A long introduction to transfer rules

Writing transfer rules seems to be tricky. People generally understand the basic concepts, but they struggle with the formalism. We think the formalism isn't that bad. And compared to many other formalisms, it's fairly straightforward. Maybe one of the reasons people struggle is that we mix declarative and procedural programming. Could be.

Some formalities

Before starting, it is important to give some idea of what we can't do, before explaining what we can. If you come at rule-learning expecting something else, then it's likely to be confusing.

• There are no recursive rules. Rules match fixed-length patterns. There is no optionality at the level of words. There is no way of saying one-or-more, it's just one.
• Apertium's rules are very tied to the Apertium stream format. If you don't understand the stream format, it will be a lot more difficult to understand the rules.
• Rules contain both declarative parts and procedural parts. You can't just expect to say what you want or how you want to do it. You need to do both -- but in different places (but it's quite intuitive).
• Patterns match only on the source side. Not on the target side.
• The structural transfer has no access to the information in the target language morphological dictionary. This means that if the transfer needs some information about the available forms of a particular word, e.g. if it is only singular, or only plural. Then this information needs to go in the bilingual dictionary.

Lexical transfer and structural transfer

See also: Bilingual dictionaries

At this point it's worth not confusing the rôles of lexical transfer and structural transfer. There is a grey area between the two, but there are also big parts that don't overlap.

• Lexical transfer:
  • Nearly always gives translations between words, not tags.
  • Can add or change tags, on a per-lemma basis.
  • Doesn't do reordering.
  • Can be used to give a head's up to the structural transfer to draw attention to missing features, or features that cannot be decided on a no-context basis. For example:
    • the <ND> and <GD> tags which say "Hey, when I'm translating this word I don't know what the gender or number should be -- structural transfer! I need your help to find out", or
    • the <sint> tag which says "¡Ojo! if you're writing a transfer rule for adjectives, and it matches this adjective then you need to think about how you're going to handle the comparative and superlative forms"

• Structural transfer:
  • Rarely gives translations between single words.
  • Often adds or changes tags on a per-category (groups of lemmas) basis.
  • Can change the order of words.

A rule-of-thumb is that if the rule applies to all words in a category, it probably wants to be treated in the structural transfer, and if it applies to just part of those words, then maybe it needs to be dealt with in the lexical transfer.

The output of the lexical transfer looks like this:

 ^slword<sometag>/tlword<sometag>$  ^slword1<sometag><blah>/tlword3<sometag><foo>$ ^slword3<sometag><blah>/tlword2<sometag><GD>$ 

Where the output of the structural transfer would look like this:

 ^tlword<sometag>$ ^tlword3<sometag><foo>$ ^tlword2<sometag><GD>$ 

That is, when you are in the first structural transfer stage you have access to both the source and target sides of the translation. After the first structural transfer stage, you only have access to the target side.

How lexical transfer is processed

Given a in input lexical unit: ^slword<tag1><tag2><tag3>$

If we have the following in the bilingual dictionary:

    <e><p><l>slword<s n="tag1"/><s n="tag2"/></l><r>tlword<s n="tag1"/></r></p></e>

We will get this output from the lexical-transfer module:

  ^slword<tag1><tag2><tag3>/tlword<tag1><tag3>$

Some preliminaries

• Pattern:

• Action:

Overview of a transfer file

It's hard to give a step-by-step overview of what a transfer file looks like because there is quite a lot of obligatory parts that need to go into even the most basic file. But, it's important to get a general view before we go into the details. Here is an example in which I'm deliberately not going to use linguistic names for the different parts, to try and avoid assumptions.

<?xml version="1.0" encoding="utf-8"?>
<transfer>
  <section-def-cats>
    <def-cat n="some_word_category">
      <cat-item tags="mytag.*"/>
    </def-cat>
  </section-def-cats>
  <section-def-attrs>
    <def-attr n="some_feature_of_a_word">
      <attr-item tags="myfeature"/>
      <attr-item tags="myotherfeature"/>
    </def-attr>
  </section-def-attrs>
  <section-def-vars>
    <def-var n="blank"/>
  </section-def-vars>  
  <section-rules>
    <rule>
      <pattern>
        <pattern-item n="some_word_category"/>
      </pattern>
      <action>
        <let><clip pos="1" side="tl" part="some_feature_of_a_word"/><lit-tag v="myotherfeature"/></let>
        <out>
          <lu><clip pos="1" side="tl" part="whole"/></lu>
        </out>
      </action>
    </rule>
  </section-rules>
</transfer>

I'll try and give a tag-by-tag account... the <transfer> and </transfer> tags don't do anything. They just encapsulate the rest of the sections.

Practical examples

Apertium 1

Input:	Otišla si	tiho	i	bez	pozdrava
Output:	You left	quietly	and	without	a word

Lexical transfer

We pass the phrase through our lexical transfer stage, and come up with the following output:

^otići<vblex><perf><iv><lp><f><sg>/leave<vblex><lp><f><sg>$ 
^biti<vbser><clt><pres><p2><sg>/be<vbser><clt><pres><p2><sg>$ 
^tiho<adv>/quietly<adv>$ 
^i<cnjcoo>/and<cnjcoo>$ 
^bez<pr>/without<pr>$ 
^pozdrav<n><mi><sg><gen>/word<n><sg><gen>$

Put the output into a file called example1.txt we're going to be needing it later.

At this point if you haven't already, it's worth trying model zero, that is no transfer rules, so try passing the output through the transfer file we made earlier and generating it:

$ cat /tmp/example1.txt | apertium-transfer -b rules.t1x rules.bin
^leave<vblex><lp><f><sg>$
^be<vbser><clt><pres><p2><sg>$ 
^quietly<adv>$ 
^and<cnjcoo>$ 
^without<pr>$ 
^word<n><sg><gen>$

Then try and pass it through a generator of English:

$ cat /tmp/example1.txt | apertium-transfer -b rules.t1x rules.bin | lt-proc -g sh-en.autogen.bin
#leave 
#be 
quietly 
and 
without 
#word

This is obviously inadequate, but don't worry, we're going to use the structural transfer module to make it adequate!

Thinking it through

Let's think about what changes we need to make in order to convert this into an adequate form for target language generation. NB: If we want to change information, it's a procedure, if we want to output it or not, it's a declaration.

Procedures

1. If the source language tag is <lp>, change the target language tag to <past>

Declarations

1. Output a subject pronoun which takes its person and number information from the main verb.
2. Output the main verb with information on category and tense (but not gender and number).
3. Not output the auxiliary verb (biti, "be").
4. Output nouns with category and number (but not case).
5. Words should be output encapsulated in ^ and $
6. Tags should be output encapsulated in < and >

Work order

So, what order do we do these in, well it doesn't really matter -- an experienced developer would probably do it in two stages, but for pedagogical purposes, we're going to split it up into five stages:

• First we're going to write a rule which matches the lp and auxiliary construction, and outputs only the main verb (declarations: 2, 3, 5)
  • Define the categories of "lp" and "auxiliary"
• Second we're going to edit that rule to change the source language tag from <lp> to <past> (procedure: 1)
  • Define the attribute of "tense"
• Third we're going to edit the same rule to not output gender and number (declaration: 2)
  • Define the attribute of "verb_type"
• Fourth we're going to edit the same rule to output a subject pronoun before the verb (declarations: 1, 5, 6)
  • Define the attributes of "person" and "number"
• Fifth we're going to write a new rule which matches the noun construction and output only category and number (declarations: 4, 5)
  • Define the category of "noun"
  • Define the attribute of "noun_type"

Cheatsheet

Here is what the input and output of each of the changes above will look like.

Change	Input	Output
1	^otići<vblex><perf><iv><lp><f><sg>/leave<vblex><lp><f><sg>$ ^biti<vbser><clt><pres><p2><sg>/be<vbser><clt><pres><p2><sg>$	^leave<vblex><lp><f><sg>$
2	^otići<vblex><perf><iv><lp><f><sg>/leave<vblex><lp><f><sg>$ ^biti<vbser><clt><pres><p2><sg>/be<vbser><clt><pres><p2><sg>$	^leave<vblex><past><f><sg>$
3	^otići<vblex><perf><iv><lp><f><sg>/leave<vblex><lp><f><sg>$ ^biti<vbser><clt><pres><p2><sg>/be<vbser><clt><pres><p2><sg>$	^leave<vblex><past>$
4	^otići<vblex><perf><iv><lp><f><sg>/leave<vblex><lp><f><sg>$ ^biti<vbser><clt><pres><p2><sg>/be<vbser><clt><pres><p2><sg>$	^prpers<prn><subj><p2><mf><sg>$ ^leave<vblex><past>$
5	^pozdrav<n><mi><sg><gen>/word<n><sg><gen>$	^word<n><sg>$

Implementation

Step 1

We're working on the output:

^otići<vblex><perf><iv><lp><f><sg>/leave<vblex><lp><f><sg>$
^biti<vbser><clt><pres><p2><sg>/be<vbser><clt><pres><p2><sg>$ 

1. Make our categories:

  <section-def-cats>
    <def-cat n="lp">
      <cat-item tags="vblex.*.*.lp.*"/>
    </def-cat>
    <def-cat n="biti-clt">
      <cat-item tags="vbser.clt.pres.*"/>
    </def-cat>
  </section-def-cats>

Why do we need .*.* ? -- Because of how the matching system in categories works. In the middle of tag sequences, a * is counted as a single tag. At the end, it is counted as any sequence of tags. So, we have <vblex> followed by any tag, followed by any tag, followed by <lp> followed by any sequence of tags.

2. Edit the example rule and replace the pattern.

      <pattern>
        <pattern-item n="lp"/>
        <pattern-item n="biti-clt"/>
      </pattern>

3. Save and compile the rule file.

$ apertium-preprocess-transfer rules.t1x rules.bin

Now test it:

$ cat example1.txt | apertium-transfer -b rules.t1x rules.bin

^leave<vblex><lp><f><sg>$ 
^quietly<adv>$ 
^and<cnjcoo>$ 
^without<pr>$ 
^word<n><sg><gen>$

Great!

Step 2

1. Add a <def-attr> giving the possible values of the tense feature.

So, now we've got rid of the verb "to be", we want to change the tag for tense from <lp> to <past>. This will involve using a statement, explicitly telling the transfer what we want to change. But before that, we need to define which possible values our feature can take. We do this with a <def-attr>. Let's call it tense. Add the following to the <section-def-attrs>.

    <def-attr n="tense">
      <attr-item tags="lp"/>
      <attr-item tags="past"/>
    </def-attr>

This says that words can have an attribute tense, which may be one of two values, either <lp> or <past>. As an exercise to the reader, you could try and add some other possible values.

2. Write a statement which changes the tense on the target-language (side="tl")

This statement should go immediately after the <action> tag and before the <out> tag.

      <action>
        <let><clip pos="1" side="tl" part="tense"/><lit-tag v="past"/></let>
        <out>

Now test it:

$ apertium-preprocess-transfer rules.t1x rules.bin

$ cat example1.txt | apertium-transfer -b rules.t1x rules.bin
^leave<vblex><past><f><sg>$ 
^quietly<adv>$ 
^and<cnjcoo>$ 
^without<pr>$ 
^word<n><sg><gen>$

Step 3

So far we have been outputting the whole target-language lexical unit. The reader will recall that the target-language lexical is produced by passing the output of the source language part-of-speech tagger through the lexical-transfer module, and replacing a prefix on the source code with a prefix on the target side. Any source-language tags not matched by the prefix in the bilingual dictionary are copied into the output on the target language side.

This is why we have:

^otići<vblex><perf><iv><lp><f><sg>/leave<vblex><lp><f><sg>$ 

Our input from the source language is otići<vblex><perf><iv><lp><f><sg>, our bilingual dictionary maps otići<vblex><perf><iv> → leave<vblex> so we are left with the tags <f><sg> on the target-language side. These tags are not needed for generation of English, so we need to declare that they should not be output.

1. Define the features that we want to output.

We have already defined the feature (def-attr) tense, the only remaining tag we need to output is the verb type. So, make a new def-attr for verb type. In Apertium usually, these are prefixed with a_ this stands for "attribute", so it isn't confused with other uses of "verb".

    <def-attr n="a_verb">
      <attr-item tags="vblex"/>
      <attr-item tags="vbser"/>
    </def-attr>

2. Declare which features we want to output.

So, now that we've got our two features defined, let's change our <out> statement:

Where we previously had:

          <lu><clip pos="1" side="tl" part="whole"/></lu>

Replace it with:

          <lu>
            <clip pos="1" side="tl" part="lem"/>
            <clip pos="1" side="tl" part="a_verb"/>
            <clip pos="1" side="tl" part="tense"/>
          </lu>

The two second features we know, as we've just added them. The "lem" feature is a pre-defined feature which corresponds to "lemma". There are others, if you remember "whole" corresponds to the whole lexical unit, and "tags" corresponds to all the tags.

Now test it:

$ apertium-preprocess-transferb rules.t1x rules.bin

$ cat example1.txt | apertium-transfer -b rules.t1x rules.bin
^leave<vblex><past>$ 
^quietly<adv>$ 
^and<cnjcoo>$ 
^without<pr>$ 
^word<n><sg><gen>$

Woo! Now we should have a lexical form for "leave" which we can generate... let's give it a go:

$ cat example1.txt | apertium-transfer -b rules.t1x rules.bin | lt-proc -g sh-en.autogen.bin
left 
quietly 
and 
without 
#word

We're getting there, now let's move onto the next step...

Step 4

To recap, our input is:

^otići<vblex><perf><iv><lp><f><sg>/leave<vblex><lp><f><sg>$
^biti<vbser><clt><pres><p2><sg>/be<vbser><clt><pres><p2><sg>$

And our current output is:

^leave<vblex><past>$

What we need to do is take the information from the (now not output) verb biti "to be" and use it to output a personal pronoun before the verb. Remember, that when we do not output something we are not deleting it, it is still there in the input, we are just choosing not to put it in the output.

The string that we want to output looks like this:

^prpers<prn><subj><p2><mf><sg>$

1. Output a skeleton pronoun.

So, we need to take the person and number information from the verb "to be", and the rest we need to just declare to be output. Let's start with outputting the string ^prpers<prn><subj>$ before the verb.

          <lu>
            <lit v="prpers"/>
            <lit-tag v="prn.subj"/>
          </lu>
          <b/>

The <lu> outputs a ^ and the </lu> outputs a $. The <lit> tag outputs a string literal, and the <lit-tag> outputs a tag sequence which starts with a <, ends with a > and where each . is replaced with ><. The <b/> outputs a single space ' ' character.

2. Defining features for person and number.

The next step is to declare that features from the verb be output. To do this, we need to again define some features and their possible values.

    <def-attr n="person">
      <attr-item tags="p1"/>
      <attr-item tags="p2"/>
      <attr-item tags="p3"/>
    </def-attr>
    <def-attr n="number">
      <attr-item tags="sg"/>
      <attr-item tags="pl"/>
    </def-attr>

We then go back to our skeleton pronoun and add the missing features:

          <lu>
            <lit v="prpers"/>
            <lit-tag v="prn.subj"/>
            <clip pos="2" side="sl" part="person"/>
            <lit-tag v="mf"/>
            <clip pos="2" side="sl" part="number"/>
          </lu>
          <b/>

There are two main things worth noting here, the first is that we are "clipping" from position 2, that is the verb "to be". If we tried to clip from position 1, we would get no result for the "person" attribute because the verb in position 1 does not contain any tags which match the tags in the def-attr for person. The second is that we've outputted the gender as a literal tag. This is because although our first verb gives us the gender, it gives us just <f>, and English doesn't express gender in the second person pronoun.

Now test it:

$ apertium-preprocess-transfer rules.t1x rules.bin

$ cat example1.txt | apertium-transfer -b rules.t1x rules.bin 
^prpers<prn><subj><p2><mf><sg>$ ^leave<vblex><past>$ 
^quietly<adv>$ 
^and<cnjcoo>$ 
^without<pr>$ 
^word<n><sg><gen>$

And then with the generator:

$ cat example1.txt | apertium-transfer -b rules.t1x rules.bin | lt-proc -g  sh-en.autogen.bin 
you left 
quietly 
and 
without 
#word

Step 5

So, the last remaining thing to do is to not output the genitive tag on the noun. We're going to have to make a whole new rule to match nouns.

1. Define a category for nouns

    <def-cat n="noun">
      <cat-item tags="n.*"/>
    </def-cat>

2. Make a new rule which matches nouns as defined by the previous category.

To start with we'll just output the whole lexical unit.

<rule>
  <pattern>
    <pattern-item n="noun"/>
  </pattern>
  <action>
    <out>
      <lu><clip pos="1" side="tl" part="whole"/></lu>
    </out>
  </action>
</rule>

3. Define a new category for noun type

    <def-attr n="a_noun">
      <attr-item tags="n"/>
      <attr-item tags="np"/>
    </def-attr>

4. Adjust the rule to output only noun type and number

Remember that we previously defined the <def-attr> for "number".

    <out>
      <lu>
        <clip pos="1" side="tl" part="lem"/>
        <clip pos="1" side="tl" part="a_noun"/>
        <clip pos="1" side="tl" part="number"/>
      </lu>
    </out>

Now test it:

$ cat example1.txt | apertium-transfer -b rules.t1x rules.bin 
^prpers<prn><subj><p2><mf><sg>$ ^leave<vblex><past>$ 
^quietly<adv>$ 
^and<cnjcoo>$ 
^without<pr>$ 
^word<n><sg>$

$ cat example1.txt | apertium-transfer -b rules.t1x rules.bin  | lt-proc -g sh-en.autogen.bin
you left
quietly
and
without
word

Success!

Step 6

Step 6 is left as an exercise for the reader. Change the noun rule to output an indefinite article before the noun. You can do this following the instructions for adding the pronoun. The string you need to output is ^a<det><ind><sg>$.

Apertium 3

Resorni ministar je navlačio ljude, kaže sejte biljku zelenu i čudo će da bude

The minister of agriculture tricks the people, he says plant the green herb and there will be a miracle