<?xml version="1.0"?>
<feed xmlns="http://www.w3.org/2005/Atom" xml:lang="en">
	<id>https://wiki.apertium.org/w/index.php?action=history&amp;feed=atom&amp;title=Ideas_for_Google_Summer_of_Code%2FBackpropagation</id>
	<title>Ideas for Google Summer of Code/Backpropagation - Revision history</title>
	<link rel="self" type="application/atom+xml" href="https://wiki.apertium.org/w/index.php?action=history&amp;feed=atom&amp;title=Ideas_for_Google_Summer_of_Code%2FBackpropagation"/>
	<link rel="alternate" type="text/html" href="https://wiki.apertium.org/w/index.php?title=Ideas_for_Google_Summer_of_Code/Backpropagation&amp;action=history"/>
	<updated>2026-04-16T20:54:33Z</updated>
	<subtitle>Revision history for this page on the wiki</subtitle>
	<generator>MediaWiki 1.34.1</generator>
	<entry>
		<id>https://wiki.apertium.org/w/index.php?title=Ideas_for_Google_Summer_of_Code/Backpropagation&amp;diff=73101&amp;oldid=prev</id>
		<title>Popcorndude: Created page with &quot;This project is kind of like backpropagation but the each layer of the network/pipeline is different and this probably needs to be taken into account.  You could probably get...&quot;</title>
		<link rel="alternate" type="text/html" href="https://wiki.apertium.org/w/index.php?title=Ideas_for_Google_Summer_of_Code/Backpropagation&amp;diff=73101&amp;oldid=prev"/>
		<updated>2021-02-03T14:54:39Z</updated>

		<summary type="html">&lt;p&gt;Created page with &amp;quot;This project is kind of like backpropagation but the each layer of the network/pipeline is different and this probably needs to be taken into account.  You could probably get...&amp;quot;&lt;/p&gt;
&lt;p&gt;&lt;b&gt;New page&lt;/b&gt;&lt;/p&gt;&lt;div&gt;This project is kind of like backpropagation but the each layer of the network/pipeline is different and this probably needs to be taken into account.&lt;br /&gt;
&lt;br /&gt;
You could probably get a lot of the information by stepping through the pipeline in reverse, but there&amp;#039;s also some need to pay attention to what each module does.&lt;br /&gt;
&lt;br /&gt;
Some thoughts:&lt;br /&gt;
&lt;br /&gt;
* For comparison, you&amp;#039;ll need to analyze the reference sentence, if you get unknowns that&amp;#039;s a target monodix issue&lt;br /&gt;
* If you have the right words in the wrong order, that&amp;#039;s probably transfer&lt;br /&gt;
* If the correct word was present after bidix but not after lexical selection, that&amp;#039;s the issue&lt;br /&gt;
* Similarly for morphological disambiguation&lt;br /&gt;
* A gender issue on pronouns or subject agreement on verbs might be anaphora (might have to restrict this logic to only pairs that use apertium-anaphora and words with anaphora attached to them)&lt;br /&gt;
* If it&amp;#039;s structurally correct but the wrong lemma, probably bidix&lt;br /&gt;
* Unknown word mark? Analyzer&lt;br /&gt;
&lt;br /&gt;
Also, problems may have more than one plausible solution!&lt;br /&gt;
&lt;br /&gt;
== Coding Challenge ==&lt;br /&gt;
&lt;br /&gt;
Write a script that takes a bilingual corpus and runs it through the full pipeline and just up to bidix and identify words with lexical selection errors (that is, the desired lemma+POS was in the bidix output but not the final output) and print them out with a 3 words of context on either side.&lt;/div&gt;</summary>
		<author><name>Popcorndude</name></author>
		
	</entry>
</feed>