The CR module

The most difficult module of my IRCRSystem was the CR module. And just now, I have solve it. I was so excited.

The CR module could add the suffix to each noun phrase. Based on the feature vector auto-extraction module. I had passed it.

Ok. Above all, I could display the final CR result by the txt format.

Tomorrow I could used the MFC to construct a beautiful interface. So terrific news for me. ^-^

This evening we had the final test of our English class. I had done my best.

Ok. Go back now, have a good rest and keep on my work for tomorrow.

The feature vector auto-extraction

Yesterday I had finished the noun phrase recognization task. But I was confounded by the feature vector auto-extraction module. Because the considering cases were too many to clear up them. And at ten o'clock last evening, I decided to process it today.

This morning, before the lab weekly meeting, I began to analyze the feature vector auto-extraction module. After my carefully consideration, I thought there were three big cases. And in each big case there were three little cases. After all the cases working glibly in my brain, I was happy. I could solve it.

This afternoon, I began to code the cases. And just now I had finished this task and tested it. But when I began to design the coreference resolution module I found I must solve the suffix problem.

Ok. The idea was clearly. I could realize it tomorrow morning. But now it's time to do the evening practice: haveing some jogging. Go!

The Noun Phrase Recognization

This was one big feaction of my CR system. Based on the name entity recognization task, I had done this task. The algorithm was same as the one in my CR paper.

But the feature vector extraction module was very complicated. I must clear up my thinking and relize it tomorrow.

Go and try!

Joshua Huang visits our lab

This afternoon, Dr.Joshua Huang visited our lab. He was the Assistant Director of the E-Business Technology Institute. He had many years experience in databases and data mining research in The Commonwealth Science and Industry Research Organization (CISRO), Australia.



After Dr.Tliu gave some introduce about our lab, he gave his introduce about his research and projects. I was interesting in his clustering algotithm based on K-NN. His idea was adding a variable w to restrict the weight of each variables when clustering. In his formula there was a variable beta that was the exponent of w. I found out in his ppt that he chose the beta factitiously. I saw he used 10, -9, -8 and -7 for beta. I asked him whether the beta could be modified heuristic or selfadapted. His answer was they had not done this deeply research.

He also gave us some analysis about the time serials predicting. He used the Multiple Regression Analysis for predicting. But as my experience I thought Multiple Regression Analysis could not give the proper curve trend.

Dr.Joshua Huang was good at data mining and business intelligence. I should study his research method and his paper on clustering.

The English oral test

This afternoon, all of our English students had the English oral test. The test content was based on the dialog of the appointed ten films. We had to pick one film. And our teacher picked stochasticly one snippet with some dialog. After the student watched the snippnet he must iterate the main idea.

I was the first one to have the test. I chose Krammer vs. Krammer. And I was appointed to listen the first snippnet. This snippnet was about Mr.Krammer talked to his boss in an office room. The time was thirty seconds. Frankly speaking, the speed was too fast. But as I had listened it many many times, I iterated it successfully. But I had one wrong pronunciation and with a little Chinese snippnet. I got 90%.

There was another exciting news that we could take part in the first mathemathcal modeling contest for graduate at Sep. 17. Just now I had trooped two of my best friends. We will struggle for it.

May us good luck!!

Send my paper

This afternoon, I sent my paper Decision tree-based Chinese Noun Phrase Coreference Resolurion to the SWCL2004 webpage.

After the paper sent out I could finish next task that was building a coreference resolution system for any free text. This afternoon, I had built the preprocessing fraction including sentences dividing, word segment, pos tagging and name entity recognization. Then I would add my noun phrase recognization module into it. And finally I could separate the coreference resolution task as two parts: pronoun coreference resolution and general noun phrase coreference resolution. Based on the decision trees that of my finished tasks I could resolove them quickly.

Now the left tasks were only of programming. I should finish them quickly. Because the next task was writing my graduating design paper.

Let me try my best!

Paper's writing skill

Writing paper needs skill. I think so.

This morning, I invited Dr.Tliu, Mrs.Qinb, Carl and Lee as a committee to discuss my paper on Decision trees-based Chinese noun phrase coreference resolution. Before this committee, I had sent my paper to Mrs.Qin and Mr.Lu and got lots of useful suggestions. This morning we would discuss my third edition paper.

Dr.Tliu emphasized I must pay more attention on the basic concept and the title. Mrs.Qin gave me some advice about how to reduce the tables. Carl thought I could add some detailed information about the test result and the adventage about decision trees used for coreference resolution. Lee suggested me to modify the authors names in English.

Yes, there were some problems in my paper. And just now, I had modified them in my opinion.

There was somebody said that you would learn more after you modified some paper once. And I had felt it now. Thanks to the teachers and the studying brothers.