sayankotor / cqas Goto Github PK

https://stackoverflow.com/questions/2165716/spaces-and-new-lines-with-html-and-css

lksjfklsdjf

this because is to be fixed

Currently some of the template output can be made more readable and coherent:

Now:

It's simple! tea is better, because: better for you, easier, better for the environment, safer.. But you should know that coffee is: greater, easier to keep your weight under control with green tea as part of your diet, caffeine, cooler.

Should be:

It's simple! Tea is better, because: better for you, better for the environment, and ** safer.** But you should know that coffee is: greater, easier to keep your weight under control with green tea as part of your diet, and cooler.
*

Below i list all fixes in the sequential order (each of them is marked using boldface):

• Capitalize the first letter in a sentence
• Include and before the last element in the sentence
• You have one extra comma.
• You shall strictly keep at most three aspect per object. How to select which to keep: (i) first pick the longest ones as they tend to contain phrases of text with complete argument, (ii) use the provided aspect score.

Use these questions as pre-filled suggestions: (randonly generate)

https://docs.google.com/spreadsheets/d/1vNSvUgZAZgM950mbS7W6axBqdO5NCTaE9pIT_wAnNtE/edit#gid=518798415

1. CTRL (100)
2. Snippets
3. Bullet

RoBERTa yields much better results thus should be showcased in the demo.

The output currently is generated in the input text field. This is wrong.

The user should not be able to edit output - it would be like editing Google outputs yourself. No reason to do it. Change it to simple

tag or something like a table.

The output below is a good example - just use similar output formatting (not input field):

Add next to Github and API link leading to something like this:

http://ltdemos.informatik.uni-hamburg.de/arg-api

Flask-Swagger for documentation.

Currently it seems that TextRank is performing summarization of ALL sentences which support both opinions. This doesn't make any sense as sentences with opposite opinion are mixed up like the first one and the highligted one:

What to do:

1. Apply TextRank twice: once for the bunch of sentence supporting object 1 and the second time for the bunch of sentences supporting object 2.

2. Combine the sentences without any newlines inside each of these two groups.

3. Add a paragraph break between these two summaries.

to the nlp.zhores.net/cqas