Write a topic evaluation. about anlp

[Samualkrish]

As per my conversation with the TA. The project that we have selected are good but there are certain things we need to elaborate more.

Listing down the order in which the TA and Professor need the abstract.

1. Explain the problem in a very good detail

• what is the current issue you are addressing
• what will be the outcome if your solution works
• why is the problem different in case there is an existing solution
• which part of NLP application will this fall into e.x. text classification, machine translation, dialogue systems

2. What is the source language other than English, target language (optional) based on your project.
3. How is your corpora implemented?
   • Are you implementing a manual way to train your vocabulary?
   • Are you using WordNet or any other known vocabulary?
4. How is your data set defined?
   • What is your source for your data set?
   • How is data set modeled?
     • How will your system understand the data set.
     • Example of data set model : ʾaʿadda pos=V,mood=SBJV,gen=FEM,per=1,voice=PASS,num=PL nuʿadda
     • pos - V => Verb
     • mood=SBJV => Subjective
     • gen=FEM => Gender-Feminine
     • per=1 => 1st Person
     • voice=PASS
     • num=PL => Plural
   • Will your data set be sufficient to handle all the vocabulary or will there be some words which might not be there? What will you do in that case?
5. How is the model going to be implemented?
   • What are the features which you would be using ex. words or phrases or sentences.
   • What are the classes (a rough estimate is fine) for classification?
   • What are the morphology, phonemes, stems and nltk packages you would use?
   • What are the steps involved during the phase, detail your phases at each step?
     • Like Part of Speech tagging, what would be the method you will use
     • What stemmer would you use for other language.
     • What are the components in your model
6. What is your evaluation metric?
   • Define a baseline (a minimum accepted solution)
     • Example a literal word to word translation can be used as base.
   • How would you define the upper bound
     • An best solution with good accuracy and F1 score.
     • If your output is nearly equal to a human translated text then your program has passed
     • If your program runs at the baseline, then you will lose points in the project.
7. Related Research and work
   • What are the works that are done about your project?
   • Links to the components that your are going to use?
   • Research papers on how to train your data if available?

Update your topic with these information, and we will post it on blackboard

[summerlight]

Because we don't have much time before the deadline, we're gonna write a proposal.