Talk:Anaphora resolution - Revision history

Khannatanmai at 17:55, 14 June 2020

2020-06-14T17:55:22Z

Francis Tyers at 12:19, 29 December 2017

2017-12-29T12:19:23Z

Francis Tyers: Created page with "

  that hèctor mentioned on list  ?  i guess it's an issue for sme-nob  oh  it is  we've used some variables and..."

2017-09-14T11:54:22Z

Created page with "<pre> <spectei> that hèctor mentioned on list <Unhammer> ? <spectei> i guess it's an issue for sme-nob <Unhammer> oh <Unhammer> it is <Unhammer> we've used some variables and..."

New page

<pre>
<spectei> that hèctor mentioned on list
<Unhammer> ?
<spectei> i guess it's an issue for sme-nob
<Unhammer> oh
<Unhammer> it is
<Unhammer> we've used some variables and such
<Unhammer> for jorgal they actually do h_n when unsure!
<Unhammer> because it happens so often
<begiak> apertium: donissegui * 82503: /languages/apertium-cat/apertium-cat.cat.dix: picotejar
<begiak> apertium: donissegui * 82504: /trunk/apertium-spa-cat/apertium-spa-cat.spa-cat.dix: picotear
<spectei> is that better than just putting the masculine pronoun ?
<Unhammer> lene et al prefer it, and it does quite often go wrong, so I'm meh
<spectei> yeah
<spectei> maybe should use 'hæn' :D
<Unhammer> heheh :)
<spectei> or is that a word
<Unhammer> hen?
<spectei> no 'hæn'
<spectei> doesn't look like a word
<Unhammer> but that's often different connotations
<spectei> although maybe a dialect word?
<Unhammer> nah LGBTQF's sometimes use «hen»
<spectei> hmm, true
<spectei> i think the underscore isn't a great option, but if they like it :D
<Unhammer> but that also means it's not "neutral" like the swedish "hen" is meant to be, it's often used for explicitly middle-gender/ungender or whatnot
<spectei> i wonder if it could be done with some kind of discourse tracking thing in transfer
<spectei> kind of like variables, but automatically set
<Unhammer> we have some things with subjects etc
<Unhammer> but it's really hard to work with
<spectei> yeah
<Unhammer> also I think we did something in CG
<Unhammer> .lex
<spectei> i don't like that, it's a bit hacky
<spectei> but i struggle to think of a way to deal with it that isn't hacky
<spectei> ideally you'd like to just store the LU of the referent that you expect
<spectei> for example, <clip pos="1" side="ref" part="a_gen"/>
<Unhammer> SELECT han IF (0 han) (0 hun) (*1 subj + masc)
<Unhammer> blah
<spectei> would give you the gender of the referent
<spectei> which would be automatically calculated
<Unhammer> automaGically
<begiak> apertium: g-ramirez * 82505: /languages/apertium-spa/apertium-spa.spa.dix: minor fixes
<spectei> hahaha
<Unhammer> ftfy
<spectei> :/
<spectei> i'm trying to think of a language independent way to conceive this
<spectei> and "apertiumy"
<spectei> e.g.
<Unhammer> what if there are several referents?
<Unhammer> possessor+possessed
<begiak> apertium: g-ramirez * 82506: /languages/apertium-cat/apertium-cat.cat.dix: minor fixes
<begiak> apertium: g-ramirez * 82507: /trunk/apertium-spa-cat/apertium-spa-cat.spa-cat.dix: minor fixes
<Unhammer> "hans" masc possessor, "mi" fem possessed
<spectei> in the input to transfer have ^a<prn>/b<prn>/reflu$
<spectei> in this case you'd only need the possessor
<spectei> because the possessed comes in the chunk
<spectei> e.g. mor mi
<spectei> maybe the way to generalise this is to have a system whereby an extra LU can be added to the end of biltrans with extra informations
<spectei> ^el seu<det><def>/hans<det><def>/hon<prn><pers><p3><f><sg><nom>$
<spectei> and if you don't have an ambiguity, then biltrans becomes:
<spectei> ^casa<n><f>/hus<n><nt>/$
<Unhammer> den boka er ikkje mi
<Unhammer> den boka eg har her er ikkje mi
<spectei> and this extra LU could be added by some "magic" (statistics) or we could have some kind of rule-based component
<Unhammer> couldn't
<Unhammer> ^casa<n><f>/hus<n><nt>/$
<Unhammer> just be
<Unhammer> ^casa<n><f>/hus<n><nt>$
<spectei> no!
<Unhammer> ?
<spectei> because then you'd have ambiguity between:
<spectei> ^poble<n><m>/folk<n><nt>/by<n><m>$
<spectei> and
<spectei> ^poble<n><m>/folk<n><nt>/by<n><m>/$
<spectei> hmm, actually in the case of 1:1 yes
<spectei> you could just have that
<spectei> but if you had
<spectei> a/b/c
<spectei> then it's ambiguous
<spectei> if the 'c' is a reference-LU or if it's a translation
<Unhammer> oh, hm
<spectei> this would be fairly straightforward to hack into the code i think
<Unhammer> if it happens after lex-sel, then you should always have 1:1 as input to the anares
<spectei> i mean, in terms of transfer
<spectei> Unhammer, no because sometimes the lex-sel won't resolve everything
<spectei> then the transfer takes the first one
<Unhammer> lrx-proc -1 ?
<Unhammer> :)
<spectei> hah
<spectei> D:
<Unhammer> I mean, it could just remove everything after since transfer takes the first anyway
<spectei> i kind of like the idea of having /$ for empty referent
<Unhammer> ok
<spectei> but what do you think ?
<spectei> changing transfer to allow for that + something like <clip pos="1" side="ref" part="a_gen"/>
<spectei> would be 1-2 days' work
* fotonzade (~memduh@78.189.158.97) ha entrat a #apertium
<spectei> and you could do some preliminary experiments using CG
<spectei> "one could" :D
</pre>

@@ Line 1: / Line 1: @@
+== Old wiki ==
+{{TOCD}}
+Apertium has a problem with [[anaphora resolution]].
+For example:
+# If you have "el seu" in Catalan and are translating to French it could be "son" (third-person singular) or "leur" (third-person plural). If you are translating to English or Russian then you also need to know the gender of the possessor (его, ее, их).
+# If you are generating subject pronouns for a language, often you need to know the gender of the pronoun, e.g. "ha arribat" could be "He has arrived" or "She has arrived". In this case the "frequent" thing to do is to use the masculine pronoun, but that just relies on the male pronouns are used more frequently (see below):
+Usually this kind of thing is done over parse trees, but Apertium doesn't have parse trees, so we'd need to find another way to do it.
+Masculine and feminine subject pronouns in English wikipedia:
+<pre>
+5682787  he
+3469648  He
+1508156  she
+She
+</pre>
+=== Ideas ===
+==== Integration with the pipeline ====
+One idea is to allow an extra LU to be passed into transfer using the following biltrans format:
+* Els grups del Parlament han mostrat aquest dimarts el seu suport al batle d'Alaró, Guillem Balboa, que va denunciar que uns desconeguts havien llançat un xai mort al pati de casa seva.
+Current output of biltrans:
+<pre>
+^El<det><def><m><pl>/The<det><def><m><pl>$ ^grup<n><m><pl>/group<n><pl>$ ^de<pr>/of<pr>/from<pr>$
+^el<det><def><m><sg>/the<det><def><m><sg>$ ^Parlament<n><m><sg>/Parliament<n><sg>$
+^haver<vbhaver><pri><p3><pl>/have<vbhaver><pri><p3><pl>$
+^mostrar<vblex><pp><m><sg>/show<vblex><pp><m><sg>/display<vblex><pp><m><sg>$
+^aquest<det><dem><m><sg>/this<det><dem><m><sg>$ ^dimarts<n><m><sp>/Tuesday<n><ND>$
+^el seu<det><pos><m><sg>/his<det><pos><m><sg>$ ^suport<n><m><sg>/support<n><sg>$
+^a<pr>/at<pr>/in<pr>/to<pr>$ ^el<det><def><m><sg>/the<det><def><m><sg>$
+^*batle/*batle$ ^de<pr>/of<pr>/from<pr>$ ^*Alaró/*Alaró$^.<sent>/.<sent>$
+</pre>
+Proposed output:
+<pre>
+^El<det><def><m><pl>/The<det><def><m><pl>/$ ^grup<n><m><pl>/group<n><pl>/$ ^de<pr>/of<pr>/from<pr>/$
+^el<det><def><m><sg>/the<det><def><m><sg>/$ ^Parlament<n><m><sg>/Parliament<n><sg>/$
+^haver<vbhaver><pri><p3><pl>/have<vbhaver><pri><p3><pl>/$
+^mostrar<vblex><pp><m><sg>/show<vblex><pp><m><sg>/display<vblex><pp><m><sg>/$
+^aquest<det><dem><m><sg>/this<det><dem><m><sg>/$ ^dimarts<n><m><sp>/Tuesday<n><ND>/$
+^el seu<det><pos><m><sg>/his<det><pos><m><sg>/group<n><pl>$ ^suport<n><m><sg>/support<n><sg>/$
+^a<pr>/at<pr>/in<pr>/to<pr>/$ ^el<det><def><m><sg>/the<det><def><m><sg>/$
+^*batle/*batle/$ ^de<pr>/of<pr>/from<pr>/$ ^*Alaró/*Alaró/$^.<sent>/.<sent>/$
+</pre>
+Note that an extra <tt>/</tt> has been added, after this <tt>/</tt> comes the "extra information", which could
+be the anaphoric referent. For example, at the moment the translation is:
+<pre>
+The groups of the Parliament have shown this Tuesday his support at the *batle of *Alaró...
+</pre>
+But if we add the information about referent:
+<pre>
+^el seu<det><pos><m><sg>/his<det><pos><m><sg>/grup<n><m><pl>$
+</pre>
+We could get "their" instead of "his". This extra information could be addressed in transfer using something like:
+<pre>
+<clip pos="1" side="ref" part="a_gen"/>
+<clip pos="1" side="ref" part="a_nbr"/>
+</pre>
+This would provide minimal disturbance to the existing transfer syntax:
+<pre>
+<choose>
+  <when>
+    <test><equal><clip pos="1" side="ref" part="a_nbr"/><lit-tag v="pl"/></equal></test>
+    <let><clip pos="1" side="tl" part="lem"/><lit v="their"/></let></when>
+  <when> <!-- Should also check for animacy: el cotxe ... el seu ... → the car ... its ...  -->
+    <test><and><equal><clip pos="1" side="ref" part="a_gen"/><lit-tag v="m"/></equal></test>
+    <let><clip pos="1" side="tl" part="lem"/><lit v="his"/></let></when>
+  <when> <!-- Should also check for animacy: la máquina ... el seu ... → the machine ... its ... -->
+    <test><equal><clip pos="1" side="ref" part="a_gen"/><lit-tag v="f"/></equal></test>
+    <let><clip pos="1" side="tl" part="lem"/><lit v="her"/></let></when>
+</choose>
+</pre>
+The actual information could be added either using CG or by developing another tool (e.g. using some kind of machine learning).
+==== Processing and learning component ====
+We'd first need a way to mark possible NPs as antecedents. We could use a transfer-pattern-like file format for this, where we have e.g.
+<pre>
+<def-cat n="det">
+  <cat-item tags="det.*"/>
+</def-cat>
+<def-cat n="adj">
+  <cat-item tags="adj.*"/>
+</def-cat>
+<def-cat n="nom">
+  <cat-item tags="n.*"/>
+</def-cat>
+...
+<markable>
+<pattern>
+  <pattern-item n="det"/>
+  <pattern-item n="adj"/>
+  <pattern-item n="nom" head/>
+</pattern>
+</markable>
+</pre>
+This would then match the sequence and store the "nom" (and it's position) as the head and a potential antecedent.
+===Further reading===
+* Ruslan Mitkov (1999) "Multilingual Anaphora Resolution". ''Machine Translation. Volume 14, Issue 3–4, pp 281–299
 ==Comments==
 <pre>