197
edits
Changes
Jump to navigation
Jump to search
*This tuning had been fruitless. I stopped the code.*Looked into a word2vec representation of the input
no edit summary
[[Minh Le]] [[Work Logs]] [[Minh Le (Work Log)|(log page)]]
2018-08-03:
*For some reason, when we search Cappital Innovators, there are more options in the "Tools" section. Need to figure out away around this. Did some quick fix around but nothing permanents.
*Finished crawling, started classifying.
*Finished classifying.
*Pushed the batch to MTurk.
2018-08-02:
*Cleaned up codes
*Published the big MTurk batch.
*Got results after 2 hours.
*Processed the data and trimmed extra columns off.
*Helped Grace with her minor code code
*Helped Maxine with the url classifier
*Improved crawler to take date arguments as per Ed request.
*Ran the crawler again.
2018-08-01:
*Built the SeedDB parser with Maxine and Connor
*Finished getting the data from Seed DB and sent it to Connor.
2018-07-31:
*Talked to Connor and Maxine to figure out SeedDB
*Published the first small batch of MTurk with interjudge reliability (2 workers per HIT) and got good results
*Tested SeedDB server
2018-07-30:
*Finalized the design for MTurk, sent to Ed for thoughts and opinions
*Tried publishing a batch on MTurk using the sandbox, and talked to Connor to test it out together.
2018-07-29:
*Worked on HTML mockup for MTurk
*Crawled Data for the Mturk
2018-07-28:
*Worked on HTML mockup for MTurk
2018-07-27:
*Worked on MTurk
2018-07-26:
*Worked on collecting data with others.
*Skyped Ed, Hira along with others.
2018-07-25:
*Worked with MTurk with Connor
*Talked with Ed about the project progress. We agreed that the RNN can wait, and focus on collecting the data because the data seems much usable now.
*Hand collect data along with fellow interns.
2018-07-24:
*Tried to tweak some more. Still no progress. I might change to word2vec finally?
*Looked into MTurk
2018-07-23:
*The tuning has not been completed yet. However, checking from the results, it seemed that the last 6 parameters did not significantly affect the result?
*This tuning had been fruitless. I stopped the code.
*Looked into using Yang's preprocessing code.
*Maxine was borrowing my crawler for her work and she found a bug in the crawler where the crawler would never take the first result. i think because google updates their web display? Anyway, fixed it.
*Worked on the wiki page
2018-07-20:
dropout_rate_firstlayer\tdropout_rate_secondlayer\trec_dropout_rate_firstlayer\trec_dropout_rate_secondlayer\tembedding_vector_length\tfirstlayer_units\tsecondlayer_units\t"dropout_rate_dropout_layer\tepochs\tbatch_size\tvalidation_split
*Talked to Ed about potentially just do a test run with the RandomForest model because we needed data soon.
2018-07-19:
