Select a natural language tool (ex. morphological analyser, parsing, text summarization, speech recognition, translation tools, etc ) which adopts some kind of probabilistic (statistical) or neural network model and evaluate it critically (effffectiveness against real data, good points,weak points, etc) using the data you choose or create. Show the reason why it is chosen or built.
- If quantitative analysis is pursued, you might have to create the correct answer, which is time-consuming.
- Specify the version number of the tool and the evaluation environment also (machine specififications, OS, etc). Use the latest version of the tool.
- Describe the model on which the tool is based brieflfly.
- It would be nice if two or more similar tools are compared.
- Note that just trying a few sentences does not suffiffiffice.
- Show in what part you have spent your time most. (ex. data creation, correct answer generation, analysis of the tool output, etc.)
- Specify used research papers or home pages, if any.
- Write your e-mail address just in case the instructor needs to contact you.
- You can write your report either in Japanese or in English. (I advise Japanese students to write it in Japanese.)
- If you want to pick up a difffferent task, contact the instructor (firstname.lastname@example.org).
We might be able to compromise.
- Any questions should be directed to the mail address above.
Deadline: 2023/01/22 (sun) 23:50
Upload your report in the PDF format to the box of “Advanced course on natural language processing” of the 9th class [report] in keio.jp K-LMS.
(If you should miss the deadline, send your report by e-mail to the instructor.)
Possibly useful URLs:
http://nlp.ist.i.kyoto-u.ac.jp/ (ex.JUMAN; Kyoto Univ.)
http://cl.aist-nara.ac.jp/ (ex. MeCab, CaboCha; NAIST)
EasyDue™ 支持PayPal, AliPay, WechatPay, Taobao等各种付款方式!
E-mail: email@example.com 微信:easydue