def convert_doc_to_latex(doc, verbatim_strings=[], use_gold_trees=False):
"""
This function expects a list of ccg_trees, and a list of tokens
(as produced by transccg). Then, it converts each pair (ccg_tree, ccg_tokens)
into a presentation MathML string, and wraps them with HTML code.
verbatim_strings contains a list of strings that should be printed
verbatim at the end of the HTML document, for debugging.
"""
ccg_trees = []
if use_gold_trees:
for sentence in doc.xpath('//sentence'):
gold_tree_index = int(sentence.get('gold_tree', '0'))
ccg_trees.append(sentence.xpath('./ccg')[gold_tree_index])
ccg_trees = [build_ccg_tree(c) for c in ccg_trees]
else:
ccg_trees = [build_ccg_tree(c) for c in doc.xpath('//sentence/ccg[1]')]
sem_trees = [build_ccg_tree(c) for c in doc.xpath('//semantics')]
if not sem_trees:
sem_trees = [None] * len(ccg_trees)
tokens = doc.xpath('//tokens')
assert len(ccg_trees) == len(tokens)
num_hypotheses = len(ccg_trees) - 1
sentence_ids = [
"Premise {0}: ".format(i + 1) for i in range(num_hypotheses)
]
sentence_ids.append("Conclusion: ")
latex_str = ""
for i in range(len(ccg_trees)):
sentence_surface = ' '.join(tokens[i].xpath('token/@surf'))
latex_str += "\n\n\\vspace{2em}\n\n\\noindent\n" + sentence_ids[i] + sentence_surface + "\n\n\medskip\n\n" \
+ convert_node_to_latex(ccg_trees[i], sem_trees[i], tokens[i])
verbatim_text = ""
if verbatim_strings:
verbatim_text = "<p>Script piped to coq</p>"
for vb_str in verbatim_strings:
verbatim_text += "<pre>\n" + vb_str + "\n</pre>\n"
html_str = "\documentclass{article}\n" \
+ "\\usepackage{proof,lscape}\n" \
+ "\\pagestyle{empty}\n" \
+ "\\newcommand{\\rulelabelsize}{\\scriptsize}\n" \
+ "\\newcommand{\\FA}{\\mbox{\\rulelabelsize $<$}}\n" \
+ "\\newcommand{\\BA}{\\mbox{\\rulelabelsize $>$}}\n" \
+ "\\newcommand{\\FC}{\\mbox{\\rulelabelsize $>\\!\\mathbf{B}$}}\n" \
+ "\\newcommand{\\BC}{\\mbox{\\rulelabelsize $>\\!\\mathbf{B}$}}\n" \
+ "\\newcommand{\\SR}[1]{\\begin{tabular}{c}$#1$\\end{tabular}}\n" \
+ "\\begin{document}\n" \
+ "\\small\n" \
html_str += "%\\begin{landscape}"
html_str += latex_str
html_str += verbatim_text
html_str += "\n%\\end{landscape}"
html_str += "\n\\end{document}"
return html_str
def convert_doc_to_mathml(doc, use_gold_trees=False):
"""
This function expects an XML <document>, which is then converted
into a presentation MathML string.
"""
num_sentences = int(doc.xpath('count(./sentences/sentence)'))
mathml_str = ""
for sent_ind, sentence in enumerate(doc.xpath('./sentences/sentence')):
gold_tree_index = int(sentence.get('gold_tree', -1))
if sent_ind < num_sentences - 1:
sentence_label = 'Premise {0}'.format(sent_ind)
else:
sentence_label = 'Conclusion'
sentence_text = get_surf_from_xml_node(sentence)
ccg_trees = sentence.xpath('./ccg')
sem_trees = sentence.xpath('./semantics')
tokens = sentence.xpath('./tokens')
if not tokens:
return mathml_str
tokens = tokens[0]
assert len(ccg_trees) >= len(sem_trees)
for i in range(len(ccg_trees)):
ccg_tree_id = ccg_trees[i].get('id', str(i))
try:
ccg_tree = build_ccg_tree(ccg_trees[i])
except ValueError:
mathml_str += "<p>{0}, tree {1}: {2}</p>\n".format(
sentence_label, ccg_tree_id, sentence_text) \
+ "<p>Syntactic parse error. Visualization skipped.</p>"
continue
if gold_tree_index == i:
ccg_tree_id += " (gold)"
sem_tree = None if i >= len(sem_trees) else sem_trees[i]
if sem_tree is not None:
sem_tree = build_ccg_tree(sem_tree)
mathml_str += "<p>{0}, tree {1}: {2}</p>\n".format(
sentence_label, ccg_tree_id, sentence_text) \
+ "<math xmlns='http://www.w3.org/1998/Math/MathML'>\n" \
+ convert_node_to_mathml(ccg_tree, sem_tree, tokens) \
+ "</math>\n"
verbatim_strings = doc.xpath(
'./proof/master_theorem/theorems/theorem/coq_script/text()')
verbatim_text = ""
if verbatim_strings:
verbatim_text = "<p>Script piped to coq</p>"
for vb_str in verbatim_strings:
verbatim_text += "<pre>\n" + vb_str + "\n</pre>\n"
doc_mathml_str = '{0}\n{1}'.format(mathml_str, verbatim_text)
return doc_mathml_str
def test_NonterminalButpreterminal(self):
sentence_str = r"""
<sentence id="s1">
<tokens>
<token surf="surf1" id="t1_1"/>
<token surf="surf2" id="t1_2"/>
</tokens>
<ccg root="sp1-5">
<span terminal="t1_1" category="cat1" id="sp1-1"/>
<span terminal="t1_2" category="cat2" id="sp1-2"/>
<span child="sp1-1" rule="lex1" category="NP1" id="sp1-3"/>
<span child="sp1-2" rule="lex2" category="NP2" id="sp1-4"/>
<span child="sp1-3 sp1-4" rule="rr" category="NPP" id="sp1-5"/>
</ccg>
</sentence>
"""
sentence = etree.fromstring(sentence_str)
ccg_tree = sentence.find("ccg")
ccg_root = build_ccg_tree(ccg_tree)
tokens = sentence.find("tokens")
attributes = get_attributes_from_ccg_node_recursively(ccg_root[1], tokens)
expected_attributes = {'category' : 'NP2',
'rule' : 'lex2',
'id' : 'sp1-4',
'child' : 'sp1-2',
'child0_terminal' : 't1_2',
'child0_surf' : 'surf2',
'child0_category' : 'cat2',
'child0_id' : 'sp1-2'}
self.assertEqual(len(expected_attributes), len(attributes),
'\n{0}\nvs.\n{1}'.format(expected_attributes, attributes))
for k in expected_attributes:
self.assertEqual(expected_attributes.get(k, None), attributes.get(k, None))
def test_nonterminal1(self):
sentence_str = r"""
<sentence id="s1">
<tokens>
<token surf="surf1" id="t1_1"/>
</tokens>
<ccg root="sp1-3">
<span terminal="t1_1" category="cat1" id="sp1-1"/>
<span child="sp1-1" rule="lex" category="NP" id="sp1-2"/>
<span child="sp1-2" rule="rr" category="NPP" id="sp1-3"/>
</ccg>
</sentence>
"""
sentence = etree.fromstring(sentence_str)
ccg_tree = sentence.find("ccg")
ccg_root = build_ccg_tree(ccg_tree)
tokens = sentence.find("tokens")
attributes = get_attributes_from_ccg_node_recursively(ccg_root, tokens)
expected_attributes = {'category' : 'NPP',
'rule' : 'rr',
'id' : 'sp1-3',
'child' : 'sp1-2',
'child0_category' : 'NP',
'child0_rule' : 'lex',
'child0_id' : 'sp1-2',
'child0_child' : 'sp1-1',
'child0_child0_terminal' : 't1_1',
'child0_child0_surf' : 'surf1',
'child0_child0_category' : 'cat1',
'child0_child0_id' : 'sp1-1'}
self.assertEqual(len(expected_attributes), len(attributes))
for k in expected_attributes:
self.assertEqual(expected_attributes.get(k, None), attributes.get(k, None))
def test_terminal(self):
sentence_str = r"""
<sentence id="s1">
<tokens>
<token base="base1" pos="pos1" surf="surf1" id="t1_1"/>
</tokens>
<ccg root="sp1-1">
<span terminal="t1_1" category="cat1" end="2" begin="1" id="sp1-1"/>
</ccg>
</sentence>
"""
sentence = etree.fromstring(sentence_str)
ccg_tree = sentence.find("ccg")
ccg_root = build_ccg_tree(ccg_tree)
tokens = sentence.find("tokens")
attributes = get_attributes_from_ccg_node_recursively(ccg_root, tokens)
expected_attributes = {'terminal' : 't1_1',
'category' : 'cat1',
'end' : '2',
'begin' : '1',
'id' : 'sp1-1',
'base' : 'base1',
'pos' : 'pos1',
'surf' : 'surf1'}
self.assertEqual(len(expected_attributes), len(attributes))
for k in expected_attributes:
self.assertEqual(expected_attributes[k], attributes[k])
def test_preterminal(self):
sentence_str = r"""
<sentence id="s1">
<tokens>
<token surf="surf1" id="t1_1"/>
</tokens>
<ccg root="sp1-2">
<span terminal="t1_1" category="cat1" id="sp1-1"/>
<span child="sp1-1" rule="lex" category="NP" id="sp1-2"/>
</ccg>
</sentence>
"""
sentence = etree.fromstring(sentence_str)
ccg_tree = sentence.find("ccg")
ccg_root = build_ccg_tree(ccg_tree)
tokens = sentence.find("tokens")
attributes = get_attributes_from_ccg_node_recursively(ccg_root, tokens)
expected_attributes = {
'category': 'NP',
'rule': 'lex',
'id': 'sp1-2',
'child': 'sp1-1',
'child0_terminal': 't1_1',
'child0_surf': 'surf1',
'child0_category': 'cat1',
'child0_id': 'sp1-1'
}
self.assertEqual(len(expected_attributes), len(attributes))
for k in expected_attributes:
self.assertEqual(expected_attributes.get(k, None),
attributes.get(k, None))
def test_terminal(self):
sentence_str = r"""
<sentence id="s1">
<tokens>
<token base="base1" pos="pos1" surf="surf1" id="t1_1"/>
</tokens>
<ccg root="sp1-1">
<span terminal="t1_1" category="cat1" end="2" begin="1" id="sp1-1"/>
</ccg>
</sentence>
"""
sentence = etree.fromstring(sentence_str)
ccg_tree = sentence.find("ccg")
ccg_root = build_ccg_tree(ccg_tree)
tokens = sentence.find("tokens")
attributes = get_attributes_from_ccg_node_recursively(ccg_root, tokens)
expected_attributes = {
'terminal': 't1_1',
'category': 'cat1',
'end': '2',
'begin': '1',
'id': 'sp1-1',
'base': 'base1',
'pos': 'pos1',
'surf': 'surf1'
}
self.assertEqual(len(expected_attributes), len(attributes))
for k in expected_attributes:
self.assertEqual(expected_attributes[k], attributes[k])
def test_nonterminal2(self):
sentence_str = r"""
<sentence id="s1">
<tokens>
<token surf="surf1" id="t1_1"/>
<token surf="surf2" id="t1_2"/>
</tokens>
<ccg root="sp1-5">
<span terminal="t1_1" category="cat1" id="sp1-1"/>
<span terminal="t1_2" category="cat2" id="sp1-2"/>
<span child="sp1-1" rule="lex1" category="NP1" id="sp1-3"/>
<span child="sp1-2" rule="lex2" category="NP2" id="sp1-4"/>
<span child="sp1-3 sp1-4" rule="rr" category="NPP" id="sp1-5"/>
</ccg>
</sentence>
"""
sentence = etree.fromstring(sentence_str)
ccg_tree = sentence.find("ccg")
ccg_root = build_ccg_tree(ccg_tree)
tokens = sentence.find("tokens")
attributes = get_attributes_from_ccg_node_recursively(ccg_root, tokens)
expected_attributes = {
'category': 'NPP',
'rule': 'rr',
'id': 'sp1-5',
'child': 'sp1-3 sp1-4',
'child0_category': 'NP1',
'child0_rule': 'lex1',
'child0_id': 'sp1-3',
'child0_child': 'sp1-1',
'child0_child0_terminal': 't1_1',
'child0_child0_surf': 'surf1',
'child0_child0_category': 'cat1',
'child0_child0_id': 'sp1-1',
'child1_category': 'NP2',
'child1_rule': 'lex2',
'child1_id': 'sp1-4',
'child1_child': 'sp1-2',
'child1_child0_terminal': 't1_2',
'child1_child0_surf': 'surf2',
'child1_child0_category': 'cat2',
'child1_child0_id': 'sp1-2'
}
self.assertEqual(
len(expected_attributes), len(attributes),
'\n{0}\nvs.\n{1}'.format(expected_attributes, attributes))
for k in expected_attributes:
self.assertEqual(expected_attributes.get(k, None),
attributes.get(k, None))
def convert_vertical_to_mathml(doc, verbatim_strings=[], use_gold_trees=False):
"""
This function expects a list of ccg_trees, and a list of tokens
(as produced by transccg). Then, it converts each pair (ccg_tree, ccg_tokens)
into a presentation MathML string, and wraps them with HTML code.
verbatim_strings contains a list of strings that should be printed
verbatim at the end of the HTML document, for debugging.
"""
ccg_trees = []
if use_gold_trees:
for sentence in doc.xpath('//sentence'):
gold_tree_index = int(sentence.get('gold_tree', '0'))
ccg_trees.append(sentence.xpath('./ccg')[gold_tree_index])
ccg_trees = [build_ccg_tree(c) for c in ccg_trees]
else:
ccg_trees = [build_ccg_tree(c) for c in doc.xpath('//sentence/ccg[1]')]
sem_trees = [build_ccg_tree(c) for c in doc.xpath('//semantics')]
if not sem_trees:
sem_trees = [None] * len(ccg_trees)
tokens = doc.xpath('//tokens')
assert len(ccg_trees) == len(tokens)
num_hypotheses = len(ccg_trees) - 1
sentence_ids = [
"Premise {0}: ".format(i + 1) for i in range(num_hypotheses)
]
sentence_ids.append("Conclusion: ")
mathml_str = ""
for i in range(len(ccg_trees)):
sentence_surface = ' '.join(tokens[i].xpath('token/@surf'))
mathml_str += "<p>" + sentence_ids[i] + sentence_surface + "</p>\n" \
+ "<math xmlns='http://www.w3.org/1998/Math/MathML'>" \
+ convert_node_to_mathml(ccg_trees[i], sem_trees[i], tokens[i]) \
+ "</math>"
verbatim_text = ""
if verbatim_strings:
verbatim_text = "<p>Script piped to coq</p>"
for vb_str in verbatim_strings:
verbatim_text += "<pre>\n" + vb_str + "\n</pre>\n"
html_str = """\
<!doctype html>
<html lang='en'>
<head>
<meta charset='UTF-8'>
<title>CCG to Lambda conversion</title>
<style>
body {
font-size: 1em;
}
</style>
<script type="text/javascript"
src="http://cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML">
</script>
<script type="text/x-mathjax-config">
MathJax.Hub.Config({
tex2jax: {
inlineMath: [['$','$'], ['\\(','\\)']],
processEscapes: true
},
CommonHTML: { matchFontHeight: false },
displayAlign: "left",
displayIndent: "2em"
});
MathJax.Hub.Config({
"HTML-CSS": {
availableFonts: [],
preferredFont: null,
webFont: "Neo-Euler"
}
});
</script>
</head>
<body>
"""
html_str += mathml_str
html_str += verbatim_text
html_str += """\
</body>
</html>
"""
return html_str
def convert_doc_to_mathml_(doc, verbatim_strings=[], use_gold_trees=False):
"""
This function expects a list of ccg_trees, and a list of tokens
(as produced by transccg). Then, it converts each pair (ccg_tree, ccg_tokens)
into a presentation MathML string, and wraps them with HTML code.
verbatim_strings contains a list of strings that should be printed
verbatim at the end of the HTML document, for debugging.
"""
ccg_trees = []
if use_gold_trees:
for sentence in doc.xpath('./sentences/sentence'):
gold_tree_index = int(sentence.get('gold_tree', '0'))
ccg_trees.append(sentence.xpath('./ccg')[gold_tree_index])
ccg_trees = [build_ccg_tree(c) for c in ccg_trees]
else:
ccg_trees = [
build_ccg_tree(c) for c in doc.xpath('./sentences/sentence/ccg[1]')
]
sem_trees = [
build_ccg_tree(c) for c in doc.xpath('./sentences/sentence/semantics')
]
if not sem_trees:
sem_trees = [None] * len(ccg_trees)
tokens = doc.xpath('./sentences/sentence/tokens')
assert len(ccg_trees) == len(tokens)
num_hypotheses = len(ccg_trees) - 1
sentence_ids = [
"Premise {0}: ".format(i + 1) for i in range(num_hypotheses)
]
sentence_ids.append("Conclusion: ")
mathml_str = ""
for i in range(len(ccg_trees)):
sentence_surface = ' '.join(tokens[i].xpath('token/@surf'))
mathml_str += "<p>" + sentence_ids[i] + sentence_surface + "</p>\n" \
+ "<math xmlns='http://www.w3.org/1998/Math/MathML'>" \
+ convert_node_to_mathml(ccg_trees[i], sem_trees[i], tokens[i]) \
+ "</math>"
verbatim_text = ""
if verbatim_strings:
verbatim_text = "<p>Script piped to coq</p>"
for vb_str in verbatim_strings:
verbatim_text += "<pre>\n" + vb_str + "\n</pre>\n"
html_str = """\
<!doctype html>
<html lang='en'>
<head>
<style>
body {
font-size: 1em;
}
</style>
<meta charset='UTF-8'>
<title>CCG to Lambda conversion</title>
<script type="text/javascript"
src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.1/MathJax.js?config=TeX-AMS-MML_HTMLorMML">
</script>
</head>
<body>
"""
html_str += mathml_str
html_str += verbatim_text
html_str += """\
</body>
</html>
"""
return cgi.escape(html_str)
def convert_root_to_mathml(root, verbatim_strings=[], use_gold_trees=False):
"""
This function expects an XML root. Then, it converts each document doc
into a presentation MathML string, and wraps them with HTML code.
verbatim_strings is a list of strings that should be printed verbatim at
the end of the HTML document, for debugging.
"""
doc_mathml_strs = []
for doc_ind, doc in enumerate(root.xpath('./document')):
doc_id = doc.get('id', doc_ind)
num_sentences = int(doc.xpath('count(./sentences/sentence)'))
mathml_str = ""
for sent_ind, sentence in enumerate(doc.xpath('./sentences/sentence')):
gold_tree_index = int(sentence.get('gold_tree', -1))
if sent_ind < num_sentences - 1:
sentence_label = 'Premise {0}'.format(sent_ind)
else:
sentence_label = 'Conclusion'
sentence_text = get_surf_from_xml_node(sentence)
ccg_trees = sentence.xpath('./ccg')
sem_trees = sentence.xpath('./semantics')
tokens = sentence.xpath('./tokens')[0]
assert len(ccg_trees) >= len(sem_trees)
for i in range(len(ccg_trees)):
ccg_tree_id = ccg_trees[i].get('id', str(i))
ccg_tree = build_ccg_tree(ccg_trees[i])
if gold_tree_index == i:
ccg_tree_id += " (gold)"
sem_tree = None if i >= len(sem_trees) else sem_trees[i]
if sem_tree is not None:
sem_tree = build_ccg_tree(sem_tree)
mathml_str += "<p>{0}, tree {1}: {2}</p>\n".format(
sentence_label, ccg_tree_id, sentence_text) \
+ "<math xmlns='http://www.w3.org/1998/Math/MathML'>\n" \
+ convert_node_to_mathml(ccg_tree, sem_tree, tokens) \
+ "</math>\n"
verbatim_text = ""
if verbatim_strings:
verbatim_text = "<p>Script piped to coq</p>"
for vb_str in verbatim_strings:
verbatim_text += "<pre>\n" + vb_str + "\n</pre>\n"
doc_mathml_str = '{0}\n{1}'.format(mathml_str, verbatim_text)
doc_mathml_strs.append(doc_mathml_str)
html_str = """\
<!doctype html>
<html lang='en'>
<head>
<meta charset='UTF-8'>
<title>CCG to Lambda conversion</title>
<style>
body {
font-size: 1em;
}
</style>
<script type="text/javascript"
src="http://cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML">
</script>
</head>
<body>
"""
html_str += '\n'.join([s for s in doc_mathml_strs])
html_str += """\
</body>
</html>
"""
return html_str
def convert_doc_to_mathml(doc, verbatim_strings = [], use_gold_trees=False):
"""
This function expects a list of ccg_trees, and a list of tokens
(as produced by transccg). Then, it converts each pair (ccg_tree, ccg_tokens)
into a presentation MathML string, and wraps them with HTML code.
verbatim_strings contains a list of strings that should be printed
verbatim at the end of the HTML document, for debugging.
"""
ccg_trees = []
if use_gold_trees:
for sentence in doc.xpath('//sentence'):
gold_tree_index = int(sentence.get('gold_tree', '0'))
ccg_trees.append(sentence.xpath('./ccg')[gold_tree_index])
ccg_trees = [build_ccg_tree(c) for c in ccg_trees]
else:
ccg_trees = [build_ccg_tree(c) for c in doc.xpath('//sentence/ccg[1]')]
sem_trees = [build_ccg_tree(c) for c in doc.xpath('//semantics')]
if not sem_trees:
sem_trees = [None] * len(ccg_trees)
tokens = doc.xpath('//tokens')
assert len(ccg_trees) == len(tokens)
num_hypotheses = len(ccg_trees) - 1
sentence_ids = ["Premise {0}: ".format(i + 1) for i in range(num_hypotheses)]
sentence_ids.append("Conclusion: ")
mathml_str = ""
for i in range(len(ccg_trees)):
sentence_surface = ' '.join(tokens[i].xpath('token/@surf'))
mathml_str += "<p>" + sentence_ids[i] + sentence_surface + "</p>\n" \
+ "<math xmlns='http://www.w3.org/1998/Math/MathML'>" \
+ convert_node_to_mathml(ccg_trees[i], sem_trees[i], tokens[i]) \
+ "</math>"
verbatim_text = ""
if verbatim_strings:
verbatim_text = "<p>Script piped to coq</p>"
for vb_str in verbatim_strings:
verbatim_text += "<pre>\n" + vb_str + "\n</pre>\n"
html_str = """\
<!doctype html>
<html lang='en'>
<head>
<meta charset='UTF-8'>
<title>CCG to Lambda conversion</title>
<style>
body {
font-size: 1em;
}
</style>
<script type="text/javascript"
src="http://cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML">
</script>
</head>
<body>
"""
html_str += mathml_str
html_str += verbatim_text
html_str += """\
</body>
</html>
"""
return html_str
