| blob | baf253f116841d20919c8311511428ba00df400d |
4 <head>
13 /*<![CDATA[*/
15 /*]]>*/
16 </style>
17 </head>
18 <body>
20 <p>This HOWTO document will describe how to start a new language
21 pair for the Apertium machine translation system from scratch.</p>
22 <p>It does not assume any knowledge of linguistics, or machine
23 translation above the level of being able to distinguish nouns from
24 verbs (and prepositions etc.)</p>
26 this, and of the intuitive approach to machine translation used in
27 Apertium, the amount of linguistic knowledge necessary about the
28 source and target language to build data for Apertium is kept to a
29 minimum, and it may be easily learned on top of basic high-school
30 grammar skills such as: morphological analysis of words:
31 parts-of-speech or lexical categories (noun, verb, preposition,
32 etc.) and basic morphology (number, gender, case, person, etc.;
33 agreement (such as gender and number agreement between nouns and
34 their modifiers: adjectives, determiners, etc.); main local
35 structural differences between the source and target language:
36 position of adjectives with respect to nouns (e.g, adjective after
37 noun in Spanish, before noun in English), prepositional regime,
40 <ul>
42 </ul>
45 Permission is granted to copy, distribute and/or modify this
46 document under the terms of the GNU Free Documentation License,
47 Version 1.2 or any later version published by the Free Software
48 Foundation; with no Invariant Sections, no Front-Cover Texts, and
49 no Back-Cover Texts. A copy of the license is included in the
50 section entitled "<a href=
53 <p>Any extracts of code are released under the GNU GPL, Version
56 <p>Apertium is, as you've probably realised by now, a machine
57 translation system. Well, not quite, its a machine translation
58 platform. It provides and engine and toolbox which allow you to
60 you need to do is write the data. The data consists on a basic
61 level, of three dictionaries and a few rules (to deal word
62 re-ordering and other grammatical stuff).</p>
63 <p>For a more detailed introduction into how it all works, there
64 are some excellent papers on the project's website <a href=
67 <ul>
71 <li>a text editor (or a specialized XML editor if you prefer
72 to)</li>
73 </ul>
74 <p>This document will not describe how to install these packages,
75 for more information on that, please see the documentation section
76 of the Apertium website.</p>
78 <p>The Apertium machine translation system is of the
79 shallow-transfer type, this basically means it works on
80 dictionaries and shallow transfer rules. Shallow transfer
81 is distinguished from "deep transfer" in that it doesn't do
82 full syntactic parsing, the rules are typically operations on
83 groups of lexical units, rather than operations on parse trees.
84 On a basic level, there are three main dictionaries:</p>
85 <!--
86 <div class="comment">Perhaps it would be a good idea to try to
87 explain what "deep transfer" would mean (full syntactic parsing with rules specifying
88 operations on parse [sub]trees, etc.)
89 so that the readers get an idea of what "shallow" would mean.</div>
91 :: I added something above, but it is difficult (for me) to
92 explain without using terminology that the readers would
93 be unfamiliar with. If it can be refined great, if not
94 I'm not sure... ~fran
95 -->
99 inflected. In our example this will be called:
103 inflected. In our example this will be called:
106 words and symbols in the two languages. In our example this will be
108 </ul>
109 <p>In a translation pair, both languages can be either source or
110 target for translation, these are relative terms.
111 <!--
112 <div class="comment">Not that <em>xx</em> and <em>yy</em> are being used, why cannot
113 we use these instead of sh and en (for the initial examples discussion)? Also,
114 I would advocate for using <em>source</em> and <em>target</em> as <em>relative</em>
115 terms; this was the initial motivation for using <em>xx</em> and <em>yy</em>.</div>
117 :: About the filenames: I wanted to give a concrete example. Regarding
118 the source/target, you're right, I've removed the 'typically we will
119 use' part. ~fran
120 -->
122 <p>There are also two files for transfer rules. These are the rules
123 which govern how words are re-ordered in sentences, e.g. <cite>chat
125 cat</cite>. It also governs agreement of gender, number etc. The
126 rules can also be used to insert or delete lexical items, as will
127 be described later. These files are:</p>
128 <ul>
137 </ul>
138 <!--
139 <div class="comment">Perhaps one should note that this
140 is Apertium "level 1".</div>
142 :: I need to note the version, but I think I'll re-do the
143 examples etc. from an Apertium 2 point of view. I remember
144 using HOWTOs before and being annoyed that they were out of
145 date! ~fran
146 -->
148 <p>Many of the language pairs currently available have other files,
149 but we won't cover them here. These files are the only ones
150 required to generate a functional system.</p>
152 <p>As may have been alluded by the file names, this HOWTO will use
153 the example of translating Serbo-Croatian to English to explain how
154 to create a basic system. This is not an ideal pair, as the system
155 works better for more closely related languages, and furthermore it
156 does not currently support the full Serbo-Croatian alphabet, but
157 that shouldn't present a problem for the simple examples we'll have
158 here.</p>
159 <!--
160 <div class="comment">We can remove the sentence "As may have been alluded..." if we use <em>xx</em>
161 and <em>yy</em> for the examples up to here.</div>
163 :: I reckon its alright, and good to give concrete examples,
164 although this may get removed when i do a re-write ~fran
165 -->
168 <p>There are number of terms that will need to be understood before
169 we continue.</p>
171 word. It is the word stripped of any grammatical information. For
173 English nouns this will typically be the singular form of the word
174 in question. For verbs, the lemma is the infinitive stripped of
181 typically given between angle brackets, as follows:</p>
182 <ul>
185 </ul>
186 <!--
187 <div class="comment">Make it clear that symbols appear in angle brackets in
188 the input and the output of modules, but not in dictionaries?</div>
190 :: Good point, done. Feel free to rephrase. ~fran
191 -->
196 noting that in many of the currently available language pairs the symbol definitions
200 tags. </p>
201 <!--
202 <div class="comment">Should we warn the reader that symbol names in many of the available language
203 pair packages are in Catalan?</div>
205 :: Good idea, done. Feel free to rephrase. ~fran
206 -->
209 Apertium system, paradigm refers to a example of how a particular
210 group of words inflect. In the morphological dictionary, lemmas
211 (see above) are linked to paradigms which allows us to describe how
212 a given lemma inflects without having to write out all of the
213 endings.</p>
214 <p>An example of the utility of this is, if we wanted to store the
216 two lots of the same thing:</p>
217 <ul>
222 </ul>
226 "<em>friendly</em> inflects like <em>happy</em>" etc. In this
227 example, happy would be the paradigm, the model for how the others
228 inflect. The precise description of how this is defined will be
229 explained shortly. Paradigms are defined in
234 <p>Lets start by making our first source language dictionary. The
235 dictionary is an XML file. Fire up your text editor and type the
236 following:</p>
239 <br />
244 encoding. There is work ongoing to port it to Unicode (indeed an
246 is available from the SVN repository on the Apertium project
247 site).</p>
248 <p>Note: It is important to have your locale set up correctly when
249 writing/reading files, you can find out your current locale setting
251 <p>So, the file so far defines that we want to start a dictionary.
252 In order for it to be useful, we need to add some more entries, the
253 first is an alphabet. This defines the set of letters that may be
254 used in the dictionary, for Serbo-Croatian. Normally it would look
255 something like the following, containing all the letters of the
256 Serbo-Croatian alphabet:</p>
257 <p>
258 <tt><alphabet>ABCČĆDDžĐEFGHIJKLLjMNNjOPRSŠTUVZŽabcčćddžđefghijklljmnnjoprsštuvzž</alphabet></tt></p>
260 <p>
262 <p>The reason for this is that, as mentioned above,
266 minuscule forms) are not found in this encoding. Some languages
268 represent the missing letters, and then transliterating. For
269 example, using the character 'ç' (c with cedilla) to represent 'ć'
270 (c with acute accent), or using 'ð' (eth) to represent 'đ' (d with
271 stroke). We will not be using this method, although an example of
272 its use may be found in the Romanian-Spanish translation pair.</p>
273 <!--
274 <div class="comment">I don't think <em>lj</em> and <em>nj</em> pose any problem, since they are not
275 special letters but digraphs. Are you really going to use these transliterations?</div>
277 :: lj and nj can be written as both digraphs (or, more 'properly' as
278 paired characters, like æ - note that we won't use them or the
279 transliterations. ~fran
280 -->
283 tag.</p>
284 <p>Next we need to define some symbols. Lets start off with the
292 <p>The symbol names do not have to be so small, in fact they could
293 be just written our in full, but as you'll be typing them a lot, it
294 makes sense to abbreviate.</p>
295 <p>Unfortunately, it isn't quite so simple, nouns in Serbo-Croatian
296 inflect for more than just number, they also inflect for gender and
297 case. However, we'll assume for the purposes of this example that
298 the noun is masculine and in the nominative case (a full example
299 may be found at the end of this document).</p>
302 <br />
306 <br />
308 <p>
310 section, which contains words, enclitics etc. The second type is an
312 punctuation etc. We don't have an inconditional section here, although
313 it will be demonstrated later.
314 </p>
315 <!--
316 <div class="comment">Perhaps we should say that each <code><section></code> is a dictionary section withcertain properties; for instance, a <code>standard</code> section contains words that will
317 only be segmented if the next character is out of the alphabet, whereas a <code>unconditional</code> section
318 contains words that will be segmented regardless of the following character, etc. What is
319 called "a section for mapping lemmas to paradigms" is simply a "dictionary section".</div>
321 :: we should include an example of 'splitting up' / i'm not
322 quite sure how this works. ~fran
324 -->
334 <br />
337 <br />
340 <p>Now we've got the skeleton in place, we can start by adding a
341 noun. The noun in question will be 'gramofon' (which means
342 'gramophone' or 'record player').</p>
343 <p>The first thing we need to do, as we have no prior paradigms, is
344 to define a paradigm.</p>
345 <p>Remember we're assuming masculine gender and nominative case.
346 The singular form of the noun is 'gramofon', and the plural is
347 'gramofoni'. So:</p>
365 denotes that there is no extra material to be added to the stem for
366 the singular.</p>
367 <p>This may seem like a rather verbose way of describing it, but
368 there are reasons for it and it quickly becomes second nature.
371 <ul>
376 </ul>
378 dictionaries will later be compiled into finite state machines.
379 Compiling them left to right produces analyses from words, and from
380 right to left produces words from analyses. For example:</p>
381 <ul>
386 </ul>
387 <p>Now we've defined a paradigm, we need to link it to its lemma,
389 we've defined.</p>
394 <ul>
397 right are the same).</li>
399 </ul>
403 and the root is that the lemma is the citation form of the word,
404 while the root is the substring of the lemma to which stems are
405 added. This will become clearer later when we show an entry where
406 the two are different.</p>
407 <p>We're now ready to test the dictionary. Save it, and then return
408 to the shell. We first need to compile it (with
411 <pre>
412 $ lt-comp lr apertium-sh-en.sh.dix sh-en.automorf.bin
413 </pre>
415 <pre>
416 main@standard 12 12
417 </pre>
418 <p>As we are compiling it left to right, we're producing an
420 <pre>
421 $ lt-comp rl apertium-sh-en.sh.dix sh-en.autogen.bin
422 </pre>
424 <!--
425 <div class="comment">Not necessarily true for the examples below in which you
426 equate <em>gramofoni</em> (nominative) and <em>gramophone</em> (accusative); when generating, you should decide which one...
427 </div>
429 :: at the moment we're not dealing with generating serbo-croatian, and
430 the 'output' refers to the 'main@standard 12 12' part. I've clarified this
431 somewhat ~fran
432 -->
434 analyser.</p>
435 <pre>
436 $ lt-proc sh-en.automorf.bin
437 </pre>
440 <pre>
442 </pre>
443 <p>Now, for the English dictionary, do the same thing, but
445 change the plural inflection. What if you want to use the more correct
446 word 'record player', well, we'll explain how to do that later.</p>
447 <!--
448 <div class="comment">Check</div>
450 :: I added that they should change the plural inflection (i -> s),
451 what this what the check referred to ? ~fran
452 -->
454 <ul>
456 basic Serbo-Croatian morphological dictionary, and</li>
458 basic English morphological dictionary.</li>
459 </ul>
461 <p>So we now have two morphological dictionaries, next thing to
462 make is the bilingual dictionary. This describes mappings between
463 words. All dictionaries use the same format (which is specified in
466 the basic skeleton:</p>
475 <br />
477 <br />
481 two words. Something like:</p>
485 <!--
486 <div class="comment">Some of this could be done with
487 <i></div>
489 :: How do you mean? ~fran
490 -->
491 <p>Because there are a lot of these entries, they're typically
492 written on one line to facilitate easier reading of the file. Again
494 compile it left to right to produce the Serbo-Croatian → English
495 dictionary, and right to left to produce the English →
496 Serbo-Croatian dictionary.</p>
498 <pre>
499 $ lt-comp lr apertium-sh-en.sh.dix sh-en.automorf.bin
500 $ lt-comp rl apertium-sh-en.sh.dix sh-en.autogen.bin
502 $ lt-comp lr apertium-sh-en.en.dix en-sh.automorf.bin
503 $ lt-comp rl apertium-sh-en.en.dix en-sh.autogen.bin
505 $ lt-comp lr apertium-sh-en.sh-en.dix sh-en.autobil.bin
506 $ lt-comp rl apertium-sh-en.sh-en.dix en-sh.autobil.bin
507 </pre>
511 <!--
512 <div class="comment">These three names come from interNOSTRUM but I
513 am not sure I like them much... unfortunately, they are used all
514 over...</div>
516 :: Aye, they're a bit cryptic, but a lot of the stuff is,
517 when you know what they mean, or where they come from
518 it makes more sense. ~fran
519 -->
521 <p>So, now we have two morphological dictionaries, and a bilingual
522 dictionary. All that we need now is a transfer rule for nouns.
524 which can be found in the Apertium package. If you need to
525 implement a rule it is often a good idea to look in the rule files
526 of other language pairs first. Many rules can be recycled/reused
527 between languages. For example the one described below would be
528 useful for any null-subject language.</p>
532 <br />
534 <p>At the moment, because we're ignoring case, we just need to make
535 a rule that takes the grammatical symbols input and outputs them
536 again.</p>
537 <p>We first need to define categories and attributes. Categories
538 and attributes both allow us to group grammatical symbols.
539 Categories allow us to group symbols for the purposes of matching
541 to group a set of symbols that can be chosen from. For example
546 <br />
549 <br />
551 <p>As we're only inflecting, nouns in singular and plural then we
552 need to add a category for nouns, and with an attribute of number.
553 Something like the following will suffice:</p>
559 anything) and refers to them as "<code>nom</code>" (we'll see how
560 thats used later). </p>
561 <!--
562 <div class="comment">Catalan names again</div>
564 :: I think its fine for now, we note that the
565 names are fairly arbitrary. ~fran
566 -->
583 <br />
585 <p>These variables are used to store or transfer attributes
586 between rules. We need only one for now,</p>
587 <pre>
589 </pre>
590 <!--
591 <div class="comment">Perhaps we should be more specific and say
592 that these are <em>state</em> variables which may be used to
593 propagate information computed in a certain application of a rule,
594 to later applications of the same or other rules.</div>
596 :: Hmm, I think this is getting a bit complicated. Although
597 it would be good to give a concrete example of where
598 they are used. But I haven't used these yet. I'll dig
599 around in the manual. ~fran
600 -->
601 <p>Finally, we need to add a rule, to take in the noun and then
602 output it in the correct form. We'll need a rules section...</p>
604 <br />
606 <p>Changing the pace from the previous examples, I'll just paste
607 this rule, then go through it, rather than the other way round.</p>
626 languages such as AWK, perl or lex. How about drawing an analogy
627 here?</div>
628 <p>The first tag is obvious, it defines a rule. The second tag,
630 pattern is found". In this example the pattern consists of a single
632 patterns are matched in a longest-match first. So if you have three
633 rules, the first catches "<prn><vblex><n>", the
634 second catches "<prn><vblex>" and the third catches
635 "<n>", the pattern matched, and rule executed will be the
636 first.</p>
637 <!--
638 <div class="comment">I would rather use an example in which rules
639 match whole lexical units, because that is the way the transfer
640 works. One cannot write patterns matching parts of lexical units,
641 as far as I know</div>
643 :: Not sure I understand, could you re-write this part? ~fran
644 -->
645 <p>For each pattern, there is an associated action, which produces
649 attributes and parts of the source language
652 <p>Let's compile it and test it. Transfer rules are compiled
653 with:</p>
654 <pre>
655 $ apertium-preprocess-transfer apertium-sh-en.trules-sh-en.xml trules-sh-en.bin
656 </pre>
658 <p>Now we're ready to test our machine translation system. There is
659 one crucial part missing, the part-of-speech (PoS) tagger, but that
660 will be explained shortly. In the meantime we can test it as
661 is:</p>
663 <pre>
664 $ echo "gramofoni" | lt-proc sh-en.automorf.bin
666 </pre>
667 <p>Now, normally here the POS tagger would choose the right version
668 based on the part of speech, but we don't have a POS tagger yet, so
669 we can use this little perl script that will just output the first
670 item retrieved.</p>
671 <!--
672 <div class="comment">I am not sure the reader would understand what
673 <em>version</em> means here unless we have an example in which the
674 morphological analyser derives more than one lexical form for a
675 given source form... Perhaps we should explain that, and give an
676 example...</div>
678 :: Good idea. I'll try and work an example in. ~fran
680 -->
681 <pre>
682 $ echo "gramofoni" | lt-proc sh-en.automorf.bin | \
685 </pre>
686 <!--
687 <div class="comment">We are writing the pipelines ourselves, but
688 Apertium uses <code>modes.xml</code> file to control this; perhaps
689 a later version of this howto can explain how to write
690 <code>modes.xml</code> files... if you look at
691 <code>apertium-en-ca</code>, you will see that the
692 <code>modes.xml</code> file contains many modes, to use parts of
693 the pipeline for diagnostics and testsing.</div>
695 :: I'll add something on modes.xml at the end. ~fran
697 -->
699 <pre>
700 $ echo "gramofoni" | lt-proc sh-en.automorf.bin | \
702 apertium-transfer apertium-sh-en.trules-sh-en.xml trules-sh-en.bin sh-en.autobil.bin
703 </pre>
705 <pre>
707 </pre>
708 <!--
709 <div class="comment">What's the "^@"?</div>
711 :: I think that its because we don't have any
712 punctuation specified. I'll check it and
713 put a note in. ~fran
715 -->
716 <ul>
723 </ul>
724 <p>Try commenting out one of these clip statements, recompiling and
725 seeing what happens.</p>
726 <p>So, now we have the output from the transfer, the only thing
727 that remains is to generate the target-language inflected forms.
730 <pre>
731 $ echo "gramofoni" | lt-proc sh-en.automorf.bin | \
733 apertium-transfer apertium-sh-en.trules-sh-en.xml trules-sh-en.bin sh-en.autobil.bin | \
734 lt-proc -g sh-en.autogen.bin
736 gramophones\@
737 </pre>
738 <!--
739 <div class="comment">What is the "\@"?</div>
741 :: I think it is as a result of not having any
742 punctuation defined. ~fran
743 -->
744 <p>And c'est ca. You now have a machine translation system that
745 translates a Serbo-Croatian noun into an English noun. Obviously
746 this isn't very useful, but we'll get onto the more complex stuff
747 soon. Oh, and don't worry about the '@' symbol, I'll explain that
748 soon too.</p>
749 <p>Think of a few other words that inflect the same as gramofon.
750 How about adding those. We don't need to add any paradigms, just
751 the entries in the main section of the monolingual and bilingual
752 dictionaries.</p>
754 <p>Ok, so we have a system that translates nouns, but thats pretty
755 useless, we want to translate verbs too, and even whole sentences!
758 null-subject language, this means that it doesn't typically use
759 personal pronouns before the conjugated form of the verb. English
762 <pre>
763 * Vidim
765 * I see
766 </pre>
768 <p>This will be important when we come to write the transfer rule
769 for verbs. Other examples of null-subject languages include:
770 Spanish, Romanian and Polish. The also has the effect that while we
771 only need to add the verb in the Serbo-Croatian morphological
772 dictionary, we need to add both the verb, and the personal pronouns
773 in the English morpohlogical dictionary. We'll go through both of
774 these.</p>
786 interests of space, we'll just add enough information to do the
787 translation and will leave filling out the paradigms (adding the
788 other conjugations of the verb) as an exercise to the reader.</p>
789 <p>The astute reader will have realised by this point that we can't
791 grammatically correct sentence in Serbo-Croatian. The correct
793 the accusative case. We'll have to add that form too, no need to
794 add the case information for now though, we just add it as another
798 <!--
799 <div class="comment">Mikel's comments so far end here.</div>
800 -->
801 <p>First thing we need to do is add some more symbols. We need to
802 first add a symbol for 'verb', which we'll call "vblex" (this means
803 lexical verb, as opposed to modal verbs and other types). Verbs
804 have 'person', and 'tense' along with number, so lets add a couple
805 of those aswell. We need to translate "I see", so for person we
807 'present indicative'.</p>
808 <pre>
812 </pre>
813 <p>After we've done this, the same with the nouns, we add a
814 paradigm for the verb conjugation. The first line will be:</p>
816 <p>The '/' is used to demarcate where the stems (the parts between
826 <p>The 'im' denotes the ending (as in 'vidim'), it is necessary to
828 the definition. The rest is fairly straightforward, 'vblex' is
829 lexical verb, 'pri' is present indicative tense, 'p1' is first
830 person and 'sg' is singular. We can also add the plural which will
831 be the same, except 'imo' instead of 'im' and 'pl' instead of
832 'sg'.</p>
833 <p>After this we need to add a lemma, paradigm mapping to the main
834 section:</p>
838 lemma.</p>
839 <p>Thats the work on the Serbo-Croatian dictionary done for now.
840 Lets compile it then test it.</p>
841 <pre>
842 $ lt-comp lr apertium-sh-en.sh.dix sh-en.automorf.bin
843 main@standard 23 25
844 $ echo "vidim" | lt-proc sh-en.automorf.bin
846 $ echo "vidimo" | lt-proc sh-en.automorf.bin
848 </pre>
849 <p>Ok, so now we do the same for the English dictionary (remember
850 to add the same symbol definitions here as you added to the
851 Serbo-Croatian one).</p>
854 <p>because the past tense is 'saw'. Now, we can do one of two
855 things, we can add both first and second person, but they are the
856 same form. In fact, all forms (except third person singular) of the
857 verb 'to see' are 'see'. So instead we make one entry for 'see' and
858 give it only the 'pri' symbol.</p>
870 <pre>
871 $ lt-comp lr apertium-sh-en.en.dix en-sh.automorf.bin
872 main@standard 18 19
874 $ echo "see" | lt-proc en-sh.automorf.bin
876 </pre>
878 <pre>
879 <e><p><l>videti<s n="vblex"/></l><r>see<s n="vblex"/></r></p></e>
880 </pre>
883 <pre>
884 $ lt-comp lr apertium-sh-en.sh-en.dix sh-en.autobil.bin
885 main@standard 18 18
886 $ lt-comp rl apertium-sh-en.sh-en.dix en-sh.autobil.bin
887 main@standard 18 18
888 </pre>
890 <pre>
891 $ echo "vidim" | lt-proc sh-en.automorf.bin | \
893 apertium-transfer apertium-sh-en.trules-sh-en.xml trules-sh-en.bin sh-en.autobil.bin
896 </pre>
897 <p>We get the analysis passed through correctly, but when we try
898 and generate a surface form from this, we get a '#', like
899 below:</p>
900 <pre>
901 $ echo "vidim" | lt-proc sh-en.automorf.bin | \
903 apertium-transfer apertium-sh-en.trules-sh-en.xml trules-sh-en.bin sh-en.autobil.bin | \
904 lt-proc -g sh-en.autogen.bin
905 #see\@
906 </pre>
907 <p>This '#' means that the generator cannot generate the correct
908 lexical form because it does not contain it. Why is this?</p>
909 <p>Basically the analyses don't match, the 'see' in the dictionary
912 Serbo-Croatian side has more information than the English side
913 requires. You can test this by adding the missing symbols to the
914 English dictionary, and then recompiling, and testing again.</p>
915 <p>However, a more paradigmatic way of taking care of this is by
916 writing a rule. So, we open up the rules file
917 (apertium-sh-en.trules-sh-en.xml in case you forgot).</p>
922 <p>We also need to add attributes for tense and for person. We'll
923 make it really simple for now, you can add p2 and p3, but I won't
924 in order to save space.</p>
928 <br />
954 <p>Remember when you tried commenting out the 'clip' tags in the
955 previous rule example and they disappeared from the transfer, well,
956 thats pretty much what we're doing here. We take in a verb with a
957 full analysis, but only output a partial analysis (lemma + verb tag
958 + tense tag).</p>
960 <pre>
961 $ echo "vidim" | lt-proc sh-en.automorf.bin | \
963 apertium-transfer apertium-sh-en.trules-sh-en.xml trules-sh-en.bin sh-en.autobil.bin
965 </pre>
967 <pre>
968 $ echo "vidim" | lt-proc sh-en.automorf.bin | \
970 apertium-transfer apertium-sh-en.trules-sh-en.xml trules-sh-en.bin sh-en.autobil.bin | \
971 lt-proc -g sh-en.autogen.bin
972 see\@
973 </pre>
974 <p>Try it with 'vidimo' (we see) to see if you get the correct
975 output.</p>
977 <pre>
978 $ echo "vidim gramofoni" | lt-proc sh-en.automorf.bin | \
980 apertium-transfer apertium-sh-en.trules-sh-en.xml trules-sh-en.bin sh-en.autobil.bin | \
981 lt-proc -g sh-en.autogen.bin
982 see gramophones\@
983 </pre>
985 <p>Well, thats great, but we're still missing the personal pronoun
986 that is necessary in English. In order to add it in, we first need
987 to edit the English morphological dictionary.</p>
988 <p>As before, the first thing to do is add the necessary
989 symbols:</p>
992 <p>Of the two symbols, prn is pronoun, and subj is subject (as in
993 the subject of a sentence).</p>
994 <p>Because there is no root, or 'lemma' for personal subject
995 pronouns, we just add the pardef as follows:</p>
1006 <p>With 'prsubj' being 'personal subject'. The rest of them (You,
1007 We etc.) are left as an exercise to the reader.</p>
1009 <pre>
1011 </pre>
1012 <p>So, save, recompile and test, and we should get something
1013 like:</p>
1014 <pre>
1015 $ echo "I" | lt-proc en-sh.automorf.bin
1017 </pre>
1019 <p>Now we need to amend the 'verb' rule to output the subject
1020 personal pronoun along with the correct verb form.</p>
1021 <p>First, add a category (this must be getting pretty pedestrian by
1022 now):</p>
1026 <p>Now add the types of pronoun as attributes, we might as well add
1027 the 'obj' type as we're at it, although we won't need to use it for
1028 now:</p>
1041 <lit
1043 <lit-tag
1045 <lit-tag
1064 <p>This is pretty much the same rule as before, only we made a
1065 couple of small changes.</p>
1069 <p>so that the generator could choose the right pronoun and the
1070 right form of the verb.</p>
1072 <ul>
1076 tags from the verb, we add these ourself, "prn" for pronoun, and
1079 </ul>
1080 <p>Note that we retrieved the information for number and tense
1081 directly from the verb.</p>
1083 <pre>
1084 $ echo "vidim gramofone" | lt-proc sh-en.automorf.bin | \
1086 apertium-transfer apertium-sh-en.trules-sh-en.xml trules-sh-en.bin sh-en.autobil.bin | \
1087 lt-proc -g sh-en.autogen.bin
1088 I see gramophones
1089 </pre>
1090 <p>Which, while it isn't exactly prize-winning prose (much like
1091 this HOWTO), is a fairly accurate translation.</p>
1093 <p>While gramophone is an English word, it isn't the best
1094 translation. Gramophone is typically used for the very old kind,
1095 you know with the needle instead of the stylus, and no
1096 amplification. A better translation would be 'record player'.
1097 Although this is more than one word, we can treat it as if it is
1099 constructions.</p>
1100 <p>We don't need to touch the Serbo-Croatian dictionary, just the
1101 English one and the bilingual one this, so open it up.</p>
1102 <p>The plural of 'record player' is 'record players', so it takes
1103 the same paradigm as gramophone (gramophone__n) — in that we just
1104 add 's'. All we need to do is add a new element to the main
1105 section.</p>
1110 tag, although this isn't entirely new as we saw it in use in the
1111 rules file.</p>
1113 <pre>
1114 $ echo "vidim gramofone" | lt-proc sh-en.automorf.bin | \
1116 apertium-transfer apertium-sh-en.trules-sh-en.xml trules-sh-en.bin sh-en.autobil.bin | \
1117 lt-proc -g sh-en.autogen.bin
1118 I see record players
1119 </pre>
1120 <p>Perfect. A big benefit of using multiwords is that you can
1121 translate idiomatic expressions verbatim, without having to do
1122 word-by-word translation.</p>
1123 </body>
1124 </html>