of seq2seq model;
*1.data used Zizhitongjian only(6,000 pairs), we can get BLEU 6 at most.
*2.data used Zizhitongjian only(12,000 pairs), we can get BLEU 7 at most.
*3.data used Shiji and Zizhitongjian(43,0000 pairs), we can get BLEU about 9.
*4.data used Shiji and Zizhitongjian(43,0000 pairs), and split the ancient language text one character by one, we can get BLEU 11.11 at most.
*The main factors now is the data(including pairs of sentence、the quality——cause the modern language text include context information.
|-
|}

NLP Status Report 2017-8-21

2017-08-21T05:03:10Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/8/14
|Jiyuan Zhang ||
*polished the couplet model
||
Code refactoring for poem system
|-
|Aodong LI ||

||

|-
|Shiyue Zhang ||

||

|-
|Shipan Ren ||
* organized the results of the experiment
* learned how to use THUMT and how did it work
||
* train translation models by using THUMT
* test the bleu of these models
* compare with our system
|-

|Jiayu Guo||
* process data and run model;
* test results.
checkpoint-100000 translation model
BLEU： 11.11

*source:在秦者名错，与张仪争论,於是惠王使错将伐蜀，遂拔，因而守之。
*target:在秦国的名叫司马错，曾与张仪发生争论，秦惠王采纳了他的意见，于是司马错率军攻蜀国，攻取后，又让他做了蜀地郡守。
*trans：当时秦国的人都很欣赏他的建议，与张仪一起商议，所以吴王派使者率军攻打蜀地，一举攻，接着又下令守城。
*source:神大用则竭，形大劳则敝，形神离则死。
*target:精神过度使用就会衰竭，形体过度劳累就会疲惫，神形分离就会死亡。
*trans: 精神过度就可衰竭,身体过度劳累就会疲惫，地形也就会死。
*source:今天子接千岁之统，封泰山，而余不得从行，是命也夫，命也夫！
*target:现天子继承汉朝千年一统的大业，在泰山举行封禅典礼而我不能随行，这是命啊，是命啊！
*trans: 现在天子可以继承帝位的成就爵位，爵位至泰山，而我却未能执行先帝的命运。

||
* read source code of seq2seq model;
*1.data used Zizhitongjian only(6,000 pairs), we can get BLEU 6 at most.
*2.data used Zizhitongjian only(12,000 pairs), we can get BLEU 7 at most.
*3.data used Shiji and Zizhitongjian(43,0000 pairs), we can get BLEU about 9.
*4.data used Shiji and Zizhitongjian(43,0000 pairs), and split the ancient language text one character by one, we can get BLEU 11.11 at most.
*The main factors now is the data(including pairs of sentence、the quality——cause the modern language text include context information.
|-
|}

NLP Status Report 2017-8-14

2017-08-14T06:15:17Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/7/3
|Jiyuan Zhang ||
*polished the couplet model
||
Code refactoring for poem system
|-
|Aodong LI ||
||
|-
|Shiyue Zhang ||
||
|-
|Shipan Ren ||
||
|-
|Jiayu Guo||
||
|-

}

NLP Status Report 2017-8-14

2017-08-14T06:14:14Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/7/3
|Jiyuan Zhang ||
*polished the couplet model
||

|-
|Aodong LI ||
||
|-
|Shiyue Zhang ||
||
|-
|Shipan Ren ||
||
|-
|Jiayu Guo||
||
|-

}

NLP Status Report 2017-8-14

2017-08-14T06:11:35Z

Zhangjy：

NLP Status Report 2017-8-14

2017-08-14T06:10:56Z

Zhangjy：

NLP Status Report 2017-8-14

2017-08-14T06:10:28Z

Zhangjy：以“{| class="wikitable" !Date !! People !! Last Week !! This Week |- | rowspan="6"|2017/7/3 |Jiyuan Zhang || *polished the couplet model || |- |Aodong LI || || |- |Sh...”为内容创建页面

NLP Status Report 2017-8-7

2017-08-07T04:29:08Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/7/3
|Jiyuan Zhang ||
*generated streame according to a couplet
*almost completed the task of filling in the blanks of a couplet
||
*continue to perfect the couplet model
|-
|Aodong LI ||
* Got 55,000+ Englsih poems and 260,000+ lines after preprocessing
* Added phase separators as the style indicator, and every line has at least one separator
* Training loss didn't decrease very much, only from 440 to 50
* The translation quality deteriorated when added language model
||
* Try to use a larger language model to decrease the training loss
* Try to use character-based MT in English-Chinese translation
|-
|Shiyue Zhang ||

||

|-
|Shipan Ren ||
* looked for the performance(the bleu value) of other models
on the WMT2014 dataset from the published papers,but not found.
* installed and built Moses on the server
||
* train statistical machine translation model and test it
toolkit: Moses
data sets:WMT2014 en-de、en-fr data sets
* collate experimental results.compare our baseline model with Moses
|-

|Jiayu Guo||
*process document.Until now, Shiji has been split up to 2,4000 pairs of sentence.
*Zizhitongjian has been split up to 1,6000 pairs.
||
*adjust jieba source code, in order to make jieba more accurate for ancient language wordpiece
*read model source code
|-
|}

NLP Status Report 2017-8-7

2017-08-07T04:27:54Z

Zhangjy：以“{| class="wikitable" !Date !! People !! Last Week !! This Week |- | rowspan="6"|2017/7/3 |Jiyuan Zhang || *made the poster for ACL [http://cslt.riit.tsinghua.edu.cn/...”为内容创建页面

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/7/3
|Jiyuan Zhang ||
*made the poster for ACL [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/9/95/Acl2017-poster.pdf]
*attempted to fix repeated word, but failed
*done some work of n-gram model of the couplet
||
*generate streame according to a couplet
*complete the task of filling in the blanks of a couplet

|-
|Aodong LI ||
* Got 55,000+ Englsih poems and 260,000+ lines after preprocessing
* Added phase separators as the style indicator, and every line has at least one separator
* Training loss didn't decrease very much, only from 440 to 50
* The translation quality deteriorated when added language model
||
* Try to use a larger language model to decrease the training loss
* Try to use character-based MT in English-Chinese translation
|-
|Shiyue Zhang ||

||

|-
|Shipan Ren ||
* looked for the performance(the bleu value) of other models
on the WMT2014 dataset from the published papers,but not found.
* installed and built Moses on the server
||
* train statistical machine translation model and test it
toolkit: Moses
data sets:WMT2014 en-de、en-fr data sets
* collate experimental results.compare our baseline model with Moses
|-

|Jiayu Guo||
*process document.Until now, Shiji has been split up to 2,4000 pairs of sentence.
*Zizhitongjian has been split up to 1,6000 pairs.
||
*adjust jieba source code, in order to make jieba more accurate for ancient language wordpiece
*read model source code
|-
|}

NLP Status Report 2017-7-31

2017-07-31T06:41:29Z

Zhangjy：

NLP Status Report 2017-7-31

2017-07-31T06:41:08Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/7/3
|Jiyuan Zhang ||
*made the poster for ACL [[http://cslt.riit.tsinghua.edu.cn/mediawiki/images/9/95/Acl2017-poster.pdf]]
*attempted to fix repeated word, but failed
*done some work of n-gram model of the couplet
||
*generate streame according to a couplet
*complete the task of filling in the blanks of a couplet

|-
|Aodong LI ||
* Got 55,000+ Englsih poems and 260,000+ lines after preprocessing
* Added phase separators as the style indicator, and every line has at least one separator
* Training loss didn't decrease very much, only from 440 to 50
* The translation quality deteriorated when added language model
||
* Try to use a larger language model to decrease the training loss
* Try to use character-based MT in English-Chinese translation
|-
|Shiyue Zhang ||

||

|-
|Shipan Ren ||
* looked for the performance(the bleu value) of other models
on the WMT2014 dataset from the published papers,but not found.
* installed and built Moses on the server
||
* train statistical machine translation model and test it
toolkit: Moses
data sets:WMT2014 en-de、en-fr data sets
* collate experimental results.compare our baseline model with Moses
|-

|Jiayu Guo||
*process document.Until now, Shiji has been split up to 2,4000 pairs of sentence.
*Zizhitongjian has been split up to 1,6000 pairs.
||
*adjust jieba source code, in order to make jieba more accurate for ancient language wordpiece
*read model source code
|-
|}

文件:Acl2017-poster.pdf

2017-07-31T06:40:28Z

Zhangjy：

NLP Status Report 2017-7-31

2017-07-31T04:38:52Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/7/3
|Jiyuan Zhang ||
*made the poster for ACL
*attempted to fix repeated word, but failed
*done some work of n-gram model of the couplet
||
*generate streame according to a couplet
*complete the task of filling in the blanks of a couplet

|-
|Aodong LI ||

||

|-
|Shiyue Zhang ||

||

|-
|Shipan Ren ||
* trained two models of the baseline using WMT2014 en-fr datasets
under training

* read some papers（memory-augmented-nmt and Memory augmented Chinese-Uyghur Neural Machine Translation）
||
* read memory-augmented-nmt code
* read papers about memory augmented NMT
|-

|Jiayu Guo||
*Until now, Shiji has been split up to 2,5000 pairs of sentence, Zizhitongjian has been split up to 2,0000 pairs.
*
||
*process document
|-
|}

NLP Status Report 2017-7-31

2017-07-31T04:36:04Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/7/3
|Jiyuan Zhang ||
*made the poster for ACL
*attempted to fix repeated word, but failed
*done some work of n-gram model of the couplet
||
*complete
|-
|Aodong LI ||

||

|-
|Shiyue Zhang ||

||

|-
|Shipan Ren ||
* trained two models of the baseline using WMT2014 en-fr datasets
under training

* read some papers（memory-augmented-nmt and Memory augmented Chinese-Uyghur Neural Machine Translation）
||
* read memory-augmented-nmt code
* read papers about memory augmented NMT
|-

|Jiayu Guo||
*Until now, Shiji has been split up to 2,5000 pairs of sentence, Zizhitongjian has been split up to 2,0000 pairs.
*
||
*process document
|-
|}

NLP Status Report 2017-7-24

2017-07-26T02:41:06Z

Zhangjy：以“{| class="wikitable" !Date !! People !! Last Week !! This Week |- | rowspan="6"|2017/7/3 |Jiyuan Zhang || * || *make the poster for ACL *complete neural model for t...”为内容创建页面

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/7/3
|Jiyuan Zhang ||
*
||
*make the poster for ACL
*complete neural model for the couplet
|-
|Aodong LI ||
*

||
*
|-
|Shiyue Zhang ||

||

|-
|Shipan Ren ||
* found ways to tokenize the WMT2014 data
rewrote prepare_data.py form moses-smt
used the tokenizer of moses-smt

*train two versions of the code on WMT2014 en-de and en-fr datasets
tested these checkpoints of en-de dataset

||
* tested these checkpoints of en-fr dataset
* record the result and do analysis
* read papers about memory augmented NMT
|-

|}

NLP Status Report 2017-7-17

2017-07-18T04:32:21Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/7/3
|Jiyuan Zhang ||
*
||
*generate streame according to a couplet
*try my best to complete the task of filling in the blanks of a couplet
|-
|Aodong LI ||
*

||
*
|-
|Shiyue Zhang ||

||

|-
|Shipan Ren ||
* found ways to tokenize the WMT2014 data
rewrote prepare_data.py form moses-smt
used the tokenizer of moses-smt

*train two versions of the code on WMT2014 en-de and en-fr datasets
tested these checkpoints of en-de dataset

||
* tested these checkpoints of en-fr dataset
* record the result and do analysis
* read papers about memory augmented NMT
|-

|}

NLP Status Report 2017-7-10

2017-07-10T04:48:29Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/7/3
|Jiyuan Zhang ||
*reproduced the couplet model using moses
||
*continue to modify the couplet
|-
|Aodong LI ||
* Tried seq2seq with or without attention model to do style transfer (cross domain) task but this didn't work due to overfitting
seq2seq with attention model: Chinese-to-English
vanilla seq2seq model: English-to-English (Unsupervised)
* Read two style controlled papers in generative model field
* Trained seq2seq with style code model
||
* Understand the model and mechanism mentioned in the two related papers
* Figure out new ways to do style transfer task
|-
|Shiyue Zhang ||

||

|-
|Shipan Ren ||
* read and run ViVi_NMT code
* read the API of tensorflow
* debugged ViVi_NMT and upgraded code version to tensorflow1.0
* found the new version saves more time，has lower complexity and better bleu than before
||
* test two versions of the code on small data sets (Chinese-English) and large data sets (Chinese-English) respectively
* test two versions of the code on WMT 2014 English-to-German parallel dataset and WMT 2014 English-French dataset respectively
* record experimental results
* read paper and try to make the bleu become a little better
|-

|}

NLP Status Report 2017-7-10

2017-07-10T04:41:28Z

Zhangjy：以“{| class="wikitable" !Date !! People !! Last Week !! This Week |- | rowspan="6"|2017/7/3 |Jiyuan Zhang || || |- |Aodong LI || * Tried seq2seq with or without attent...”为内容创建页面

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/7/3
|Jiyuan Zhang ||
||
|-
|Aodong LI ||
* Tried seq2seq with or without attention model to do style transfer (cross domain) task but this didn't work due to overfitting
seq2seq with attention model: Chinese-to-English
vanilla seq2seq model: English-to-English (Unsupervised)
* Read two style controlled papers in generative model field
* Trained seq2seq with style code model
||
* Understand the model and mechanism mentioned in the two related papers
* Figure out new ways to do style transfer task
|-
|Shiyue Zhang ||

||

|-
|Shipan Ren ||
* read and run ViVi_NMT code
* read the API of tensorflow
* debugged ViVi_NMT and upgraded code version to tensorflow1.0
* found the new version saves more time，has lower complexity and better bleu than before
||
* test two versions of the code on small data sets (Chinese-English) and large data sets (Chinese-English) respectively
* test two versions of the code on WMT 2014 English-to-German parallel dataset and WMT 2014 English-French dataset respectively
* record experimental results
* read paper and try to make the bleu become a little better
|-

|}

Weekly meeting

2017-06-26T09:41:23Z

Zhangjy：

*Location: FIT-1-304
*Time: Monday, 7:00 PM

{| class="wikitable"
! Date !! Speaker!! Title !! Materials !! On duty
|-
| 2012/08/27 ||Dong Wang || Heterogeneous Convolutive Non-negative Sparse Coding ||[[媒体文件:Heterogeneous_convolutive_non-negative_sparse_coding.pdf|slides]] [http://homepages.inf.ed.ac.uk/v1dwang2/public/pdf/inerspeech2012-hetero.pdf paper] ||
|-
|2012/09/03 ||NO Meeting|| || ||
|-
|2012/09/10 || NO Meeting|| || ||
|-
|2012/09/17 ||WALEED ABDULLA||Auditory Based Feature Vectors for Speech Recognition ||[[媒体文件:AuditoryBasedFeatureVectors.pdf|slides]]||范淼
|-
| rowspan="2"|2012/09/24 ||刘超|| N-gram FST indexing for Spoken Term Detection || [[媒体文件:120924-N_gram_FST_indexing_for_Spoken_Term_Detection-LC-0.pdf|slides]] ||尹聪
|-
|范淼||Micro-blogging, Wikipedia, Folksonomy, What's Next? ||[[媒体文件:120924-Micro-blogging, Wikipedia, Folksonomy, What's Next-FM--01-FM-.pdf|slides]] ||
|-
| 2012/10/08 ||NO Meeting|| || ||
|-
| 2012/10/15 ||NO Meeting|| || ||
|-
|2012/10/22||Wu Xiaojun||speaker recognition in CSLT ||[[媒体文件:VPR_in_CSLT.pdf|slides]]||卡尔
|-
| rowspan="1"|2012/10/29 ||王军||An overview of Automatic Speaker Diarization Systems || [[媒体文件:121027-Speaker Diarization-WJ.pdf|slides]] ||别凡虎
|-
| rowspan="1"|2012/11/05 ||别凡虎||Experiments on Emotional Speaker Recognition||[[媒体文件:121104-Experiments_on_Emotional_Speaker_Recognition-BFH.pdf|slides]] ||刘超
|-
| rowspan="1"|2012/11/12 ||唐国瑜||Statistical Word Sense Improves Document Clustering ||[[媒体文件:121112_Statistical_Word_Sense_Improves_Document_Clustering_TGY.pdf‎ |slides]]||邱晗
|-
| rowspan="1"|2012/11/19 ||张陈昊||TDSR with Long-term Features Based on Functional Data Analysis||[[媒体文件:121118-ISCSLP-FDA_SR-ZCH.pdf|slides]] ||王俊俊
|-
| rowspan="1"|2012/11/26 ||王琳琳||Time-Varying Speaker Recognition: An Introduction||[[媒体文件:121126-Time_Varying_Speaker_Recognition_I-Wll.pdf‎|slides]] ||龚宬
|-
| rowspan="1"|2012/12/03 ||No meeting|| || ||
|-
| rowspan="1"|2012/12/10 ||No meeting|| || ||
|-
| rowspan="1"|2012/12/17 ||No meeting|| || ||

|-
| rowspan="1"|2012/01/07 || || || ||
|-
|2012/01/07 ||王军||基于DF-MAP的说话人模型训练方法||[[媒体文件:130107-基于DFMAP的说话人模型训练方法-WJ.pdf|slides]] ||唐国瑜
|-
| rowspan="1"|2012/01/14 ||王东|| Computing in CSLT ||[[媒体文件:Computing_in_CSLT.pdf|slides]] ||王琳琳
|-

|-
| rowspan="1"|2013/03/04 ||王军||Sequential Adaptive Learning for Speaker Verification ||[[媒体文件:130301-Sequential adaptive learning for speaker verification-WJ.pdf|slides]] ||别凡虎
|-
| rowspan="1"|2013/03/11 || Du Jinle|| VAD stuff || ||
|-
| rowspan="1"|2013/03/18 || || || ||
|-
| rowspan="1"|2013/03/25 || || || ||
|-
| rowspan="1"|2013/04/01 || || || ||
|-
| rowspan="1"|2013/04/08 || 张陈昊|| A Fishervoice based Feature Fusion Method for SUSR ||[[媒体文件:130408-FisherVoice-ZCH.pdf|slides]] ||谢仲达
|-
| rowspan="1"|2013/04/15 ||龚宬|| An Exploration on Influence Factors of VAD's Performance in Speaker Recognition ||[[媒体文件:130415-An_Exploration_on_Influence_Factors_of_VAD-GC.pdf|slides]] ||
|-
| rowspan="1"|2013/04/22 ||王俊俊 || Understanding the Query: THCIB and THUIS at NTCIR-10 Intent Task ||[[媒体文件:130422-Understanding_the_Query-WJJ.pdf|slides‎]] ||
|-
| rowspan="1"|2013/04/29 || || || ||
|-
| rowspan="1"|2013/05/06 ||别凡虎 ||MLLR on Emotional Speaker Recognition ||[[媒体文件:130506-MLLR on Emotional Speaker Recognition-BFH.pdf|slides]] ||
|-
| rowspan="1"|2013/05/13 ||刘超 || The Use of Deep Neural Network for Speech Recognition || [[媒体文件:130513-the_use_of_dnn_for_asr-lc.pdf|slides]] ||
|-
| rowspan="1"|2013/05/20 || || || ||
|-
| rowspan="1"|2013/05/27 ||王琳琳|| 说话人识别中的时变鲁棒性问题研究 || [[媒体文件:130527-TVSV-Wll.pdf|slides]] ||
|-
| rowspan="1"|2013/06/03 ||王俊俊|| 汉语搜索结果聚类系统研究与实现 || [[媒体文件:130601-毕业答辩-02-WJJ.pdf|slides]] ||
|-
| rowspan="1"|2013/06/10 || || || ||
|-
| rowspan="1"|2013/06/17 ||范淼 || Relation Extraction ||[[媒体文件:130617-relation_extraction-fm.pdf|slides]] ||
|-
| rowspan="1"|2013/06/24 ||唐国瑜 || Incorporating Statistical Word Senses in Topic Model ||[[媒体文件:130624_Incorporating Statistical Word Senses in Topic Model_TGY.pdf|slides]] ||
|-
| rowspan="1"|2013/07/01 || || || ||
|-
| rowspan="1"|2013/07/08 || || || ||
|-
| rowspan="1"|2013/07/15 || || || ||
|-
| rowspan="1"|2013/09/09 ||王东 || Research Frontier in Speech Technology||[[媒体文件:Research Frontier in Speech Technology.pdf|slides]] ||
|-
| rowspan="1"|2013/09/16 || || || ||
|-
| rowspan="1"|2013/09/23 || || || ||
|-
| rowspan="1"|2013/09/30 || || || ||
|-
| rowspan="1"|2013/10/07 || || || ||
|-
| rowspan="1"|2013/10/14 || || || ||
|-
| rowspan="1"|2013/10/21 ||范淼 ||Transduction Classification with Matrix Completion （中文报告）||[[媒体文件: Transduction_Classifiction_with_Matrix_Completion.pdf‎|slides]] [http://pages.cs.wisc.edu/~jerryzhu/pub/mc4ssl_FINAL.pdf paper]|| 李蓝天
|-
| rowspan="1"|2013/10/28 || || || ||
|-
| rowspan="1"|2013/11/04 || 王军 || 基于i-vector的intersession补偿及打分方法(综述) || [[媒体文件:131104-ivecto下intersession补偿及打分方法--01-WJ-.pdf‎|slides]]||
|-
| rowspan="1"|2013/11/11 ||张陈昊 ||PLDA介绍及PLDA在说话人识别中的应用 ||[[媒体文件:PLDA.pdf|slides]] || 唐国瑜
|-
| rowspan="1"|2013/11/18 ||别凡虎 ||i-vector理论介绍（讨论）||[[媒体文件:131118-i-vector_and_GMM-UBM-BFH.pdf|slides]]‎ ||王军
|-
| rowspan="1"|2013/11/25 ||刘超 || Pruning Neural Networks By Optimal Brain Damage(综述)||[[媒体文件:131125-OBD-LC-01.pdf|slides]] ||范淼
|-
| rowspan="1"|2013/12/02 ||范淼 ||Distant Supervision for Relation Extraction with Matrix Completion （英文报告）||[[媒体文件:131202-DRMC-FM-01.pdf|slides]] || 李蓝天
|-
| rowspan="1"|2013/12/09 || Dong Wang|| Introduction to the HMM-based speech synthesis||[http://hts.sp.nitech.ac.jp/archives/2.2/HTS_Slides.zip slides] ||
|-
| rowspan="1"|2013/12/16 ||张陈昊 ||语音研究中的基元介绍 ||[[媒体文件:131215-Phonology-ZCH.pdf|slides]] ||
|-
| rowspan="1"|2013/12/23 || Dong Wang|| Introduction to the HMM-based speech synthesis (2)||[http://hts.sp.nitech.ac.jp/archives/2.2/HTS_Slides.zip slides] ||
|-
| rowspan="1"|2013/12/23 || || || ||
|-
| rowspan="1"|2013/12/30 ||刘荣 || continuous space language model||[[媒体文件:Cslm-cslt.pdf|slides]] ||刘超
|-
| rowspan="1"|2014/01/06 || || || ||
|-
| rowspan="1"|2014/01/13 || || || ||
|-
| rowspan="1"|2014/01/20 || || || ||
|-
| rowspan="1"|2014/02/24 || || || ||
|-
| rowspan="1"|2014/03/03 || || || ||
|-
| rowspan="1"|2014/03/10 ||范淼|| Distant Supervision for Information Extraction (英文报告)|| || 李蓝天
|-
| rowspan="1"|2014/03/17 ||唐国瑜 || Topic Models Incorporating Statistical Word Senses || [[媒体文件:TMISWS_For_CICLing2014.pdf|slides]]||
|-
| rowspan="1"|2014/03/24 ||孟祥涛 || Noisy training for Deep Neural Networks|| ||
|-
| rowspan="1"|2014/03/31 ||范淼|| Translating Embeddings for Modeling Multi-relational Data （中文报告） || [https://www.hds.utc.fr/everest/lib/exe/fetch.php?id=en%3Atranse&cache=cache&media=en:cr_paper_nips13.pdf paper]||李蓝天
|-
| rowspan="1"|2014/04/07 || || || ||
|-
| rowspan="1"|2014/04/14 || Wang Jun|| I-vector and PLDA in depth ||[[媒体文件:131104-ivector-microsoft-wj.pdf|slides]] ||
|-
| rowspan="1"|2014/04/21 || 邱晗||汉语事件句式规范化处理 ||[[媒体文件:140421-汉语事件句式规范化-QH.pdf‎|slides]] ||
|-
| rowspan="1"|2014/04/28 || 唐国瑜|| Some papers in　CICLing2014 ||[[媒体文件:Some_papers_in_CICling2014.pdf|slides]] ||刘超
|-
| rowspan="1"|2014/05/05 || || || ||
|-
| rowspan="1"|2014/05/12 || 卡尔|| paper introduction || [[媒体文件:Acoustic Factor Analysis.pdf|slides]] || 邱晗
|-
| rowspan="2"|2014/05/19 || 邱晗|| 汉语事件句式CCG推导树重构 ||[[媒体文件:140519-CCG_reConstruction.pdf‎|slides]]‎|| 卡尔
|-
|Liu Chao|| master proposal: sparse and deep neural networks || [[媒体文件:140519-proposal-LC-01.pdf|slides]] ||
|-
| rowspan="1"| || Liu Chao|| 2nd master proposal: sparse and deep neural networks|| ||
|-
| rowspan="1"|2014/06/16 || 别凡虎 || Truncated Wave based VPR and Some Recent Work || [[媒体文件:140614-Truncated_Speech_based_VPR.pdf‎|slides]]‎ || 别凡虎
|-
| rowspan="1"|2014/06/23 || 别凡虎 || Block-wise training for I-vector || [[媒体文件:140623-Block-wise training for I-vector.pdf‎|slides]]‎ || 别凡虎
|-
| rowspan="1"| 2014/07/07||王军 ||Discriminative Scoring for Speaker Recognition Based on I-vectors || [[媒体文件:140707-work_report.pdf|slides]]|| 王军
|-
| rowspan="1"| 2014/09/01|| || || ||
|-
| rowspan="1"|2014/09/09 ||别凡虎 ||Reseach on Truncated Wave based VPR||[[媒体文件:140909-Truncated Speech based VPR.pdf|slides]] || 别凡虎
|-
| rowspan="1"| 2014/09/15|| || || ||
|-
| rowspan="1"|2014/09/22 || Miao Fan|| Large-scale Entity Relation Extraction based on Low-dimensional Representations (中文报告，博士开题)
||[[媒体文件:基于低维表示的大规模实体关系挖掘技术.pdf‎|slides]] || Lan TianLi
|-
| rowspan="1"| 2014/09/29 || || || ||
|-
| rowspan="1"|2014/10/13 || Miao Fan|| The Frontier of Knowledge Embedding （英文报告）|| [[媒体文件:The_Frontier_of_Knowledge_Embedding.pdf‎|slides]]|| Lan TianLi
|-
| rowspan="1"|2014/10/20 || || || ||
|-
| rowspan="1"|2014/10/27 || Li Yi || Phonemes, Features, and Syllables: Converting Onset and Rime Inventories to Consonants and Vowels||[[媒体文件:Lanzhou Phonemes, Features, and Syllables- fianl.pdf|paper]] [[媒体文件:Syllables and phonemes - 20141027.pdf|slides]]||
|-
| rowspan="1"|2014/11/3 || 米吉提|| Automatic Speech Recognition of Agglutinative Language based on Lexicon Optimization||[[媒体文件:Mijit-slides-清华大学-2014-11-3.pdf|slides]] ||
|-
| rowspan="1"|2014/11/10 || || || ||
|-
| rowspan="1"|2014/11/17 ||Dong Wang || Highly restricted keyword spotting for Uyghur using sparse analysis|| [[媒体文件:Highly Restricted Keyword Selection Based on Sparse Analysis.pdf|slides]]||
|-
| rowspan="1"|2014/11/24 || || || ||
|-
| rowspan="1"|2014/12/1 ||ZhongDa Xie ||Incorporating Fine-Grained Ontological Relations in Medical Document Ranking || [[媒体文件:Fine-grained_relations.pdf|slides]]|| Lantian Li
|-
| rowspan="1"|2014/12/8 || || || ||
|-
| rowspan="1"|2014/12/15 || 唐国瑜 || 跨语言话题分析关键技术研究 ||[[媒体文件:141205-答辩-TGY.pdf|slides]] ||
|-
| rowspan="1"|2014/12/22 || || || ||
|-
| rowspan="1"|2014/12/29 || Askar || Language Mismatch in Speaker Recognition System||[[媒体文件:141229--askar.pdf|slides]] ||
|-
| rowspan="1"|2015/1/5 ||Lantian Li || Deep Neural Networks for Speaker Recognition || [[媒体文件:150104_Deep_Neural_Networks_for_Speaker_Recognition_LLT.pdf|slides]]||
|-
| rowspan="1"|2015/1/12 || || || ||
|-
| rowspan="1"|2015/1/19 || Dong Wang || Machine Learning Paradigms for Speech Recognition||[[媒体文件:Machine Learning Paradigms for Speech Recognition.pdf|slides]] [http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6423821 paper] ||
|-
| rowspan="1"|2015/1/26 || Chen Guorong || Information Transmission and Distribution on Web ||[[媒体文件:An_introduction_of_complex_network1.pdf|slides]] ||
|-
| rowspan="1" |2015/3/9 || Dong Wang || Joint Deep Learning || [[媒体文件:Joint Deep Learning.pdf|slides]] ||
|-
| rowspan="1"|2015/3/16 || Dongxu Zhang || Knowledge learning from text data and knowledge bases || [[媒体文件:Joint Deep Learning.pdf|slides]] ||
|-
| rowspan="1"|2015/4/13 || Xuewei Zhang || Lasso-based Reverberation Suppression In Automatic Speech Recognition || [[媒体文件:Lasso-based Reverberation Suppression In Automatic Speech Recognition.pdf|slides]] ||
|-
| rowspan="1"|2015/5/11 || Dong Wang ||ASR and SID Research Frontier ||[[媒体文件:ASR and SID Research Frontier.pdf|slides]] ||
|-
| rowspan="1"|2015/11/23 || Zhiyuan Tang|| CTC learning|| [[媒体文件:CTC.pdf|slides]] ||
|-
| rowspan="1"|2015/11/30 || Mengyuan Zhao|| CNN-based music removal|| [[媒体文件:Music Removal by Convolutional Denoising.pdf | slides]] ||
|-
| rowspan="1"|2015/12/3 || Zhiyuan Tang|| Networks of Memory|| [[媒体文件:Memory_net.pdf|slides]] ||
|-
| rowspan="1"|2015/12/7 || Yiqiao Pan|| Document Classification with Spherical Word Vectors||[[媒体文件:Document Classification with Spherical Word Vectors.pdf|slides]] ||
|-
| rowspan="1"|2015/12/14 || Dong Wang || Transfer Learning for Speech and Language Processing ||[[媒体文件:Transfer_Learning_for_Speech_and_Language_Processing.pdf|slides]] ||
|-
| rowspan="1"|2015/12/21 || Qixin Wang || Attention for poem generation ||[[媒体文件:Ijcai 2016.pptx|slides]] ||
|-
| rowspan="1"|2015/12/28 || Lantian Li || Max-margin metric learning for speaker recognition || [[媒体文件:Max-margin-Metric-Learning.pdf|slides]]||
|-
| rowspan="1"|2016/1/4 || Zhiyong Zhang || Parallel training,MPE and natural gradient||[[媒体文件:20160104_张之勇_Large-scale Parallel Training in Speech Recognition.pdf|slides]]||
|-
| rowspan="1"|2016/1/18 || Dongxu Zhang || Memoryless Document Vector ||[[媒体文件:Memoryless_document_vector.pdf|slides]]||
|-
| rowspan="1"|2016/3/14 || Zhiyuan Tang|| Oral presentation for "vMF-SNE: Embedding for Spherical Data"|| [[媒体文件:embedding.pdf|slides]] ||
|-
| rowspan="1"|2016/3/28 || Tianyi Luo || Review for Neural QA || [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/2/29/CSLT_Weekly_Report--20160328.pdf slides] ||
|-
| rowspan="1"|2016/4/11 || Rong Liu || Recommendation in Youku || [http://cslt.riit.tsinghua.edu.cn/mediawiki/index.php/%E6%96%87%E4%BB%B6:Cslt%E5%AE%9E%E9%AA%8C%E5%AE%A4%E4%BA%A4%E6%B5%81.pptx slides] ||
|-
| rowspan="1"|2016/5/09 || Miao Fan || Learning contextual embeddings of knowledge base with entity descriptions.|| [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/9/9c/Techreport_CSLT_2016_M.F..pdf slides] ||
|-
| rowspan="1"|2016/5/16 || Yang Wang || Research on conversation thread detection. || [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/b/bb/%E6%B1%AA%E6%B4%8B-%E6%AF%95%E8%AE%BE-CSLT.pdf slides] ||
|-
| rowspan="1"|2016/5/20 || Yang Wang & Maoning Wang || Research on portfolio selection. || [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/8/89/%E6%B1%AA%E6%B4%8B-%E9%87%91%E8%9E%8D%E7%AC%AC%E4%B8%80%E6%AC%A1%E5%88%86%E4%BA%AB.pdf slides1] [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/b/bb/%E6%B1%87%E6%8A%A5_%E8%B5%84%E4%BA%A7%E7%BB%84%E5%90%88%E4%B8%AD%E5%87%A0%E4%B8%AA%E8%AF%84%E4%BB%B7%E6%8C%87%E6%A0%87%E7%9A%84%E8%A7%A3%E9%87%8A.pdf slides2]||
|-
| rowspan="1"|2016/5/20 || Zhiyuan Tang || ICASSP 2016 summary || [[媒体文件:Note icassp16.pdf|slides]] ||
|-
| rowspan="1"|2016/5/23 || Dong Wang || graphical model and neural model || [[媒体文件:Graphic Model and Neural Model.pdf|slides]] [[媒体文件:Generative-Pdf.rar|papers]] ||
|-
| rowspan="1"|2016/8/02 || Zhiyuan Tang || Visualizing, Measuring and Understanding Neural Networks: A Brief Survey|| [[媒体文件:Nn analysis.pdf|slides]] ||
|-
| rowspan="1"|2016/8/03 || Yang Wang || Neural networks and genetic programming for financial forecasting || [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/7/79/GeneticNN.pdf slides] ||
|-
| rowspan="1"|2016/11/05 || Yang Wang || Reinforcement Learning Models and Simulations || [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/c/ca/RRL_and_sim.pdf slides] ||
|-
| rowspan="1"|2016/11/08 || April Pu || SOFTWARE DEVELIPMENT METHODOLOGIES || [http://wangd.cslt.org/talks/pdf/april_software.pptx slides] ||
|-
| rowspan="1"|2016/11/12 || Yang Wang || Generative Adversarial Nets || [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/c/c9/Generative_adversarial_network.pdf slides] ||
|-
| rowspan="1"|2016/11/22 || Zhiyuan Tang || INTERSPEECH 2016 summary || [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/6/65/Interspeech16_review.pdf slides] ||
|-
| rowspan="1"|2016/11/30 || Dong Wang || Deep and sparse learning in speech and language: an overview || [http://wangd.cslt.org/talks/pdf/bics2016.pptx slides] ||
|-
| rowspan="1"|2017/2/17 || Yang Wang || Review understanding deep learning requires rethinking generalization || [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/3/3b/Review_understanding_deep_learning_requires_rethinking_generalization.pdf slides] ||
|-
| rowspan="1"|2017/6/5 || Dong Wang || Deep speech factorization || [http://wangd.cslt.org/talks/pdf/Deep-Speech-Factorization.pdf slides] ||
|-
| rowspan="1"|2017/6/8 || Shiyue Zhang || Convolutional Sequence to Sequence Learning || [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/f/f3/Conv_seq2seq.pptx slides] ||
|-
| rowspan="1"|2017/6/12 || Shiyue Zhang || Memory-augmented Neural Machine Translation || [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/3/36/Memory-augmented_Neural_Machine_Translation_.pptx slides] ||
|-
| rowspan="1"|2017/6/21 || Shiyue Zhang || Attention Is All You Need || [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/6/68/Attention_is_all_you_need.pptx slides] ||
|-
| rowspan="1"|2017/6/26 || Jiyuan Zhang || Chinese poem generation using neural model || [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/5/50/Flexible_and_Creative_Chinese_Poetry_Generation_Using_Neural_Memory_.pptx slides] ||
|-
|}

文件:Flexible and Creative Chinese Poetry Generation Using Neural Memory .pptx

2017-06-26T09:36:31Z

Zhangjy：

文件:Memory-augmented Neural Machine Translation .pptx

2017-06-25T13:38:27Z

Zhangjy：Zhangjy上传“文件:Memory-augmented Neural Machine Translation .pptx”的新版本

Memory-augmented_Neural_Machine_Translation report

文件:Memory-augmented Neural Machine Translation .pptx

2017-06-25T08:05:35Z

Zhangjy：Zhangjy上传“文件:Memory-augmented Neural Machine Translation .pptx”的新版本

Memory-augmented_Neural_Machine_Translation report

文件:Nlp team bi monthly report.pdf

2017-05-09T04:52:29Z

Zhangjy：Zhangjy上传“文件:Nlp team bi monthly report.pdf”的新版本

Bi-monthly-2017-04-language

2017-05-09T04:39:23Z

Zhangjy：

Team [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/a/a8/Nlp_team_bi_monthly_report.pdf]

Jiyuan Zhang [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/e/e4/Bi-monthly_report_zhangjy.pdf]

Shiyue Zhang

Aodong Li

Bi-monthly-2017-04-language

2017-05-09T04:39:04Z

Zhangjy：

Team [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/a/a8/Nlp_team_bi_monthly_report.pdf]

Jiyuan Zhang http://cslt.riit.tsinghua.edu.cn/mediawiki/images/e/e4/Bi-monthly_report_zhangjy.pdf

Shiyue Zhang

Aodong Li

文件:Bi-monthly report zhangjy.pdf

2017-05-09T04:38:38Z

Zhangjy：

Bi-monthly-2017-04-language

2017-05-09T04:38:21Z

Zhangjy：

Team [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/a/a8/Nlp_team_bi_monthly_report.pdf]

Jiyuan Zhang

Shiyue Zhang

Aodong Li

文件:Nlp team bi monthly report.pdf

2017-05-09T04:37:48Z

Zhangjy：

Bi-monthly-2017-04-language

2017-05-09T04:37:26Z

Zhangjy：

Team

Jiyuan Zhang

Shiyue Zhang

Aodong Li

Bi-monthly-2017-04-language

2017-05-09T04:36:59Z

Zhangjy：以“Team Jiyuan Zhang Shiyue Zhang Aodong Li”为内容创建页面

Team

Jiyuan Zhang

Shiyue Zhang

Aodong Li

NLP Status Report 2017-4-24

2017-05-03T02:47:34Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/4/5
|Yang Feng ||

||
|-
|Jiyuan Zhang ||
*run the qx's model using different dataset
*read some papers
||
|-
|Andi Zhang ||

||
|-
|Shiyue Zhang ||
||
|-
|Peilun Xiao ||

||

|}

NLP Status Report 2017-4-24

2017-05-03T02:47:19Z

Zhangjy：以“{| class="wikitable" !Date !! People !! Last Week !! This Week |- | rowspan="6"|2017/4/5 |Yang Feng || || |- |Jiyuan Zhang || *run the qx's model using different d...”为内容创建页面

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/4/5
|Yang Feng ||

||
|-
|Jiyuan Zhang ||
*run the qx's model using different dataset
*read some paper
||
|-
|Andi Zhang ||

||
|-
|Shiyue Zhang ||
||
|-
|Peilun Xiao ||

||

|}

NLP Status Report 2017-4-17

2017-05-03T02:45:05Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/4/5
|Yang Feng ||

||
|-
|Jiyuan Zhang ||
*run the ppg model using different datasets
*check the emnlp paper
||
|-
|Andi Zhang ||

||
|-
|Shiyue Zhang ||
||
|-
|Peilun Xiao ||

||

|}

NLP Status Report 2017-4-17

2017-05-03T02:44:25Z

Zhangjy：以“{| class="wikitable" !Date !! People !! Last Week !! This Week |- | rowspan="6"|2017/4/5 |Yang Feng || * Got the sampled 100w good data and ran Moses (BLEU: 30.6) *...”为内容创建页面

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/4/5
|Yang Feng ||
* Got the sampled 100w good data and ran Moses (BLEU: 30.6)
* Reimplemented the idea of ACL (added some optimization to the previous code) and check the performance in the following gradual steps: 1. use s_i-1 as memory query; 2. use s_i-1+c_i as memory query; 3. use y as the memory states for attention; 4. use y + smt_attentions * h as memory states for attention.
* ran experiments for the above steps but the loss was inf. I am looking for reasons.
||
*do experiments and write the paper
|-
|Jiyuan Zhang ||
*run the ppg model using different datasets
*check the emnlp paper
||
*improve the effect of the qx's model
|-
|Andi Zhang ||
*revise the original oov model so that it can automatically detect oov words and translate them
*deal with the situation that source word is oov but target word is not oov first
*it didn't predict right
||
*make the model work as what we wanted
*deal with the situation that source word is oov and target word is also oov, then other situations
|-
|Shiyue Zhang ||
* got a reasonable baseline on big zhen data
||
* implement mem model on this baseline, and test on big data
|-
|Peilun Xiao ||

||

|}

NLP Status Report 2017-4-10

2017-05-03T02:41:45Z

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/4/5
|Yang Feng ||
* Got the sampled 100w good data and ran Moses (BLEU: 30.6)
* Reimplemented the idea of ACL (added some optimization to the previous code) and check the performance in the following gradual steps: 1. use s_i-1 as memory query; 2. use s_i-1+c_i as memory query; 3. use y as the memory states for attention; 4. use y + smt_attentions * h as memory states for attention.
* ran experiments for the above steps but the loss was inf. I am looking for reasons.
||
*do experiments and write the paper
|-
|Jiyuan Zhang ||
*convert the style of the paper to EMNLP
*contact the ppg's author to get the code
||
*improve the effect of the qx's model
|-
|Andi Zhang ||
*revise the original oov model so that it can automatically detect oov words and translate them
*deal with the situation that source word is oov but target word is not oov first
*it didn't predict right
||
*make the model work as what we wanted
*deal with the situation that source word is oov and target word is also oov, then other situations
|-
|Shiyue Zhang ||
* got a reasonable baseline on big zhen data
||
* implement mem model on this baseline, and test on big data
|-
|Peilun Xiao ||

||

|}

NLP Status Report 2017-4-5

2017-04-05T02:04:09Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/3/27
|Yang Feng ||
*tested for the baseline but cannot get the reasonable result.
*debug the baseline to try to reproduce the good result but failed.
*fixed the problem of nan in alpha-gamma method but the result is not good.
*changed the calculation of probability for alpha-gamma method but the result is neither good.
*ran Moses for cwmt zh-en translation, but the training data is case-sensitive, so need to rerun.
||
*rerun Moses for cwmt zh-en and cs-en
*decide to use tensorflow or theano
*run experiments based on the chosen platform
|-
|Jiyuan Zhang ||
* I did keyword expansion on the qx's model
*fixed some bugs
*read two papers
||
*improve the effect of the qx's model
|-
|Andi Zhang ||
*fixed the bug, turns out it rises from the unfamiliarity with numpy.resize() function
*the demo model can deal with oov problem(both source word and target word are oov)
||
*some paper work about graduation design
*run some experiments using theano on old data set and new zh2en from lihang
|-
|Shiyue Zhang ||
* got a reasonable baseline on big zhen data
||
* implement mem model on this baseline, and test on big data
|-
|Peilun Xiao ||

||

|}

NLP Status Report 2017-3-20

2017-04-05T01:44:29Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/3/20
|Yang Feng ||
*went through the code and made different attempt, and managed to produce the good result (47-->50 on the small zh-en data set)
*wrote the cross-entropy method for alpha-gamma method, but found it is different from the built-in method.
*changed to use the build-in soft-cross-entropy method and ran experiments
||
*get the result for alpha-gamma method
*run experiments on the big data
|-
|Jiyuan Zhang ||
*response ACL reviewers' questions
||
*improve the effect of the qx's model
|-
|Andi Zhang ||
*tried to fix the bug that prevents the prediction of EOS symbol
*the output_projection matrix is out of order after the copying vector process
||
*fix the bug
|-
|Shiyue Zhang ||
* learned to use google's new seq2seq code(gnmt)
* ran gnmt on en-de , small zh-en, cs-en, big zh-en
||
* run gnmt on new big zh-en
* try to find how to implement our model on gnmt
|-
|Peilun Xiao ||

||

|}

NLP Status Report 2017-3-27

2017-04-05T01:43:38Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/3/27
|Yang Feng ||
*tested for the baseline but cannot get the reasonable result.
*debug the baseline to try to reproduce the good result but failed.
*fixed the problem of nan in alpha-gamma method but the result is not good.
*changed the calculation of probability for alpha-gamma method but the result is neither good.
*ran Moses for cwmt zh-en translation, but the training data is case-sensitive, so need to rerun.
||
*rerun Moses for cwmt zh-en and cs-en
*decide to use tensorflow or theano
*run experiments based on the chosen platform
|-
|Jiyuan Zhang ||
* I did nothing (I found my ACL mark was at borderline, so I didn't have a mind to work)
||
*improve the effect of the qx's model
|-
|Andi Zhang ||
*fixed the bug, turns out it rises from the unfamiliarity with numpy.resize() function
*the demo model can deal with oov problem(both source word and target word are oov)
||
*some paper work about graduation design
*run some experiments using theano on old data set and new zh2en from lihang
|-
|Shiyue Zhang ||
* got good results from gnmt
* but haven't found the way to implement our model on gnmt
* trying to modify our code to make it work on big data
||
* go on trying to modify our code to make it work on big data
* go on looking into gnmt code
|-
|Peilun Xiao ||

||

|}

NLP Status Report 2017-3-27

2017-04-05T01:42:51Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/3/27
|Yang Feng ||
*tested for the baseline but cannot get the reasonable result.
*debug the baseline to try to reproduce the good result but failed.
*fixed the problem of nan in alpha-gamma method but the result is not good.
*changed the calculation of probability for alpha-gamma method but the result is neither good.
*ran Moses for cwmt zh-en translation, but the training data is case-sensitive, so need to rerun.
||
*rerun Moses for cwmt zh-en and cs-en
*decide to use tensorflow or theano
*run experiments based on the chosen platform
|-
|Jiyuan Zhang ||
* I did nothing (I found my ACL mark was at borderline, so I didn't have a mind to work)
||
improve the effect of the qx's model
|-
|Andi Zhang ||
*fixed the bug, turns out it rises from the unfamiliarity with numpy.resize() function
*the demo model can deal with oov problem(both source word and target word are oov)
||
*some paper work about graduation design
*run some experiments using theano on old data set and new zh2en from lihang
|-
|Shiyue Zhang ||
* got good results from gnmt
* but haven't found the way to implement our model on gnmt
* trying to modify our code to make it work on big data
||
* go on trying to modify our code to make it work on big data
* go on looking into gnmt code
|-
|Peilun Xiao ||

||

|}

NLP Status Report 2017-3-27

2017-04-05T01:39:52Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/3/27
|Yang Feng ||
*tested for the baseline but cannot get the reasonable result.
*debug the baseline to try to reproduce the good result but failed.
*fixed the problem of nan in alpha-gamma method but the result is not good.
*changed the calculation of probability for alpha-gamma method but the result is neither good.
*ran Moses for cwmt zh-en translation, but the training data is case-sensitive, so need to rerun.
||
*rerun Moses for cwmt zh-en and cs-en
*decide to use tensorflow or theano
*run experiments based on the chosen platform
|-
|Jiyuan Zhang ||
* I did nothing (I found my ACL mask was at borderline, so I didn't have a mind to work)
||
improve the effect of the qx's model
|-
|Andi Zhang ||

||

|-
|Shiyue Zhang ||
* got good results from gnmt
* but haven't found the way to implement our model on gnmt
* trying to modify our code to make it work on big data
||
* go on trying to modify our code to make it work on big data
* go on looking into gnmt code
|-
|Peilun Xiao ||

||

|}

NLP Status Report 2017-3-20

2017-04-05T01:34:02Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/3/20
|Yang Feng ||
*went through the code and made different attempt, and managed to produce the good result (47-->50 on the small zh-en data set)
*wrote the cross-entropy method for alpha-gamma method, but found it is different from the built-in method.
*changed to use the build-in soft-cross-entropy method and ran experiments
||
*get the result for alpha-gamma method
*run experiments on the big data
|-
|Jiyuan Zhang ||
*response ACL reviewers' suggestions
||
*improve the effect of the qx's model
|-
|Andi Zhang ||
*tried to fix the bug that prevents the prediction of EOS symbol
*the output_projection matrix is out of order after the copying vector process
||
*fix the bug
|-
|Shiyue Zhang ||
* learned to use google's new seq2seq code(gnmt)
* ran gnmt on en-de , small zh-en, cs-en, big zh-en
||
* run gnmt on new big zh-en
* try to find how to implement our model on gnmt
|-
|Peilun Xiao ||

||

|}

NLP Status Report 2017-3-20

2017-04-05T01:31:41Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/3/20
|Yang Feng ||
*went through the code and made different attempt, and managed to produce the good result (47-->50 on the small zh-en data set)
*wrote the cross-entropy method for alpha-gamma method, but found it is different from the built-in method.
*changed to use the build-in soft-cross-entropy method and ran experiments
||
*get the result for alpha-gamma method
*run experiments on the big data
|-
|Jiyuan Zhang ||
*response ACL reviewers' suggestions
||

|-
|Andi Zhang ||
*tried to fix the bug that prevents the prediction of EOS symbol
*the output_projection matrix is out of order after the copying vector process
||
*fix the bug
|-
|Shiyue Zhang ||
* learned to use google's new seq2seq code(gnmt)
* ran gnmt on en-de , small zh-en, cs-en, big zh-en
||
* run gnmt on new big zh-en
* try to find how to implement our model on gnmt
|-
|Peilun Xiao ||

||

|}

NLP Status Report 2017-3-13

2017-03-13T07:10:10Z

Zhangjy：

{| class="wikitable"
!Date !! People !! Last Week !! This Week
|-
| rowspan="6"|2017/1/3
|Yang Feng ||
* tested and analyzed the results on the cs-en data set (30.4 on the heldout-training set and 7.3 on the dev set);
* added masks to the baseline (44.4 on the cn-en);
* added encoder-masks and memory-masks to alpha-gamma method and fixed the bugs. Got an improvement of 0.5 again the masked baseline [[http://cslt.riit.tsinghua.edu.cn/mediawiki/images/b/b8/Nmt_mn_report_continue.pdf report]];
* To avoid doing softmax twice, rewrite the softmax_cross_entropy function myself. (under-training)
||
* analyze and improve the alpha-gamma method.
|-
|Jiyuan Zhang ||
*completed to reproduce planning neural network
*chose best attention_memory model for huilian and ran big train dataset(about 370k) [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/b/b9/Model_with_different_dataset.pdf result]

||
*Keyword expansion model
*collect more poem from Internet
*recruiting

|-
|Andi Zhang ||

||

|-
|Shiyue Zhang ||

||

|-
|Peilun Xiao ||

||

|}