Tagger training

From Apertium
Jump to navigation Jump to search

En français

Once your dictionaries are of a reasonable size, say perhaps 3,000 lemmata in total, it is worth training the tagger. To do this, you'll need a couple of things, a decent sized corpus, either tagged or untagged, and a .tsx file. The basic instructions may be found below.

Creating a corpus[edit]

Wikipedia[edit]

A basic corpus can be retrieved from a Wikipedia dump (see here) as follows:

$ bzcat afwiki-20070508-pages-articles.xml.bz2 | grep '^[A-Z]' | sed 's/$/\n/g' | sed 's/\[\[.*|//g' | sed 's/\]\]//g' | sed 's/\[\[//g' | sed 's/&.*;/ /g' > mycorpus.txt

Another option for stripping Wikipedia, which will probably result in a higher quality corpus, is as follows. First download the Wikipedia extractor script from here, then:

$ bzcat enwiki-20081008-pages-articles.xml.bz2.part > enwiki.xml

Now use the script above to get enwiki.txt and then

$ cat enwiki-20091001-pages-articles.txt | grep -v "''" | grep -v http | grep -v "#" | grep -v "@" |\
grep -e '................................................' | sort -fiu | sort -R | nl -s ". " > enwiki.crp.txt

The last 3 commands are not strictly necesary. They sort and finds only uniqe lines, then sorts randomly (mix the sentences) and adds line numbers.

Other sources[edit]

Some pre-processed corpora can be found here and here.

Writing a TSX file[edit]

See also: TSX format

A .tsx file is a tag definition file, it turns the fine tags from the morphological analyser into coarse tags for the tagger. The DTD is in tagger.dtd, although it is probably easier to take a look at one of the pre-written ones in other language pairs.

The file should be in the language pair directory and be called (in for example English-Afrikaans), apertium-en-af.en.tsx for the English tagger, and apertium-en-af.af.tsx for the Afrikaans tagger.

The TSX file defines a set of "coarse tags" for groups of "fine tags", this is done because the POS tagging module does not need so much information as is defined in the fine tags. It also allows the user to apply a set of restrictions or enforcements. For example to forbid a relative adverb at the start of a sentence (SENT RELADV), or to forbid a pronoun after a noun (NOM PRNPERS).

You can also write lexical rules, so for example in Afrikaans, the word "deur" is ambiguous, one meaning is "by" (as a preposition) and the other is "door" (as a noun). So we can define two coarse tags, DEURNOM and DEURPR, and then a forbid rule to say "forbid 'door' before 'the'".

It is worth considering this file carefully and probably also consulting with a linguist, as the tagger can make a big difference to the quality of the final translation. The example below gives the basic structure of the file:

<?xml version="1.0" encoding="UTF-8"?>
<tagger name="afrikaans">
  <tagset>
    <def-label name="DEURNOM" closed="true">
      <tags-item lemma="deur" tags="n.*"/>
    </def-label> 
    <def-label name="DEURPR" closed="true">
      <tags-item lemma="deur" tags="pr"/>
    </def-label>     
    <def-label name="NOM">
      <tags-item tags="n.*"/>
    </def-label> 
    <def-label name="PRNPERS" closed="true">
      <tags-item tags="prpers.*"/>
    </def-label> 
    <def-label name="DET" closed="true">
      <tags-item tags="det.*"/>
    </def-label> 
  </tagset>
  <forbid>
    <label-sequence>
      <label-item label="NOM"/>
      <label-item label="PRNPERS"/>
    </label-sequence>
    <label-sequence>
      <label-item label="DEURNOM"/>
      <label-item label="DET"/>
    </label-sequence>
  </forbid>
</tagger>

You will need enough coarse tags to cover all the fine tags in your dictionaries.

Training the tagger[edit]

A brief note on the various kinds of training that you can do:

  • Unsupervised — This uses a large (hundreds of thousands of words) untagged corpus and the iterative Baum-Welch algorithm in a wholely unsupervised manner. This is the least effective way of training the tagger, but is also the cheapest in terms of time and resources.
  • Supervised — This uses a medium sized (minimum 30,000 words) tagged corpus.
  • Using apertium-tagger-trainer — This uses a large untagged corpus in the target language, a previously trained .prob file and an existing translator. It performs as well as supervised training without the need of hand-tagging a corpus, at the expense of being a bit tricky to set up.

At the moment apertium-tagger-trainer only works with apertium 1, so it's not an option for most pairs.--Jacob Nordfalk 06:15, 17 September 2008 (UTC) (Clarification: it only works with one-stage transfer, so Apertium 3 pairs which only have t1x can still use it.)

Unsupervised[edit]

Main article: Unsupervised tagger training

Supervised[edit]

Main article: Supervised tagger training

Target language tagger training[edit]

Main article: Target language tagger training

There is a package called apertium-tagger-training-tools that trains taggers based on both source and target language information. The resulting probability files are as good as supervised training for machine translation purposes, but much quicker to produce, and with less effort.

See also[edit]

Further reading[edit]