m. polinsky commited on
Commit
5356e0c
1 Parent(s): de6f645

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +7 -7
README.md CHANGED
@@ -22,19 +22,19 @@ The method of article selection here is arbitrary. Pre-assigned article tags co
22
  * IR for QA and comprehension
23
  * A cheap and quick way to explore area of research dominated by large, end-to-end trained models like RAG and NewsSum or w.e....TK
24
  * News delivery and access
25
- * CNN provides summaries but there's a huge difference between being served something and being able to "create" my news.
26
- ** Sneaks around headlines...what's in the article? Headlines can push and pull....
27
  * removes control over our attention but enables empowered consumption while keeping news production in the hands of pros.
28
  * Editorial ideation
29
- ** Can be used to find implied but uncovered stories by creating news assemblages without knowing eactly what you'll get. Even though an editor knows what they're currently covering, imagine them writing a sentence describing each article on a piece of paper -- that's not the same as seeing the information in the final articles assembled and juxtaposed like this.
30
  * Cross-article information access
31
- ** Information that's related and that paints a picture can be broken across multiple articles from different times...There are more stories lying latent in the told stories.
32
  * Whole news article summarization pitfalls and windfalls.
33
- ** Doing whole articles...technique and results.
34
  * Community pantry principle
35
- ** No free lunch but there is a community pantry. It only gets you so close.
36
  * Evaluating summarization
37
- ** Difficult to objectively evaluate summarization capability beyond a general level.
38
 
39
 
40
  This application was created as the culmination of a semester of independent graduate research into NLP and transformers.
 
22
  * IR for QA and comprehension
23
  * A cheap and quick way to explore area of research dominated by large, end-to-end trained models like RAG and NewsSum or w.e....TK
24
  * News delivery and access
25
+ * CNN provides summaries but there's a huge difference between being served something and being able to "create" my news.
26
+ * Sneaks around headlines...what's in the article? Headlines can push and pull....
27
  * removes control over our attention but enables empowered consumption while keeping news production in the hands of pros.
28
  * Editorial ideation
29
+ * Can be used to find implied but uncovered stories by creating news assemblages without knowing eactly what you'll get. Even though an editor knows what they're currently covering, imagine them writing a sentence describing each article on a piece of paper -- that's not the same as seeing the information in the final articles assembled and juxtaposed like this.
30
  * Cross-article information access
31
+ * Information that's related and that paints a picture can be broken across multiple articles from different times...There are more stories lying latent in the told stories.
32
  * Whole news article summarization pitfalls and windfalls.
33
+ * Doing whole articles...technique and results.
34
  * Community pantry principle
35
+ * No free lunch but there is a community pantry. It only gets you so close.
36
  * Evaluating summarization
37
+ * Difficult to objectively evaluate summarization capability beyond a general level.
38
 
39
 
40
  This application was created as the culmination of a semester of independent graduate research into NLP and transformers.