[hltdi-l3] push by onlysk...@gmail.com - LG-LP paper down to 5 pages. on 2014-05-16 07:05 GMT
0 views
Skip to first unread message
hltd...@googlecode.com
May 16, 2014, 3:05:32 AM5/16/14
to hltdi-...@googlegroups.com
Revision: f6c3e2277d61
Branch: default
Author: Michael Gasser <gas...@cs.indiana.edu>
Date: Fri May 16 07:05:16 2014 UTC
Log: LG-LP paper down to 5 pages.
http://code.google.com/p/hltdi-l3/source/detail?r=f6c3e2277d61
Modified:
/hiiktuu/sentence.py
/paperdrafts/lglp/lglp14.pdf
/paperdrafts/lglp/lglp14.tex
/paperdrafts/lglp/mind.png
=======================================
--- /hiiktuu/sentence.py Fri May 16 04:02:37 2014 UTC
+++ /hiiktuu/sentence.py Fri May 16 07:05:16 2014 UTC
@@ -56,6 +56,7 @@
# 2014.05.15
# -- Fixed how group trees are worked out: using the snode->gnodes
variables
# rather than merger-related variables and tree variables
+# -- Search in group selection and output ordering.
import itertools, copy
# ui.py loads language, etc.
=======================================
--- /paperdrafts/lglp/lglp14.pdf Fri May 16 04:02:37 2014 UTC
+++ /paperdrafts/lglp/lglp14.pdf Fri May 16 07:05:16 2014 UTC
Binary file, no diff available.
=======================================
--- /paperdrafts/lglp/lglp14.tex Fri May 16 04:02:37 2014 UTC
+++ /paperdrafts/lglp/lglp14.tex Fri May 16 07:05:16 2014 UTC
@@ -120,16 +120,15 @@
\url{http://creativecommons.org/licenses/by/4.0/}
}
-For the great majority of the world's languages we lack
+For the majority of the world's languages we lack
adequate resources to make use of the machine learning techniques that
-have become the standard for modern computational linguists.
+have become the standard for modern computational linguistics.
Languages with inadequate resources include not only those with few
speakers, many of them endangered, but also a number of Asian and African
languages with tens of
millions of speakers, such as Telugu, Burmese, Oromo, and Hausa.
For machine translation (MT) and computer-assisted translation (CAT),
the lack is even more serious because what is
-required for machine learning is bitext, sentence-aligned translations
between the language
-in question and another language.
+required for machine learning is bitext, sentence-aligned translations.
For these reasons, work on many such languages will continue to
consist in large part in the writing of computational grammars and
@@ -138,21 +137,19 @@
notoriously time-consuming, there is a need for tools to permit
researchers and language technology users to ``get off the ground''
with these languages, that is, to create rudimentary grammars and lexica
that
-will support some basic applications and facilitate the documentation
process.
+will support basic applications and facilitate the language documentation
process.
-We focus on MT and CAT because for most of the languages in question, the
lack of
-linguistic resources correlates with a lack of written material in the
language, and
-we would like to develop tools to aid human translators, including
non-professional ones,
-in translating documents into these languages.
+We focus on MT and CAT because a lack of linguistic resources correlates
with a lack of written material, and
+we would like to develop tools to aid human translators in translating
documents into these languages.
Our long-term goal is a system that allows users with little or no
linguistic experience
to write bilingual lexicon-grammars
-for low-resource languages that can also be updated on the basis of
monolingual and bilingual corpora,
-to the extent these are available, and that can be easily integrated into
a CAT system.
+for low-resource languages that can also be updated on the basis of
corpora,
+when these are available, and that can be easily integrated into a CAT
system.
In this paper we describe the initial steps in developing
Hiiktuu,\footnote{\textit{Hiiktuu} is the Oromo
word for a (female) translator.} a lexical-grammatical framework for MT
and CAT.
Although our focus is on the language pairs Spanish-Guarani and
Amharic-Oromo, we illustrate
-Hiiktuu with examples from English-Spanish in this paper.
+Hiiktuu with examples from English-Spanish.
\section{Lexica and grammars}
@@ -162,21 +159,18 @@
The idea of treating phrases rather than individual words as the basic
units of a language
goes back at least to the proposal of a Phrasal Lexicon by Becker
\shortcite{becker}.
In recent years, the idea has gained currency within the related
frameworks of Construction Grammar \cite{steels}
-and Frame Semantics \cite{fillmoreFS} as well as in phrase-based
statistical machine translation (PBSMT), which
-in one form or another now dominates the MT field.
+and Frame Semantics \cite{fillmoreFS} as well as in phrase-based
statistical machine translation (PBSMT).
Arguments in favor of phrasal units are often framed in terms of the
ubiquity of idiomaticity, that is, departure
-to one degree or another from strict compositionality.
+from strict compositionality.
Seen another way, phrasal units address the ubiquity of lexical ambiguity.
-If a verb's interpretation depends on its object or subject, then it may
make more sense to the combination
+If a verb's interpretation depends on its object or subject, then it may
make more sense to treat the combination
of the verb and particular objects or subjects as units in their own right.
-Arguments based on idiomaticity and ambiguity are semantic; they point out
the advantages of
-phrasal units in systems that deal with meaning.
-But the arguments extend naturally to translation.
+Arguments based on idiomaticity and ambiguity are semantic, but they
extend naturally to translation.
If the meaning of a source-language phrase fails to be the strict
combination of the meanings
of the words in the phrase, then it is unlikely that the translation of
the phrase will be the
combination of the translations of the words.
-Adding lexical context to an ambiguous noun or verb could permit an MT
+Adding lexical context to an ambiguous noun or verb can sometimes permit
an MT
system to select the appropriate translation.
\subsection{A simple phrasal lexicon}
@@ -189,14 +183,13 @@
For example, in the sentence \textit{I gave her a piece of my mind},
\{\textit{I, gave}\} and \{\textit{gave, her, piece}\}
are among the catenae but not the constituents of the sentence.
-A catena has a head, and each Hiiktuu group must also have a head, the
main function of which is to
-index the group within the lexicon.
-A group's entry may also specify translations to groups in one or more
other languages.
-For each translation, the group's entry gives an \textbf{alignment}
between the groups, representing correspondences between
-group elements, as in the phrase tables of PBSMT.
+A catena has a head, and each Hiiktuu group must also have a head, which
indexes the group within the lexicon.
+A group's entry also specifies translations to groups in one or more other
languages.
+For each translation, the group's entry gives an \textbf{alignment},
representing inter-group correspondences between
+elements, as in the phrase tables of PBSMT.
Entry~\ref{entry:end} shows a simple group entry of this sort.
The English group \textit{the end of the world} with head \textit{end} has
as its Spanish translation
-the group \textit{el fin del mundo} (which must have an entry in the
Spanish lexicon).
+the group \textit{el fin del mundo} (which has its own entry in the
Spanish lexicon).
In the alignment, each word other than the fourth word (\textit{the}) in
the English group is associated with the position
of a word in the Spanish group.
@@ -211,7 +204,7 @@
\end{verbatim}
\normalsize
%\end{spacing}
-\caption{Entry for \textit{the end of the world} and its Spanish
translation}
+\caption{Group entry for \textit{the end of the world} and its Spanish
translation}
\label{entry:end}
\end{entry}
@@ -219,47 +212,44 @@
\label{subsect:lexgram}
A rudimentary lexicon with entries like the one in Entry~\ref{entry:end}
is simple
-in two senses: a user with no knowledge of linguistics or the grammar of
either
-language can add entries in a straightforward manner, and the resulting
entries are
+in two senses: a user with no formal knowledge of linguistics can add
entries in a
+straightforward manner, and the resulting entries are
easily understood.
Such a lexicon permits the translation of sentences consisting of verbatim
combinations
-of the word forms in the group entries, as long as group order is
preserved across
+of the wordforms in the group entries, as long as group order is preserved
across
the languages and there are no constraints between groups that would
affect the form
of the target-language words.
-
However, since it contains no grammatical information, such a lexicon
permits no
-{\em generalization} to combinations of wordforms that are not explicit in
group
-lexical entries.
-We are left with the need to include group entries for every reasonably
possible combination of
+{\em generalization} to combinations of wordforms that are not explicit in
the lexicon.
+Such a system would require a group entry for every reasonably possible
combination of
wordforms.
-Even when enormous bitext corpora are available, as for language pairs
like English-Spanish,
-SMT researchers have discovered the need to incorporate some syntax in
their systems.
+%Even for language pairs with enormous available bitext corpora,
+%SMT researchers have discovered the need to incorporate some syntax in
their systems.
-At the other other extreme from this simple lexicon is a full-blown
grammar that is driven by
-the traditional linguistic concern --- one might even say obsession ---
with maximum parsimony:
+At the other extreme from this simple lexicon is a full-blown grammar that
is driven by
+the traditional linguistic concern
+%--- one might even say obsession ---
+with parsimony:
every possible generalization must be ``captured''.
-Although it has the advantage of compactness and of possibly reflecting
general principles
+Although it has the advantage of compactness and of reflecting general
principles
of linguistic structure, such a grammar is difficult
to write, to debug, and to understand, requiring significant knowledge of
linguistics.
-As we have seen, abstract word-based grammars also miss the information
that is inherent
-in words in context.
+%As we have seen, abstract word-based grammars also miss the information
that is inherent
+%in words in context.
In the Hiiktuu project, the goal is to permit a range of possibilities
along the continuum from
purely lexical (and phrasal) to syntactic/grammatical, with the emphasis
on ease of entry
creation and interpretation.
-
-\section{Generalization to grammar}
\subsection{Lexemes}
\label{subsect:lexeme}
-We can achieve significant generalization over simple groups consisting of
wordforms only by
+We can achieve significant generalization over simple groups consisting of
wordforms by
permitting lexemes in groups.
As an example, consider the English group \textit{pass\_v the buck}, where
\textit{pass\_v} is
the verb lexeme \textit{pass}.
-
In order to make such a group usable, the lexicon also requires
\textbf{form} entries,
-giving the lexeme roots as well as grammatical features for specific word
forms.
+giving the lexeme roots as well as grammatical features for specific
wordforms.
Some of these, along with the group entry, are shown in
Entry~\ref{entry:pass}.
\begin{entry}
@@ -269,10 +259,12 @@
groups:
pass_v:
- words: [pass_v, the, buck]
+ spa:
+ - [escurrir_el_bulto,
+ {align: [1,2,3], agr: [{tns: tmp, prs: prs, num: num}, 0, 0]}]
forms:
pass:
- root: pass_v, features: {prs: 1, tns: prs}
- - root: pass_v, features: {prs: 2, tns: prs}
- root: pass_v, features: {prs: 3, num: plr, tns: prs}
- root: pass_n, features: {num: sng}
passes:
@@ -286,64 +278,39 @@
\label{entry:pass}
\end{entry}
-Because the entry for \textit{pass the buck} accommodates multiple
sequences of English word forms,
-there needs to be a way to map these onto the appropriate sequences in the
target language.
-In Hiiktuu, the simplest way to accomplish this is a set of pairs of
agreement features
-for the lexeme that constrain the corresponding target language form to
agree with the source
+Because this entry accommodates multiple sequences of English word forms,
+we need to map these onto appropriate target-language sequences.
+This is accomplished through pairs of agreement features
+for the lexeme, constraining the corresponding target language form to
agree with the source
form on those features.
-In Entry~\ref{entry:passtrans}, we see \textit{agr} attributes for the
translation, none for
-the words \textit{the} and \textit{buck}, and for the head
\textit{pass\_v}, agreement between
-the tense and \textit{tiempo}, person and \textit{persona}, and number and
\textit{nœmero} features.
-For example, if this group is activated in the translation of the sentence
\textit{Carl passes the buck},
-the head of the Spanish translation of the group will be constrained to be
third person singular present tense (tiempo):
+In the example, the
+head \textit{pass\_v} and its translation in the Spanish group agree on
+tense and \textit{tiempo}, person and \textit{persona}, and number and
\textit{nœmero} features.
+For example, if this group is selected in the translation of the sentence
\textit{Carl passes the buck},
+the head of the corresponding Spanish group will be constrained to be
third person singular present tense (tiempo):
\textit{Carl \textbf{escurre} el bulto}.
-
-\begin{entry}
-%\begin{spacing}{.85}
-\small
-\begin{verbatim}
-pass_v:
-- words: [pass_v, the, buck]
- spa:
- - [escurrir_el_bulto,
- {align: [1,2,3], agr: [{tns: tmp, prs: prs, num: num}, 0, 0]}]
-\end{verbatim}
-\normalsize
-%\end{spacing}
-\caption{Group entry for \textit{pass the buck} with Spanish translation}
-\label{entry:passtrans}
-\end{entry}
\subsection{Lexical/grammatical categories}
\label{subsect:cats}
Another simple way to generalize across groups is to introduce syntactic
or semantic categories.
-Consider again the English expression text it{give somebody a piece of
one's mind}.
+Consider the English expression \textit{give somebody a piece of one's
mind}.
We can generalize across specific word sequences such as \textit{gave me a
piece of his mind}
-and \textit{gave them a piece of my mind} by replacing the specific word
forms in positions
+and \textit{gave them a piece of my mind} by replacing the specific
wordforms in positions
2 and 6 in the group with categories that include the wordforms that can
fill those positions.
-This requires the form dictionary to record the categories that wordforms
belong to.
-Entry~\ref{entry:mind} shows how the entry for \textit{give somebody a
piece of one's mind} would
-record this information.
+This requires the forms dictionary to record the categories that wordforms
belong to.
+Entry~\ref{entry:mind} shows how this information would be recorded.
Category names are preceded by \$.
\begin{entry}
-%\begin{spacing}{.85}
\small
\begin{verbatim}
groups:
give_v:
- words: [give_v, $sbd, a, piece, of, $sbds, mind]
agr: [[2, 6, {prs: prs, num: num}]]
- spa:
- - [cantar_a_$algn_las_cuarenta,
- {align: [1,3,4,5,0,0,0],
- agr: [{tns: tmp, prs: prs, num: num}, 0,0,0,0,0,0]}]
my:
- words: [my]
- spa:
- - [mi]
- - [mis]
mayor:
- words: [the, mayor]
forms:
@@ -351,15 +318,14 @@
mayor: [{cats: [$sbd]}]
\end{verbatim}
\normalsize
-%\end{spacing}
-\caption{Group entry for \textit{give somebody a piece of one's mind} and
a few associated form entries}
+\caption{Three group entries and a few associated form entries}
\label{entry:mind}
\end{entry}
Because group positions that are filled by categories do not specify a
surface form,
-for parsing and generation of sentences, they must be merged with other
groups that match
+for parsing and generation of sentences they must be merged with other
groups that match
the category and do specify a form.
-For example, to parse the sentence \textit{I gave the mayor a piece of my
mind} requires
+For example, to parse or translate the sentence \textit{I gave the mayor a
piece of my mind} requires
that positions 2 and 6 in the group
\textit{give\_v\_$sbd\_a\_piece\_of\_$sbds\_mind} be
filled by the heads of the groups \textit{the\_mayor} and \textit{my}.
This \textbf{node merging} process is illustrated in Figure~\ref{fig:mind}.
@@ -377,56 +343,51 @@
\textit{one's} must agree with the subject of the sentence.
Since the group contains no subject, we constrain it to agree with the
person and number
of the verb.
-Thus the entry for this group also contains the agreement attribute:
-\texttt{agr: [[2, 6, [prs, prs], [num, num]]]}.
-This states that the sixth element must agree with the second on person
and number features.
+Thus the entry for this group also contains an agreement attribute
specifying that
+the the sixth element must agree with the second on person and number
features.
\section{Constraint satisfaction and translation}
\label{sect:cs}
Translation in Hiiktuu takes place in three phases: analysis, transfer,
and realization.
-Analysis begins with a lexical lookup of the wordforms in the
source-language sentence.
-The forms dictionary includes roots and grammatical features for some
words.\footnote{In future versions
+Analysis of the source-language sentence begins with a lexical lookup of
the wordforms in the forms dictionary for
+the source language.\footnote{In future versions
of the system, it will be possible to call a morphological analyzer on the
input forms at
this stage.}
-The forms resulting from this first pass are then used to look up
candidate groups in the
-group dictionary.
-Next the system assigns a set of groups to the input sentence, effectively
chunking the sentence.
+The words or lexemes resulting from this first pass are then used to look
up candidate groups in the
+groups dictionary.
+Next the system assigns a set of groups to the input sentence.
A successful group assignment satisfies several constraints: (1)~each word
in the input sentence
-must be assigned to zero, one, or, in the case of node merging, two group
elements.
-(2)~Each element in a selected group must be assigned to one word in the
sentence.
+is assigned to zero, one, or (in the case of node merging) two group
elements.
+(2)~Each element in a selected group is assigned to one word in the
sentence.
(3)~For each selected group, within-group agreement restrictions are
satisfied.
(4)~For each category element in a selected group, it is merged with a
non-category element in another
selected group (see the two examples in Figure~\ref{fig:mind}).
Analysis is a robust process; some words in the input sentence may end up
unassigned to any group.
-Analysis is implemented in the form of constraint satisfaction, making use
of a number of the insights
+Analysis is implemented in the form of constraint satisfaction, making use
of insights
from the Extensive Dependency Grammar framework (XDG) \cite{debusmann}.
-Although considerable source-sentence ambiguity is eliminated because
groups (unless they consist of single words),
-incorporate context, ambiguity is still possible, particularly in the
context of figurative expressions
+Although considerable source-sentence ambiguity is eliminated because
groups
+incorporate context, ambiguity is still possible, particularly for
figurative expressions
that also have a literal interpretation.
In this case, the constraint satisfaction process undertakes a search
through the space of possible group
assignments, creating an analysis for each successful assignment.
-Again this process relies on notions from XDG.
During the transfer phase, a source-language group assignment is converted
to an assignment of target-language
-groups to the sentence.
-In this process some target-language items are assigned grammatical
features on the basis of cross-language
-agreement constraints or within-group agreement constraints in the target
language.
-For example, it is during this stage in the translation of the English
sentence \textit{the mayor passes the buck}
-to Spanish that the Spanish verb assigned to the head of the group
\textit{escurrir el bulto} would be
-assigned the tense (\textit{tiempo}), person and number features
\texttt{tmp=prs, prs=3, num=1}.
-As with within-source ambiguity, some cross-language ambiguity is
eliminated because it is multi-word units
-that are being translated, but a source-language group may still have more
than one translation.
+groups.
+In this process some target-language items are assigned grammatical
features on the basis of agreement constraints.
+For example, in the translation of the English sentence \textit{the mayor
passes the buck},
+the Spanish verb that is the head of the group \textit{escurrir el bulto}
would be
+assigned the tense (\textit{tiempo}), person and number features
\texttt{tmp=prs, prs=3, num=1}: \textit{escurre}.
+A source-language group may have more than one translation.
The transfer phase creates a separate target-language group assignment for
each combination of translations of the
source-language groups.
During the realization phase, for each target-language group assignment,
-target-language surface forms are generated based on the lexemes and
grammatical features that resulted from
+surface forms are generated based on the lexemes and grammatical features
that resulted from
the transfer phase.
In the current version of the system, this is accomplished through a
dictionary that maps
-lexemes and feature sets to surface forms.
-\footnote{In future versions
+lexemes and feature sets to surface forms.\footnote{In future versions
of the system, it will be possible to call a morphological generator at
this stage.}
Finally, target-language words are sequenced in a way that satisfies
word-order
@@ -436,56 +397,48 @@
\section{Related work}
\label{sect:related}
-Our goals are most similar to those of the Apertium \cite{apertium}
project.
-As with Apertium, we are developing open-source, mostly rule-based systems
for MT.
-Also in common with Apertium, we work within the framework of relatively
shallow, chunk grammars.
-We differ mainly in our concern for flexibility, robustness, and
transparency.
-We are willing to sacrifice linguistic coverage and parsimony to achieve
these goals.
-We accommodate a range of lexical/grammatical possibilities, from the
completely
-lexical on the one extreme to phrasal units consisting of a single lexeme
and one or syntactic/semantic
+Our goals are similar to those of the Apertium \cite{apertium} project.
+As with Apertium, we are developing open-source, rule-based systems for
MT, and
+we work within the framework of relatively shallow, chunking grammars.
+We differ mainly in our willingness to sacrifice linguistic coverage
+to achieve our goals of flexibility, robustness, and transparency.
+We accommodate a range of lexical-grammatical possibilities, from the
completely
+lexical on the one extreme to phrasal units consisting of a single lexeme
and one or more syntactic/semantic
categories on the other, and we are not so concerned that Hiiktuu grammars
will accept many ungrammatical source-language
-sentences or even that they will output ungrammatical (along with
grammatical) target-language
+sentences or even output ungrammatical (along with grammatical)
translations.
-With respect to our long-term goals, Hiiktuu also resembles the Expedition
-project \cite{mcshane+nirenburg}, which aims to make use of knowledge
acquisition
-techniques and naive monolingual informants
-in the development of rule-based MT systems that translate low-resource
source languages into Engish.
-Although it is likely we will make use of some of the insights of
Expedition,
-our project differs first, in assuming bilingual informants and second, in
aiming to
+In terms of long-term goals, Hiiktuu also resembles the Expedition
+project \cite{mcshane+nirenburg}, which makes use of knowledge acquisition
+techniques and naive monolingual informants to develop
+MT systems that translate low-resource languages into English.
+Our project differs first, in assuming bilingual informants and second, in
aiming to
develop systems that are unrestricted with respect to target language.
In fact we are more interested in MT systems with low-resource languages
as target languages
because of the lack of documents in such languages.
-Although we would not want Hiiktuu to be taken seriously as a linguistic
theory, it is worth
+Although Hiiktuu is not intended as a linguistic theory, it is worth
mentioning which theories it has the most in common with.
Like Construction Grammar \cite{steels} and Frame Semantics
\cite{fillmoreFS},
it treats linguistic knowledge as essentially phrasal.
Like synchronous context-free grammar (SCFG) \cite{chiang}, it associates
multi-word units in
two languages, aligning the elements of the units and representing word
order within each.
Hiiktuu differs from SCFG in having nothing like rewrite rules or
non-terminals.
-
Hiiktuu belongs to the family of dependency grammar (DG) theories because
the heads of its
phrasal units are words or lexemes rather than non-terminals.
-It has the most in common with those computational DG theories that parse
sentences using
+It shares most with those computational DG theories that rely on
constraint satisfaction \cite{bojar04,debusmann,foth+menzel,wang+harper}.
-However, in its current version, it remains an extremely primitive form of
dependency grammar,
+However, it remains an extremely primitive form of DG,
permitting only flat structures with unlabeled arcs and no relations
between groups
other than through the merge operation described in \ref{subsect:cats}.
This means that complex grammatical phenomena such as long-distance
dependencies and
word-order variability can only be captured through specific groups.
-We prefer this approach because it remains easier for non-linguists to
understand.
-
-Finally, though we have not yet looked into the details,
-the theory's relative simplicity and flexibility should allow it to be
converted
-to other more elaborate formalisms, for example, SCFG.
\section{Status of project, ongoing and future work}
\label{sect:status}
-Hiiktuu is written in Python;
-the code and implemented lexical/grammatical examples
-are available at [\textit{URL omitted from submission to preserve
anonymity}]
+The code for Hiiktuu and a set of lexical-grammatical examples
+are available at [\textit{URL omitted to preserve anonymity}]
under the GPL license.
To date, we have only tested the framework on a limited number of
Amharic-to-Oromo
translations.
@@ -493,7 +446,7 @@
we are currently working on methods for automatically extracting groups
from
dictionaries in various formats and from the limited bilingual data that
are available.
-As a part of this work, it will be crucial to determine whether,
+As a part of this work, it will be crucial to determine whether
it is simpler to extract Hiiktuu groups from data than to extract
grammars of other sorts, for example, SCFG.
We are also implementing a GUI that will allow naive bilingual users to
@@ -504,18 +457,22 @@
for example, when two entries resemble one another, users could be queried
about the
value of collapsing them into a more abstract entry.
-As far as the grammatical framework is concerned, as noted above,
+As far as the grammatical framework is concerned,
the lack of dependencies between the heads of groups leaves the system
-without the capacity to represent agreement constraints, for example,
agreement
-between a verb+object group and a group representing the verb's subject,
+without the capacity to represent some agreement constraints, for example,
agreement
+between a verb+object group and the verb's subject,
or major constituent order differences between source and target
language.\footnote{
The only way to implement such constraints in the current version of
Hiiktuu is through
-larger groups that incorporate, for example, subjects in verb-headed
groups, as in
+groups that incorporate, for example, subjects in verb-headed groups, as in
\textit{\$sbd kick\_v \$sth}.}
-To alleviate this problem, we will be implementing the possibility
-of dependencies between the chunks that are assigned to groups, much as in
the
+To alleviate this problem, we will be implementing
+dependencies between group heads, much as in the
``interchunk module'' of Apertium.
+%Finally, though we have not yet looked into the details,
+%the theory's relative simplicity and flexibility should allow it to be
converted
+%to other more elaborate formalisms, for example, SCFG.
+
\section{Conclusions}
\label{sect:conclusions}
@@ -526,9 +483,7 @@
relatively even more disadvantaged than they were before the digital
revolution.
What is called for are methods that can be quickly and easily deployed to
begin to record the grammars and lexica of these languages and to use these
-tools for the benefit of the linguistic communities, in part to aid in the
translation
-of documents into the languages.
-
+tools for the benefit of the linguistic communities.
The Hiiktuu project is designed with these needs in mind.
Though far from achieving our ultimate goals, we have developed a simple,
flexible, and robust
framework for bilingual lexicon-grammars and MT/CAT that we hope will be a
starting
=======================================
--- /paperdrafts/lglp/mind.png Fri May 16 04:02:37 2014 UTC
+++ /paperdrafts/lglp/mind.png Fri May 16 07:05:16 2014 UTC
Binary file, no diff available.
Branch: default
Author: Michael Gasser <gas...@cs.indiana.edu>
Date: Fri May 16 07:05:16 2014 UTC
Log: LG-LP paper down to 5 pages.
http://code.google.com/p/hltdi-l3/source/detail?r=f6c3e2277d61
Modified:
/hiiktuu/sentence.py
/paperdrafts/lglp/lglp14.pdf
/paperdrafts/lglp/lglp14.tex
/paperdrafts/lglp/mind.png
=======================================
--- /hiiktuu/sentence.py Fri May 16 04:02:37 2014 UTC
+++ /hiiktuu/sentence.py Fri May 16 07:05:16 2014 UTC
@@ -56,6 +56,7 @@
# 2014.05.15
# -- Fixed how group trees are worked out: using the snode->gnodes
variables
# rather than merger-related variables and tree variables
+# -- Search in group selection and output ordering.
import itertools, copy
# ui.py loads language, etc.
=======================================
--- /paperdrafts/lglp/lglp14.pdf Fri May 16 04:02:37 2014 UTC
+++ /paperdrafts/lglp/lglp14.pdf Fri May 16 07:05:16 2014 UTC
Binary file, no diff available.
=======================================
--- /paperdrafts/lglp/lglp14.tex Fri May 16 04:02:37 2014 UTC
+++ /paperdrafts/lglp/lglp14.tex Fri May 16 07:05:16 2014 UTC
@@ -120,16 +120,15 @@
\url{http://creativecommons.org/licenses/by/4.0/}
}
-For the great majority of the world's languages we lack
+For the majority of the world's languages we lack
adequate resources to make use of the machine learning techniques that
-have become the standard for modern computational linguists.
+have become the standard for modern computational linguistics.
Languages with inadequate resources include not only those with few
speakers, many of them endangered, but also a number of Asian and African
languages with tens of
millions of speakers, such as Telugu, Burmese, Oromo, and Hausa.
For machine translation (MT) and computer-assisted translation (CAT),
the lack is even more serious because what is
-required for machine learning is bitext, sentence-aligned translations
between the language
-in question and another language.
+required for machine learning is bitext, sentence-aligned translations.
For these reasons, work on many such languages will continue to
consist in large part in the writing of computational grammars and
@@ -138,21 +137,19 @@
notoriously time-consuming, there is a need for tools to permit
researchers and language technology users to ``get off the ground''
with these languages, that is, to create rudimentary grammars and lexica
that
-will support some basic applications and facilitate the documentation
process.
+will support basic applications and facilitate the language documentation
process.
-We focus on MT and CAT because for most of the languages in question, the
lack of
-linguistic resources correlates with a lack of written material in the
language, and
-we would like to develop tools to aid human translators, including
non-professional ones,
-in translating documents into these languages.
+We focus on MT and CAT because a lack of linguistic resources correlates
with a lack of written material, and
+we would like to develop tools to aid human translators in translating
documents into these languages.
Our long-term goal is a system that allows users with little or no
linguistic experience
to write bilingual lexicon-grammars
-for low-resource languages that can also be updated on the basis of
monolingual and bilingual corpora,
-to the extent these are available, and that can be easily integrated into
a CAT system.
+for low-resource languages that can also be updated on the basis of
corpora,
+when these are available, and that can be easily integrated into a CAT
system.
In this paper we describe the initial steps in developing
Hiiktuu,\footnote{\textit{Hiiktuu} is the Oromo
word for a (female) translator.} a lexical-grammatical framework for MT
and CAT.
Although our focus is on the language pairs Spanish-Guarani and
Amharic-Oromo, we illustrate
-Hiiktuu with examples from English-Spanish in this paper.
+Hiiktuu with examples from English-Spanish.
\section{Lexica and grammars}
@@ -162,21 +159,18 @@
The idea of treating phrases rather than individual words as the basic
units of a language
goes back at least to the proposal of a Phrasal Lexicon by Becker
\shortcite{becker}.
In recent years, the idea has gained currency within the related
frameworks of Construction Grammar \cite{steels}
-and Frame Semantics \cite{fillmoreFS} as well as in phrase-based
statistical machine translation (PBSMT), which
-in one form or another now dominates the MT field.
+and Frame Semantics \cite{fillmoreFS} as well as in phrase-based
statistical machine translation (PBSMT).
Arguments in favor of phrasal units are often framed in terms of the
ubiquity of idiomaticity, that is, departure
-to one degree or another from strict compositionality.
+from strict compositionality.
Seen another way, phrasal units address the ubiquity of lexical ambiguity.
-If a verb's interpretation depends on its object or subject, then it may
make more sense to the combination
+If a verb's interpretation depends on its object or subject, then it may
make more sense to treat the combination
of the verb and particular objects or subjects as units in their own right.
-Arguments based on idiomaticity and ambiguity are semantic; they point out
the advantages of
-phrasal units in systems that deal with meaning.
-But the arguments extend naturally to translation.
+Arguments based on idiomaticity and ambiguity are semantic, but they
extend naturally to translation.
If the meaning of a source-language phrase fails to be the strict
combination of the meanings
of the words in the phrase, then it is unlikely that the translation of
the phrase will be the
combination of the translations of the words.
-Adding lexical context to an ambiguous noun or verb could permit an MT
+Adding lexical context to an ambiguous noun or verb can sometimes permit
an MT
system to select the appropriate translation.
\subsection{A simple phrasal lexicon}
@@ -189,14 +183,13 @@
For example, in the sentence \textit{I gave her a piece of my mind},
\{\textit{I, gave}\} and \{\textit{gave, her, piece}\}
are among the catenae but not the constituents of the sentence.
-A catena has a head, and each Hiiktuu group must also have a head, the
main function of which is to
-index the group within the lexicon.
-A group's entry may also specify translations to groups in one or more
other languages.
-For each translation, the group's entry gives an \textbf{alignment}
between the groups, representing correspondences between
-group elements, as in the phrase tables of PBSMT.
+A catena has a head, and each Hiiktuu group must also have a head, which
indexes the group within the lexicon.
+A group's entry also specifies translations to groups in one or more other
languages.
+For each translation, the group's entry gives an \textbf{alignment},
representing inter-group correspondences between
+elements, as in the phrase tables of PBSMT.
Entry~\ref{entry:end} shows a simple group entry of this sort.
The English group \textit{the end of the world} with head \textit{end} has
as its Spanish translation
-the group \textit{el fin del mundo} (which must have an entry in the
Spanish lexicon).
+the group \textit{el fin del mundo} (which has its own entry in the
Spanish lexicon).
In the alignment, each word other than the fourth word (\textit{the}) in
the English group is associated with the position
of a word in the Spanish group.
@@ -211,7 +204,7 @@
\end{verbatim}
\normalsize
%\end{spacing}
-\caption{Entry for \textit{the end of the world} and its Spanish
translation}
+\caption{Group entry for \textit{the end of the world} and its Spanish
translation}
\label{entry:end}
\end{entry}
@@ -219,47 +212,44 @@
\label{subsect:lexgram}
A rudimentary lexicon with entries like the one in Entry~\ref{entry:end}
is simple
-in two senses: a user with no knowledge of linguistics or the grammar of
either
-language can add entries in a straightforward manner, and the resulting
entries are
+in two senses: a user with no formal knowledge of linguistics can add
entries in a
+straightforward manner, and the resulting entries are
easily understood.
Such a lexicon permits the translation of sentences consisting of verbatim
combinations
-of the word forms in the group entries, as long as group order is
preserved across
+of the wordforms in the group entries, as long as group order is preserved
across
the languages and there are no constraints between groups that would
affect the form
of the target-language words.
-
However, since it contains no grammatical information, such a lexicon
permits no
-{\em generalization} to combinations of wordforms that are not explicit in
group
-lexical entries.
-We are left with the need to include group entries for every reasonably
possible combination of
+{\em generalization} to combinations of wordforms that are not explicit in
the lexicon.
+Such a system would require a group entry for every reasonably possible
combination of
wordforms.
-Even when enormous bitext corpora are available, as for language pairs
like English-Spanish,
-SMT researchers have discovered the need to incorporate some syntax in
their systems.
+%Even for language pairs with enormous available bitext corpora,
+%SMT researchers have discovered the need to incorporate some syntax in
their systems.
-At the other other extreme from this simple lexicon is a full-blown
grammar that is driven by
-the traditional linguistic concern --- one might even say obsession ---
with maximum parsimony:
+At the other extreme from this simple lexicon is a full-blown grammar that
is driven by
+the traditional linguistic concern
+%--- one might even say obsession ---
+with parsimony:
every possible generalization must be ``captured''.
-Although it has the advantage of compactness and of possibly reflecting
general principles
+Although it has the advantage of compactness and of reflecting general
principles
of linguistic structure, such a grammar is difficult
to write, to debug, and to understand, requiring significant knowledge of
linguistics.
-As we have seen, abstract word-based grammars also miss the information
that is inherent
-in words in context.
+%As we have seen, abstract word-based grammars also miss the information
that is inherent
+%in words in context.
In the Hiiktuu project, the goal is to permit a range of possibilities
along the continuum from
purely lexical (and phrasal) to syntactic/grammatical, with the emphasis
on ease of entry
creation and interpretation.
-
-\section{Generalization to grammar}
\subsection{Lexemes}
\label{subsect:lexeme}
-We can achieve significant generalization over simple groups consisting of
wordforms only by
+We can achieve significant generalization over simple groups consisting of
wordforms by
permitting lexemes in groups.
As an example, consider the English group \textit{pass\_v the buck}, where
\textit{pass\_v} is
the verb lexeme \textit{pass}.
-
In order to make such a group usable, the lexicon also requires
\textbf{form} entries,
-giving the lexeme roots as well as grammatical features for specific word
forms.
+giving the lexeme roots as well as grammatical features for specific
wordforms.
Some of these, along with the group entry, are shown in
Entry~\ref{entry:pass}.
\begin{entry}
@@ -269,10 +259,12 @@
groups:
pass_v:
- words: [pass_v, the, buck]
+ spa:
+ - [escurrir_el_bulto,
+ {align: [1,2,3], agr: [{tns: tmp, prs: prs, num: num}, 0, 0]}]
forms:
pass:
- root: pass_v, features: {prs: 1, tns: prs}
- - root: pass_v, features: {prs: 2, tns: prs}
- root: pass_v, features: {prs: 3, num: plr, tns: prs}
- root: pass_n, features: {num: sng}
passes:
@@ -286,64 +278,39 @@
\label{entry:pass}
\end{entry}
-Because the entry for \textit{pass the buck} accommodates multiple
sequences of English word forms,
-there needs to be a way to map these onto the appropriate sequences in the
target language.
-In Hiiktuu, the simplest way to accomplish this is a set of pairs of
agreement features
-for the lexeme that constrain the corresponding target language form to
agree with the source
+Because this entry accommodates multiple sequences of English word forms,
+we need to map these onto appropriate target-language sequences.
+This is accomplished through pairs of agreement features
+for the lexeme, constraining the corresponding target language form to
agree with the source
form on those features.
-In Entry~\ref{entry:passtrans}, we see \textit{agr} attributes for the
translation, none for
-the words \textit{the} and \textit{buck}, and for the head
\textit{pass\_v}, agreement between
-the tense and \textit{tiempo}, person and \textit{persona}, and number and
\textit{nœmero} features.
-For example, if this group is activated in the translation of the sentence
\textit{Carl passes the buck},
-the head of the Spanish translation of the group will be constrained to be
third person singular present tense (tiempo):
+In the example, the
+head \textit{pass\_v} and its translation in the Spanish group agree on
+tense and \textit{tiempo}, person and \textit{persona}, and number and
\textit{nœmero} features.
+For example, if this group is selected in the translation of the sentence
\textit{Carl passes the buck},
+the head of the corresponding Spanish group will be constrained to be
third person singular present tense (tiempo):
\textit{Carl \textbf{escurre} el bulto}.
-
-\begin{entry}
-%\begin{spacing}{.85}
-\small
-\begin{verbatim}
-pass_v:
-- words: [pass_v, the, buck]
- spa:
- - [escurrir_el_bulto,
- {align: [1,2,3], agr: [{tns: tmp, prs: prs, num: num}, 0, 0]}]
-\end{verbatim}
-\normalsize
-%\end{spacing}
-\caption{Group entry for \textit{pass the buck} with Spanish translation}
-\label{entry:passtrans}
-\end{entry}
\subsection{Lexical/grammatical categories}
\label{subsect:cats}
Another simple way to generalize across groups is to introduce syntactic
or semantic categories.
-Consider again the English expression text it{give somebody a piece of
one's mind}.
+Consider the English expression \textit{give somebody a piece of one's
mind}.
We can generalize across specific word sequences such as \textit{gave me a
piece of his mind}
-and \textit{gave them a piece of my mind} by replacing the specific word
forms in positions
+and \textit{gave them a piece of my mind} by replacing the specific
wordforms in positions
2 and 6 in the group with categories that include the wordforms that can
fill those positions.
-This requires the form dictionary to record the categories that wordforms
belong to.
-Entry~\ref{entry:mind} shows how the entry for \textit{give somebody a
piece of one's mind} would
-record this information.
+This requires the forms dictionary to record the categories that wordforms
belong to.
+Entry~\ref{entry:mind} shows how this information would be recorded.
Category names are preceded by \$.
\begin{entry}
-%\begin{spacing}{.85}
\small
\begin{verbatim}
groups:
give_v:
- words: [give_v, $sbd, a, piece, of, $sbds, mind]
agr: [[2, 6, {prs: prs, num: num}]]
- spa:
- - [cantar_a_$algn_las_cuarenta,
- {align: [1,3,4,5,0,0,0],
- agr: [{tns: tmp, prs: prs, num: num}, 0,0,0,0,0,0]}]
my:
- words: [my]
- spa:
- - [mi]
- - [mis]
mayor:
- words: [the, mayor]
forms:
@@ -351,15 +318,14 @@
mayor: [{cats: [$sbd]}]
\end{verbatim}
\normalsize
-%\end{spacing}
-\caption{Group entry for \textit{give somebody a piece of one's mind} and
a few associated form entries}
+\caption{Three group entries and a few associated form entries}
\label{entry:mind}
\end{entry}
Because group positions that are filled by categories do not specify a
surface form,
-for parsing and generation of sentences, they must be merged with other
groups that match
+for parsing and generation of sentences they must be merged with other
groups that match
the category and do specify a form.
-For example, to parse the sentence \textit{I gave the mayor a piece of my
mind} requires
+For example, to parse or translate the sentence \textit{I gave the mayor a
piece of my mind} requires
that positions 2 and 6 in the group
\textit{give\_v\_$sbd\_a\_piece\_of\_$sbds\_mind} be
filled by the heads of the groups \textit{the\_mayor} and \textit{my}.
This \textbf{node merging} process is illustrated in Figure~\ref{fig:mind}.
@@ -377,56 +343,51 @@
\textit{one's} must agree with the subject of the sentence.
Since the group contains no subject, we constrain it to agree with the
person and number
of the verb.
-Thus the entry for this group also contains the agreement attribute:
-\texttt{agr: [[2, 6, [prs, prs], [num, num]]]}.
-This states that the sixth element must agree with the second on person
and number features.
+Thus the entry for this group also contains an agreement attribute
specifying that
+the the sixth element must agree with the second on person and number
features.
\section{Constraint satisfaction and translation}
\label{sect:cs}
Translation in Hiiktuu takes place in three phases: analysis, transfer,
and realization.
-Analysis begins with a lexical lookup of the wordforms in the
source-language sentence.
-The forms dictionary includes roots and grammatical features for some
words.\footnote{In future versions
+Analysis of the source-language sentence begins with a lexical lookup of
the wordforms in the forms dictionary for
+the source language.\footnote{In future versions
of the system, it will be possible to call a morphological analyzer on the
input forms at
this stage.}
-The forms resulting from this first pass are then used to look up
candidate groups in the
-group dictionary.
-Next the system assigns a set of groups to the input sentence, effectively
chunking the sentence.
+The words or lexemes resulting from this first pass are then used to look
up candidate groups in the
+groups dictionary.
+Next the system assigns a set of groups to the input sentence.
A successful group assignment satisfies several constraints: (1)~each word
in the input sentence
-must be assigned to zero, one, or, in the case of node merging, two group
elements.
-(2)~Each element in a selected group must be assigned to one word in the
sentence.
+is assigned to zero, one, or (in the case of node merging) two group
elements.
+(2)~Each element in a selected group is assigned to one word in the
sentence.
(3)~For each selected group, within-group agreement restrictions are
satisfied.
(4)~For each category element in a selected group, it is merged with a
non-category element in another
selected group (see the two examples in Figure~\ref{fig:mind}).
Analysis is a robust process; some words in the input sentence may end up
unassigned to any group.
-Analysis is implemented in the form of constraint satisfaction, making use
of a number of the insights
+Analysis is implemented in the form of constraint satisfaction, making use
of insights
from the Extensive Dependency Grammar framework (XDG) \cite{debusmann}.
-Although considerable source-sentence ambiguity is eliminated because
groups (unless they consist of single words),
-incorporate context, ambiguity is still possible, particularly in the
context of figurative expressions
+Although considerable source-sentence ambiguity is eliminated because
groups
+incorporate context, ambiguity is still possible, particularly for
figurative expressions
that also have a literal interpretation.
In this case, the constraint satisfaction process undertakes a search
through the space of possible group
assignments, creating an analysis for each successful assignment.
-Again this process relies on notions from XDG.
During the transfer phase, a source-language group assignment is converted
to an assignment of target-language
-groups to the sentence.
-In this process some target-language items are assigned grammatical
features on the basis of cross-language
-agreement constraints or within-group agreement constraints in the target
language.
-For example, it is during this stage in the translation of the English
sentence \textit{the mayor passes the buck}
-to Spanish that the Spanish verb assigned to the head of the group
\textit{escurrir el bulto} would be
-assigned the tense (\textit{tiempo}), person and number features
\texttt{tmp=prs, prs=3, num=1}.
-As with within-source ambiguity, some cross-language ambiguity is
eliminated because it is multi-word units
-that are being translated, but a source-language group may still have more
than one translation.
+groups.
+In this process some target-language items are assigned grammatical
features on the basis of agreement constraints.
+For example, in the translation of the English sentence \textit{the mayor
passes the buck},
+the Spanish verb that is the head of the group \textit{escurrir el bulto}
would be
+assigned the tense (\textit{tiempo}), person and number features
\texttt{tmp=prs, prs=3, num=1}: \textit{escurre}.
+A source-language group may have more than one translation.
The transfer phase creates a separate target-language group assignment for
each combination of translations of the
source-language groups.
During the realization phase, for each target-language group assignment,
-target-language surface forms are generated based on the lexemes and
grammatical features that resulted from
+surface forms are generated based on the lexemes and grammatical features
that resulted from
the transfer phase.
In the current version of the system, this is accomplished through a
dictionary that maps
-lexemes and feature sets to surface forms.
-\footnote{In future versions
+lexemes and feature sets to surface forms.\footnote{In future versions
of the system, it will be possible to call a morphological generator at
this stage.}
Finally, target-language words are sequenced in a way that satisfies
word-order
@@ -436,56 +397,48 @@
\section{Related work}
\label{sect:related}
-Our goals are most similar to those of the Apertium \cite{apertium}
project.
-As with Apertium, we are developing open-source, mostly rule-based systems
for MT.
-Also in common with Apertium, we work within the framework of relatively
shallow, chunk grammars.
-We differ mainly in our concern for flexibility, robustness, and
transparency.
-We are willing to sacrifice linguistic coverage and parsimony to achieve
these goals.
-We accommodate a range of lexical/grammatical possibilities, from the
completely
-lexical on the one extreme to phrasal units consisting of a single lexeme
and one or syntactic/semantic
+Our goals are similar to those of the Apertium \cite{apertium} project.
+As with Apertium, we are developing open-source, rule-based systems for
MT, and
+we work within the framework of relatively shallow, chunking grammars.
+We differ mainly in our willingness to sacrifice linguistic coverage
+to achieve our goals of flexibility, robustness, and transparency.
+We accommodate a range of lexical-grammatical possibilities, from the
completely
+lexical on the one extreme to phrasal units consisting of a single lexeme
and one or more syntactic/semantic
categories on the other, and we are not so concerned that Hiiktuu grammars
will accept many ungrammatical source-language
-sentences or even that they will output ungrammatical (along with
grammatical) target-language
+sentences or even output ungrammatical (along with grammatical)
translations.
-With respect to our long-term goals, Hiiktuu also resembles the Expedition
-project \cite{mcshane+nirenburg}, which aims to make use of knowledge
acquisition
-techniques and naive monolingual informants
-in the development of rule-based MT systems that translate low-resource
source languages into Engish.
-Although it is likely we will make use of some of the insights of
Expedition,
-our project differs first, in assuming bilingual informants and second, in
aiming to
+In terms of long-term goals, Hiiktuu also resembles the Expedition
+project \cite{mcshane+nirenburg}, which makes use of knowledge acquisition
+techniques and naive monolingual informants to develop
+MT systems that translate low-resource languages into English.
+Our project differs first, in assuming bilingual informants and second, in
aiming to
develop systems that are unrestricted with respect to target language.
In fact we are more interested in MT systems with low-resource languages
as target languages
because of the lack of documents in such languages.
-Although we would not want Hiiktuu to be taken seriously as a linguistic
theory, it is worth
+Although Hiiktuu is not intended as a linguistic theory, it is worth
mentioning which theories it has the most in common with.
Like Construction Grammar \cite{steels} and Frame Semantics
\cite{fillmoreFS},
it treats linguistic knowledge as essentially phrasal.
Like synchronous context-free grammar (SCFG) \cite{chiang}, it associates
multi-word units in
two languages, aligning the elements of the units and representing word
order within each.
Hiiktuu differs from SCFG in having nothing like rewrite rules or
non-terminals.
-
Hiiktuu belongs to the family of dependency grammar (DG) theories because
the heads of its
phrasal units are words or lexemes rather than non-terminals.
-It has the most in common with those computational DG theories that parse
sentences using
+It shares most with those computational DG theories that rely on
constraint satisfaction \cite{bojar04,debusmann,foth+menzel,wang+harper}.
-However, in its current version, it remains an extremely primitive form of
dependency grammar,
+However, it remains an extremely primitive form of DG,
permitting only flat structures with unlabeled arcs and no relations
between groups
other than through the merge operation described in \ref{subsect:cats}.
This means that complex grammatical phenomena such as long-distance
dependencies and
word-order variability can only be captured through specific groups.
-We prefer this approach because it remains easier for non-linguists to
understand.
-
-Finally, though we have not yet looked into the details,
-the theory's relative simplicity and flexibility should allow it to be
converted
-to other more elaborate formalisms, for example, SCFG.
\section{Status of project, ongoing and future work}
\label{sect:status}
-Hiiktuu is written in Python;
-the code and implemented lexical/grammatical examples
-are available at [\textit{URL omitted from submission to preserve
anonymity}]
+The code for Hiiktuu and a set of lexical-grammatical examples
+are available at [\textit{URL omitted to preserve anonymity}]
under the GPL license.
To date, we have only tested the framework on a limited number of
Amharic-to-Oromo
translations.
@@ -493,7 +446,7 @@
we are currently working on methods for automatically extracting groups
from
dictionaries in various formats and from the limited bilingual data that
are available.
-As a part of this work, it will be crucial to determine whether,
+As a part of this work, it will be crucial to determine whether
it is simpler to extract Hiiktuu groups from data than to extract
grammars of other sorts, for example, SCFG.
We are also implementing a GUI that will allow naive bilingual users to
@@ -504,18 +457,22 @@
for example, when two entries resemble one another, users could be queried
about the
value of collapsing them into a more abstract entry.
-As far as the grammatical framework is concerned, as noted above,
+As far as the grammatical framework is concerned,
the lack of dependencies between the heads of groups leaves the system
-without the capacity to represent agreement constraints, for example,
agreement
-between a verb+object group and a group representing the verb's subject,
+without the capacity to represent some agreement constraints, for example,
agreement
+between a verb+object group and the verb's subject,
or major constituent order differences between source and target
language.\footnote{
The only way to implement such constraints in the current version of
Hiiktuu is through
-larger groups that incorporate, for example, subjects in verb-headed
groups, as in
+groups that incorporate, for example, subjects in verb-headed groups, as in
\textit{\$sbd kick\_v \$sth}.}
-To alleviate this problem, we will be implementing the possibility
-of dependencies between the chunks that are assigned to groups, much as in
the
+To alleviate this problem, we will be implementing
+dependencies between group heads, much as in the
``interchunk module'' of Apertium.
+%Finally, though we have not yet looked into the details,
+%the theory's relative simplicity and flexibility should allow it to be
converted
+%to other more elaborate formalisms, for example, SCFG.
+
\section{Conclusions}
\label{sect:conclusions}
@@ -526,9 +483,7 @@
relatively even more disadvantaged than they were before the digital
revolution.
What is called for are methods that can be quickly and easily deployed to
begin to record the grammars and lexica of these languages and to use these
-tools for the benefit of the linguistic communities, in part to aid in the
translation
-of documents into the languages.
-
+tools for the benefit of the linguistic communities.
The Hiiktuu project is designed with these needs in mind.
Though far from achieving our ultimate goals, we have developed a simple,
flexible, and robust
framework for bilingual lexicon-grammars and MT/CAT that we hope will be a
starting
=======================================
--- /paperdrafts/lglp/mind.png Fri May 16 04:02:37 2014 UTC
+++ /paperdrafts/lglp/mind.png Fri May 16 07:05:16 2014 UTC
Binary file, no diff available.
Reply all
Reply to author
Forward
0 new messages