Browse Source

Push some notes

Tyler Hallada 7 years ago
parent
commit
da13d51028
1 changed files with 25 additions and 0 deletions
  1. 25 0
      notes.md

+ 25 - 0
notes.md

@@ -11,3 +11,28 @@ functions. I can think of these off the top of my head:
11 11
 * Assembler loop - takes structure above and outputs one word
12 12
     - Maybe should wrap in a sentence loop, line-by-line loop, paragraph loop,
13 13
       etc.
14
+
15
+Syntax aware generate is actually pretty bad. I think it forces it to be too
16
+random. The POS tagging is too error prone and fine-detailed.
17
+
18
+Ideas for the future:
19
+
20
+Pick one or two lines of the haiku from actual haiku or other poems. Then add a
21
+line or two from the corpus (e.g. trump tweets) that both fits the syllables and
22
+rhymes with the end(s) of the real poetic line. I think both sources could be
23
+ngram generated, but I think it would be ideal if they were picked wholesale
24
+from the source. The problem with that approach is that you'd also have to find
25
+a common word between the two source extractions so that the sentence doesn't
26
+abruptly shift between lines. Or, maybe that's a good thing? I guess I should
27
+try both.
28
+
29
+Maybe try just switching out the nouns, verbs, adjectives, and adverbs leaving
30
+the rest of the sentence structure largely intact after the tree replace?
31
+
32
+Use word similarity vectors to construct a sentence (or poem) around a central
33
+theme. E.g. construct something like the buffalo sentence: syntactically correct
34
+sentences, and maybe even semantically meaningful sentences, but in a totally
35
+novel form because of some arbitrary restriction (can only use the word
36
+"buffalo", or animal words, or onomatopoeias, or etc.).
37
+
38
+Integrate alliteration and rhyming.