commit 73270757e4b0f3d76767fe265bc6e0cc75beb4b0
parent ac34ae80c3be3f9f6b3569420478fb6b536c7a19
Author: Antoine Amarilli <a3nm@a3nm.net>
Date: Sun, 11 Aug 2019 19:09:41 +0200
restructure TODO file
Diffstat:
| TODO | | | 50 | ++++++++++++++++++++++---------------------------- |
| test.sh | | | 4 | ++++ |
2 files changed, 26 insertions(+), 28 deletions(-)
diff --git a/TODO b/TODO
@@ -1,42 +1,36 @@
+== Bugs ==
+
./train_diaeresis.sh does not correctly handle "oublieux" despite many
occurrences in additions, what is going on?
-- "paon" en 1 mais "chaos" en 2: utiliser la prononciation de lexique comme
- lower bound ?
-
-- fuzz testing
-- log exceptions separately for plint_web
+== Ideas ==
-- change additions to give numerical syllable count for affected syllable
-- adjust the diaeresis uncertainty threshold to assert that all verses are
+- Compare to the latest lexique: in how many cases is our number of syllables
+ less than the one indicated in lexique? ("paon" en 1 mais "chaos" en 2)
+- The "additions" file should give the number of syllables of each line
+- Adjust the diaeresis uncertainty threshold to assert that all verses are
correct (change your opinion about the syllable with least confidence)
-- document expected errors of test.sh and commit necessary files
+- Document expected errors of test.sh and commit necessary files
- remove kludge for invalid characters, split them in specific chunks
-- improve performance with profiling
-
-== Error reporting ==
+- Improve performance with profiling
+- Only indicate hemistiche status when there is a problem with hemistiches
-- produce error reports first to structured and then to text and json
-- and then from json to html
-- indiquer l'état des hémistiches seulement s'il y a un problème, et seulement
- pour les syllabes plausibles
-- highlight possible hemistiche positions where an hemistiche could have been placed
+== Low priority ==
-== Evaluation ==
+=== Error reporting ===
-- évaluer sur la base de données lexique, sur la qualité de la génération, sur
- le nombre d'erreurs reconnues en cross-validation
+- Produce error reports first in a structured format (JSON) and translate that
+ into text or into HTML
+- When reporting hemistiche errors, highlight possible hemistiche positions
+ where an hemistiche could have been placed
-== Rules ==
+=== Other approaches ===
-- other examples of "mentions", "notions" ? ...
-
-== Other approaches ==
-
-- learn rhyme and gender agnostically by clustering: prepare an undirected graph
+- Learn rhyme and gender agnostically by clustering: prepare an undirected graph
of rhyming verses, factor out suffixes, do SCC, prepare a trie
-== Notes ==
+=== Misc ===
+
+- Fuzz testing: try giving random input to plint and check that it behaves
+- Better exception logging for the Web frontend
-hiatus malherbe:
-http://books.google.de/books?id=qpQGAAAAQAAJ&pg=PA40&lpg=PA40&dq=Et+que+pour+retarder+une+heure+seulement+%22Pour+ne+mourir+jamais%22,+meure+%C3%A9ternellement&source=bl&ots=G8VEFnXkmB&sig=qzmHRiloQpIp6Ebb-9aJrYOoIM0&hl=en&sa=X&ei=3oL8T9WiJ6aB4gTK8pWPBw&redir_esc=y#v=onepage&q=Et%20que%20pour%20retarder%20une%20heure%20seulement%20%22Pour%20ne%20mourir%20jamais%22%2C%20meure%20%C3%A9ternellement&f=false
diff --git a/test.sh b/test.sh
@@ -1,4 +1,8 @@
#!/bin/bash
+
+# about the hiatus error found in malherbe, see:
+# http://books.google.de/books?id=qpQGAAAAQAAJ&pg=PA40&lpg=PA40&dq=Et+que+pour+retarder+une+heure+seulement+%22Pour+ne+mourir+jamais%22,+meure+%C3%A9ternellement&source=bl&ots=G8VEFnXkmB&sig=qzmHRiloQpIp6Ebb-9aJrYOoIM0&hl=en&sa=X&ei=3oL8T9WiJ6aB4gTK8pWPBw&redir_esc=y#v=onepage&q=Et%20que%20pour%20retarder%20une%20heure%20seulement%20%22Pour%20ne%20mourir%20jamais%22%2C%20meure%20%C3%A9ternellement&f=false
+
echo "It is normal that some errors occur when running this script"
for a in test/*.tpl; do
echo "$a"
