diff --git a/libcorpus2/ann/annotatedsentence.cpp b/libcorpus2/ann/annotatedsentence.cpp
index 6a7bd60e30512e2b33a93b7d3b04060fab79768c..68b561f92d1af65a3412bb6dbf420cf5c1bca184 100644
--- a/libcorpus2/ann/annotatedsentence.cpp
+++ b/libcorpus2/ann/annotatedsentence.cpp
@@ -24,6 +24,27 @@ Sentence::Ptr AnnotatedSentence::clone_shared() const
return copy;
}
+boost::shared_ptr<AnnotatedSentence> AnnotatedSentence::wrap_sentence(
+ const boost::shared_ptr<Sentence>& s)
+{
+ boost::shared_ptr<AnnotatedSentence> a = boost::make_shared<AnnotatedSentence>();
+ foreach (Token* t, s->tokens()) {
+ a->append(t);
+ }
+ s->release_tokens();
+ return a;
+}
+
+boost::shared_ptr<AnnotatedSentence> AnnotatedSentence::wrap_sentence_clone(
+ const boost::shared_ptr<Sentence>& s)
+{
+ boost::shared_ptr<AnnotatedSentence> a = boost::make_shared<AnnotatedSentence>();
+ foreach (Token* t, s->tokens()) {
+ a->append(t->clone());
+ }
+ return a;
+}
+
void AnnotatedSentence::create_channel(const std::string& name)
{
channels_.insert(std::make_pair(name, AnnotationChannel(tokens_.size())));
diff --git a/libcorpus2/ann/annotatedsentence.h b/libcorpus2/ann/annotatedsentence.h
index 86f0d257f0b2464417d63b27de11ccce29381614..db2d4eade8c42a946a8db84071a096f057eba027 100644
--- a/libcorpus2/ann/annotatedsentence.h
+++ b/libcorpus2/ann/annotatedsentence.h
@@ -9,6 +9,9 @@ namespace Corpus2 {
class AnnotationView;
+/**
+ * Exception class for use when a requested annotation channel does not exist
+ */
class MissingAnnotationChannel : public Corpus2Error
{
public:
@@ -22,22 +25,56 @@ public:
}
};
-
+/**
+ * A class describing Sentences with additional information in the form of
+ * annotation channels.
+ *
+ * Note: channels are not automatiaclly resized. The sentence should not have
+ * tokens added or removed after annotation channels are created.
+ */
class AnnotatedSentence : public Corpus2::Sentence
{
public:
+ /**
+ * Create an empty AnnotatedSentece with no tokens and no channels
+ */
AnnotatedSentence();
~AnnotatedSentence();
- Ptr clone_shared() const;
-
+ Sentence::Ptr clone_shared() const;
+
+ /**
+ * Create an AnnotatedSentence from a Sentence, grabing all the tokens
+ * directly (afterwards the source Sentence has no tokens).
+ */
+ static boost::shared_ptr<AnnotatedSentence> wrap_sentence(
+ const boost::shared_ptr<Sentence>& s);
+
+ /**
+ * Create an AnnotatedSentence from a Sentence, cloning all the tokens.
+ * The source Sentence is not modified.
+ */
+ static boost::shared_ptr<AnnotatedSentence> wrap_sentence_clone(
+ const boost::shared_ptr<Sentence>& s);
+
+ /**
+ * Create an annotation channel named name in this annotated sentence.
+ * If the channel already exists, nothing happens.
+ */
void create_channel(const std::string& name);
+ /**
+ * @return true iif this sentence has an annotation channel named name
+ */
bool has_channel(const std::string& name) const {
return channels_.find(name) != channels_.end();
}
+ /**
+ * Return the annotation channel by name or throw MissingAnnotationChannel
+ * if there is no such channel
+ */
AnnotationChannel& get_channel(const std::string& name) {
chan_map_t::iterator i = channels_.find(name);
if (i == channels_.end()) {
@@ -46,6 +83,9 @@ public:
return i->second;
}
+ /**
+ * Const version of get_channel
+ */
const AnnotationChannel& get_channel(const std::string& name) const {
chan_map_t::const_iterator i = channels_.find(name);
if (i == channels_.end()) {
@@ -55,11 +95,20 @@ public:
}
private:
+ /// typedef for tha channels
typedef std::map<std::string, AnnotationChannel> chan_map_t;
+ /// the actual channels
chan_map_t channels_;
};
+/**
+ * Create an AnnotationView pseudo-sentence from an AnnotatedSentence that
+ * behaves like a sentence viewed through an annotation.
+ *
+ * This is a free function, not a member of AnnotatedSentence, because it is
+ * mandatory that the AnnotatedSentence be passed via a shared_ptr.
+ */
boost::shared_ptr<AnnotationView> create_view(
const boost::shared_ptr<AnnotatedSentence>& s,
const std::string& ann_name);
diff --git a/libcorpus2/ann/channel.cpp b/libcorpus2/ann/channel.cpp
index 601b65cbce90e9af28ffe6b25bd299dee7407810..5ff16f92bc8d83b2b5189b67ab0871a5e153a48c 100644
--- a/libcorpus2/ann/channel.cpp
+++ b/libcorpus2/ann/channel.cpp
@@ -48,6 +48,25 @@ void AnnotationChannel::make_segments_from_iob()
}
}
+int AnnotationChannel::renumber_segments()
+{
+ std::map<int, int> re;
+ int next = 0;
+ for (size_t i = 0; i < segments_.size(); ++i) {
+ if (segments_[i] > 0) {
+ std::map<int, int>::const_iterator ci = re.find(segments_[i]);
+ if (ci != re.end()) {
+ segments_[i] = ci->second;
+ } else {
+ ++next;
+ re.insert(std::make_pair(segments_[i], next));
+ segments_[i] = next;
+ }
+ }
+ }
+ return next;
+}
+
IOB::Enum AnnotationChannel::get_iob_at(int idx)
{
if (idx >= 0 && idx < static_cast<int>(iobs_.size())) {
@@ -78,6 +97,7 @@ std::vector<Annotation> AnnotationChannel::make_annotation_vector() const
rv[s].head_index = i;
}
}
+ std::sort(rv.begin(), rv.end(), AnnotationHeadCompare());
rv.erase(std::remove_if(rv.begin(), rv.end(),
boost::bind(&Annotation::empty, _1)));
return rv;
diff --git a/libcorpus2/ann/channel.h b/libcorpus2/ann/channel.h
index eb8618f74db23074d900fc736861146067f72c85..0637638a46b95ecdff866f5a3581b9df8d7d6fdf 100644
--- a/libcorpus2/ann/channel.h
+++ b/libcorpus2/ann/channel.h
@@ -4,47 +4,112 @@
#include <libcorpus2/ann/iob.h>
#include <vector>
#include <boost/utility.hpp>
+#include <algorithm>
+#include <map>
namespace Corpus2 {
+
+/**
+ * A general indexed annotation spanning a possibly disjoint group of
+ * tokens, with a distinguished 'head' token.
+ */
struct Annotation
{
Annotation() : indices(), head_index(-1) {}
std::vector<int> indices;
int head_index;
- bool empty() const { return indices.empty(); }
+ bool empty() const {
+ return indices.empty();
+ }
+ bool sane() const {
+ return empty() ||
+ std::find(indices.begin(), indices.end(), head_index) != indices.end();
+ }
+};
+
+struct AnnotationHeadCompare
+{
+ bool operator()(const Annotation& a1, const Annotation& a2) {
+ return a1.head_index < a2.head_index;
+ }
};
+/**
+ * An annotation channel in IOB format and segment index format, with manual
+ * sync between the two.
+ */
class AnnotationChannel
{
public:
+ /**
+ * Empty ctor
+ */
AnnotationChannel();
+ /**
+ * Ctor for an annotation of a specific size
+ */
explicit AnnotationChannel(int size);
+ /**
+ * Discard IOB annotation information, regenerate it from the segment info.
+ */
void make_iob_from_segments();
+ /**
+ * Discard indexed annotation info, regenerate from IOB.
+ */
void make_segments_from_iob();
+ /**
+ * Number segments from 1 so there are no ommited indices, return the
+ * highest used index
+ */
+ int renumber_segments();
+
+ /**
+ * Create a vector of AnnotationSegment objects, each corresponding to
+ * an annotation, with the annotations possibly being disjoint.
+ */
std::vector<Annotation> make_annotation_vector() const;
+ /**
+ * Create a vector of AnnotationSegment objects, each corresponding to
+ * an annotation, forcing the annotations to be continous (disjoint
+ * annotations are split)
+ */
std::vector<Annotation> make_continuous_annotation_vector() const;
+ /**
+ * The segment-index array accesor
+ */
const std::vector<int>& segments() const {
return segments_;
}
+ /**
+ * The IOB data vector
+ */
const std::vector<IOB::Enum>& iobs() const {
return iobs_;
}
+ /**
+ * IOB getter, returns IOB::O if idx is out of range
+ */
IOB::Enum get_iob_at(int idx);
+ /**
+ * IOB setter, out of range indices are not processed.
+ */
void set_iob_at(int idx, IOB::Enum iob);
private:
+ /// segment indices
std::vector<int> segments_;
+ /// IOB data
std::vector<IOB::Enum> iobs_;
};
diff --git a/libcorpus2/ann/iob.h b/libcorpus2/ann/iob.h
index 68d94f2805e3f6cf65b7d3c492dd419d85099c42..7be4086926855240d7dacdd77f3ff8d2d2bc0f4a 100644
--- a/libcorpus2/ann/iob.h
+++ b/libcorpus2/ann/iob.h
@@ -8,6 +8,10 @@ namespace Corpus2 {
namespace IOB {
/**
* A simple enumeration for IOB annotation chunk tagging
+ * O indicates not part of a chunk
+ * B indicates beginning of a new chunk
+ * I indicates continuation of a chunk started by a preceeding BI* sequence
+ * The only invalid sequence is O followed by I
*/
enum Enum {
O = 0,
@@ -16,6 +20,7 @@ namespace IOB {
PostLast
};
/// Convert an enum value to an uppercase I, O or B string
+ /// Returns ? on invalid value
const char* to_string(Enum iob);
/// Create an enum value from an I, O or B string
Enum from_string(const std::string& s);
diff --git a/libcorpus2/ann/view.cpp b/libcorpus2/ann/view.cpp
index 68047870c59ad1796e8ecd7aa58e673f6d491e61..ac06b1e2a82029130700a2b3fc5371f61426ca9e 100644
--- a/libcorpus2/ann/view.cpp
+++ b/libcorpus2/ann/view.cpp
@@ -25,6 +25,11 @@ Sentence::Ptr AnnotationView::clone_shared() const
return copy;
}
+void AnnotationView::release_original()
+{
+ original_.reset();
+}
+
void AnnotationView::commit()
{
commit_to(original_, ann_name_);
@@ -34,6 +39,9 @@ void AnnotationView::commit_to(
const boost::shared_ptr<AnnotatedSentence> &original,
const std::string &ann_name)
{
+ if (!original) {
+ throw AnnotationViewOutOfSync("null-pointer-to-original");
+ }
if (!original->has_channel(ann_name)) {
throw MissingAnnotationChannel(ann_name);
}
diff --git a/libcorpus2/ann/view.h b/libcorpus2/ann/view.h
index d6790962daa694dd9bcbbcbde9550c01a964cd0a..45e75fecc9739314b74008d1fa8de1ea98217dc9 100644
--- a/libcorpus2/ann/view.h
+++ b/libcorpus2/ann/view.h
@@ -7,9 +7,38 @@ namespace Corpus2 {
class AnnotatedSentence;
+/**
+ * Exception class for signalling OOS between a View and a Sentence during
+ * e.g. commits
+ */
+class AnnotationViewOutOfSync : public Corpus2Error
+{
+public:
+ AnnotationViewOutOfSync(const std::string& reason)
+ : Corpus2Error("AnnotationView out of sync with base sentence: " + reason)
+ {
+ }
+
+ ~AnnotationViewOutOfSync() throw()
+ {
+ }
+};
+
+/**
+ * A class representing an AnnotatedSentence viewed through the segmentation
+ * some Annotation specifies.
+ *
+ * It has its own Tokens but can update the original Sentence via commit().
+ */
class AnnotationView : public Sentence
{
public:
+ /**
+ * Create an AnnotationView from a sentence and an annotation name.
+ * The AnnotationView keeps a handle to the sentence.
+ *
+ * This shoudl not be used directly, use create_view.
+ */
AnnotationView(const boost::shared_ptr<AnnotatedSentence>& original,
const std::string& ann_name_);
@@ -17,10 +46,27 @@ public:
Ptr clone_shared() const;
+ /**
+ * Push changes to the original Sentence.
+ */
void commit();
+ /**
+ * Push changes to a different Sentence, but same channel
+ */
+ void commit_to(const boost::shared_ptr<AnnotatedSentence>& original);
+
+ /**
+ * Push changes to ana arbitary channel in an arbitrary sentence
+ */
void commit_to(const boost::shared_ptr<AnnotatedSentence>& original,
const std::string& ann_name);
+
+ /**
+ * Release the original Sentence. Use this when you no longer expect
+ * to need to go back or commit() changes.
+ */
+ void release_original();
private:
boost::shared_ptr<AnnotatedSentence> original_;
const std::string ann_name_;
diff --git a/tests/CMakeLists.txt b/tests/CMakeLists.txt
index d0f1de97eafde2af57d3fe013ddda8e2e3b2b711..874b60a5a4bb419e23bc7a65fbee11e8ceeb530d 100644
--- a/tests/CMakeLists.txt
+++ b/tests/CMakeLists.txt
@@ -6,6 +6,7 @@ add_definitions(-DLIBCORPUS2_TEST_DATA_DIR="${PROJECT_SOURCE_DIR}/")
add_executable( tests
main.cpp
+ ann_basic.cpp
basic.cpp
tag_split.cpp
tagset_parse.cpp
diff --git a/tests/ann_basic.cpp b/tests/ann_basic.cpp
new file mode 100644
index 0000000000000000000000000000000000000000..e564a39663ca38c948cea3b84d776681cec51802
--- /dev/null
+++ b/tests/ann_basic.cpp
@@ -0,0 +1,45 @@
+/*
+ Copyright (C) 2010 Tomasz Åšniatowski, Adam Radziszewski
+ Part of the libcorpus2 project
+
+ This program is free software; you can redistribute it and/or modify it
+under the terms of the GNU General Public License as published by the Free
+Software Foundation; either version 3 of the License, or (at your option)
+any later version.
+
+ This program is distributed in the hope that it will be useful, but
+WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+or FITNESS FOR A PARTICULAR PURPOSE.
+
+ See the LICENSE and COPYING files for more details.
+*/
+
+#include <boost/test/unit_test.hpp>
+#include <boost/make_shared.hpp>
+#include <libcorpus2/ann/annotatedsentence.h>
+
+BOOST_AUTO_TEST_SUITE(ann)
+
+BOOST_AUTO_TEST_CASE( wrap )
+{
+ Corpus2::Sentence::Ptr s = boost::make_shared<Corpus2::Sentence>();
+ s->append(new Corpus2::Token(UnicodeString::fromUTF8("orth"), PwrNlp::Whitespace::None));
+ Corpus2::Sentence::Ptr a = Corpus2::AnnotatedSentence::wrap_sentence(s);
+ BOOST_CHECK(s->empty());
+ BOOST_REQUIRE(!a->empty());
+ BOOST_CHECK_EQUAL(a->tokens()[0]->orth_utf8(), "orth");
+}
+
+BOOST_AUTO_TEST_CASE( wrap_clone )
+{
+ Corpus2::Sentence::Ptr s = boost::make_shared<Corpus2::Sentence>();
+ s->append(new Corpus2::Token(UnicodeString::fromUTF8("orth"), PwrNlp::Whitespace::None));
+ Corpus2::Sentence::Ptr a = Corpus2::AnnotatedSentence::wrap_sentence_clone(s);
+ BOOST_REQUIRE(!s->empty());
+ s->tokens()[0]->set_orth(UnicodeString::fromUTF8("aaa"));
+ BOOST_REQUIRE(!a->empty());
+ BOOST_CHECK_EQUAL(a->tokens()[0]->orth_utf8(), "orth");
+}
+
+
+BOOST_AUTO_TEST_SUITE_END()