def subtrees(tree):
#setOfTrees = set()
#setOfSubTrees = set()
setOfTrees = []
setOfSubTrees = []
if tree.isPreTerminal():
tree.children[0].wasTerminal = True
#setOfTrees.add(tree)
#setOfSubTrees.add(tree)
setOfTrees.append(tree)
setOfSubTrees.append(tree)
else:
baseChildrenList = [[]]
for child in tree.children:
newBaseChildrenList = []
c_setOfTrees, c_setOfSubtrees = DT.subtrees(child)
#setOfSubTrees = setOfSubTrees.union(c_setOfSubtrees)
setOfSubTrees = setOfSubTrees + c_setOfSubtrees
for treeSubChild in c_setOfTrees:
for treeSub in baseChildrenList:
newBaseChildrenList.append(treeSub + [treeSubChild])
baseChildrenList = newBaseChildrenList
for children in baseChildrenList:
newTree = Tree(root=tree.root,children=children, id=tree.id())
#setOfTrees.add(newTree)
#setOfSubTrees.add(newTree)
setOfTrees.append(newTree)
setOfSubTrees.append(newTree)
#setOfTrees.add(Tree(root=tree.root, id=tree.id()))
setOfTrees.append(Tree(root=tree.root, id=tree.id()))
return setOfTrees,setOfSubTrees
def activationQC(self, relevance, tree):
'''
:param relevance: np.array() dim D
:param tree: str
:return: ({'tree': str(tree), 'act_sub_trees':[(subtree, np.array())]},{'tree': str(tree), 'act_sub_trees':[(subtree, np.array())]})
'''
# calcolo i dtf dei sottoalberi di tree, cosi già ce li ho in cache
tree = Tree(string=tree)
self.calculateDTFs(tree)
# genero la matrice M, proprietà: M*M^T = I_nn
m = self.generateMatrixSubTree()
# print(m.shape, m.transpose().shape, relevance.shape)
# mmt = np.dot(m, m.transpose())
# a questo punto moltiplico il vettore della rilevanza per la matrice trasposta
# prendo prima tree1
relevance_tree = self.createRelevanceVector(relevance[:4000],
m.transpose())
# genero i 4 dizionari che mi servono per risalire ai sottoalberi
tree_index, tree_index_dict, embedding_matrix, tree_index_dict_inverse = self.relevance_matrix(
)
act_tree = self.saveActivation(tree, relevance_tree, tree_index_dict)
return act_tree
def activationRTE(self, relevance, tree1, tree2):
'''
It takes two trees in parenthetical form (str) as input.
1) convert trees to kerMIT.Tree
2) I calculate the dtf for each sub-tree of both trees
3) I take the generated sub trees from the cache and construct a matrix (N, D) N=number of subtrees, D=dimension dtf
4) multiply r * M^T so as to obtain a vector of dim (, N) => (, D) * (D, N) = (, N).
This vector tells us: how much each subtree (in the form of an index) is relevant for that input phrase
5) from the index step to the subtree itself
6) I return a list of pairs: (subtree, relevance)
:param relevance: np.array() dim D
:param tree1: str
:param tree2: str
:return: ({'tree': str(tree), 'act_sub_trees':[(subtree, np.array())]},{'tree': str(tree), 'act_sub_trees':[(subtree, np.array())]})
'''
# calcolo i dtf dei sottoalberi di tree1 e tree2, cosi già ce li ho in cache
tree1 = Tree(string=tree1)
tree2 = Tree(string=tree2)
self.calculateDTFs(tree1)
self.calculateDTFs(tree2)
# genero la matrice M, proprietà: M*M^T = I_nn
m = self.generateMatrixSubTree()
#print(m.shape, m.transpose().shape, relevance.shape)
#mmt = np.dot(m, m.transpose())
# a questo punto moltiplico il vettore della rilevanza per la matrice trasposta
# prendo prima tree1
relevance_tree_1 = self.createRelevanceVector(relevance[:4000],
m.transpose())
# prendo poi tree2
relevance_tree_2 = self.createRelevanceVector(relevance[4000:],
m.transpose())
# genero i 4 dizionari che mi servono per risalire ai sottoalberi
tree_index, tree_index_dict, embedding_matrix, tree_index_dict_inverse = self.relevance_matrix(
)
#print(relevance_tree_1)
#print(relevance_tree_2)
act_tree1 = self.saveActivation(tree1, relevance_tree_1,
tree_index_dict)
act_tree2 = self.saveActivation(tree2, relevance_tree_2,
tree_index_dict)
return act_tree1, act_tree2
def createDTK(parenthetical_tree):
tree = Tree(string=parenthetical_tree)
dtCalculator = DT(dimension=4000,
LAMBDA=0.4,
operation=fast_shuffled_convolution)
distributedTree1 = dtCalculator.dt(tree=tree)
return distributedTree1
def kernel(self, t1, t2):
try:
f = self.type_dic[self.kernelType]
except KeyError:
print("kernelType ", self.kernelType, " not implemented.")
return f(t1).dot(f(t2))
if __name__ == "__main__":
#s = "(S#trade$v:-1:4095 (NP#oil$n:-1:63 (NP#oil$n:-1:63 (JJ#Crude$j:1:7 Crude)(NN#oil$n:2:56 oil)))(VP#trade$v:-1:4032 (VBD#trade$v:-1:0 traded)(PP#barrel$n:-1:4032 (IN#at$i:-1:0 at)(NP#barrel$n:-1:4032 (NP#37.80$c:-1:448 ($#$$$:-1:0 $)(CD#37.80$c:3:448 37.80))(NP#barrel$n:-1:3584 (DT#a$d:-1:0 a)(NN#barrel$n:4:3584 barrel))))))"
ss = "(S (@S (NP (NP (@NP (DT The) (NN wait)) (NN time)) (PP (IN for) (NP (@NP (DT a) (JJ green)) (NN card)))) (VP (AUX has) (VP (@VP (VBN risen) (PP (IN from) (NP (@NP (NP (CD 21) (NNS months)) (TO to)) (NP (CD 33) (NNS months))))) (PP (IN in) (NP (@NP (DT those) (JJ same)) (NNS regions)))))) (. .))"
ss = '(NOTYPE##ROOT(NOTYPE##NP(NOTYPE##S(NOTYPE##NP(NOTYPE##NNP(NOTYPE##Children)))(NOTYPE##VP-REL(NOTYPE##VBG-REL(NOTYPE##W))(NOTYPE##CC(NOTYPE##and))(NOTYPE##VBG(NOTYPE##waving))(NOTYPE##PP(NOTYPE##IN(NOTYPE##W))(NOTYPE##NP(NOTYPE##NN(NOTYPE##camera))))))))'
ss = ss.replace(")", ") ").replace("(", " (")
t = Tree(string=ss)
kernel = DT(dimension=4000,
LAMBDA=0.6,
operation=op.fast_shuffled_convolution)
v = kernel.drule(t)
vv = kernel.dt(t)
for r in t.allRules():
print(r)
print(v)
print(vv)
print(np.dot(v, vv))
#print(kernel.kernel(t,frag))
def demo():
import random
def fill(cw):
cw['fill'] = '#%06d' % random.randint(0, 999999)
cf = CanvasFrame(width=550, height=450, closeenough=2)
t = Tree.fromstring('''
(S (NP the very big cat)
(VP (Adv sorta) (V saw) (NP (Det the) (N dog))))''')
tc = TreeWidget(cf.canvas(),
t,
draggable=1,
node_font=('helvetica', -14, 'bold'),
leaf_font=('helvetica', -12, 'italic'),
roof_fill='white',
roof_color='black',
leaf_color='green4',
node_color='blue2')
cf.add_widget(tc, 10, 10)
def boxit(canvas, text):
big = ('helvetica', -16, 'bold')
return BoxWidget(canvas,
TextWidget(canvas, text, font=big),
fill='green')
def ovalit(canvas, text):
return OvalWidget(canvas, TextWidget(canvas, text), fill='cyan')
treetok = Tree.fromstring(
'(S (NP this tree) (VP (V is) (AdjP shapeable)))')
tc2 = TreeWidget(cf.canvas(), treetok, boxit, ovalit, shapeable=1)
def color(node):
node['color'] = '#%04d00' % random.randint(0, 9999)
def color2(treeseg):
treeseg.label()['fill'] = '#%06d' % random.randint(0, 9999)
treeseg.label().child()['color'] = 'white'
tc.bind_click_trees(tc.toggle_collapsed)
tc2.bind_click_trees(tc2.toggle_collapsed)
tc.bind_click_nodes(color, 3)
tc2.expanded_tree(1).bind_click(color2, 3)
tc2.expanded_tree().bind_click(color2, 3)
paren = ParenWidget(cf.canvas(), tc2)
cf.add_widget(paren, tc.bbox()[2] + 10, 10)
tree3 = Tree.fromstring('''
(S (NP this tree) (AUX was)
(VP (V built) (PP (P with) (NP (N tree_to_treesegment)))))''')
tc3 = tree_to_treesegment(cf.canvas(),
tree3,
tree_color='green4',
tree_xspace=2,
tree_width=2)
tc3['draggable'] = 1
cf.add_widget(tc3, 10, tc.bbox()[3] + 10)
def orientswitch(treewidget):
if treewidget['orientation'] == 'horizontal':
treewidget.expanded_tree(1, 1).subtrees()[0].set_text('vertical')
treewidget.collapsed_tree(1, 1).subtrees()[0].set_text('vertical')
treewidget.collapsed_tree(1).subtrees()[1].set_text('vertical')
treewidget.collapsed_tree().subtrees()[3].set_text('vertical')
treewidget['orientation'] = 'vertical'
else:
treewidget.expanded_tree(1, 1).subtrees()[0].set_text('horizontal')
treewidget.collapsed_tree(1,
1).subtrees()[0].set_text('horizontal')
treewidget.collapsed_tree(1).subtrees()[1].set_text('horizontal')
treewidget.collapsed_tree().subtrees()[3].set_text('horizontal')
treewidget['orientation'] = 'horizontal'
text = """
Try clicking, right clicking, and dragging
different elements of each of the trees.
The top-left tree is a TreeWidget built from
a Tree. The top-right is a TreeWidget built
from a Tree, using non-default widget
constructors for the nodes & leaves (BoxWidget
and OvalWidget). The bottom-left tree is
built from tree_to_treesegment."""
twidget = TextWidget(cf.canvas(), text.strip())
textbox = BoxWidget(cf.canvas(), twidget, fill='white', draggable=1)
cf.add_widget(textbox, tc3.bbox()[2] + 10, tc2.bbox()[3] + 10)
tree4 = Tree.fromstring('(S (NP this tree) (VP (V is) (Adj horizontal)))')
tc4 = TreeWidget(cf.canvas(),
tree4,
draggable=1,
line_color='brown2',
roof_color='brown2',
node_font=('helvetica', -12, 'bold'),
node_color='brown4',
orientation='horizontal')
tc4.manage()
cf.add_widget(tc4, tc3.bbox()[2] + 10, textbox.bbox()[3] + 10)
tc4.bind_click(orientswitch)
tc4.bind_click_trees(tc4.toggle_collapsed, 3)
# Run mainloop
cf.mainloop()
return self.spectrum
if __name__ == "__main__":
#s = "(S#trade$v:-1:4095 (NP#oil$n:-1:63 (NP#oil$n:-1:63 (JJ#Crude$j:1:7 Crude)(NN#oil$n:2:56 oil)))(VP#trade$v:-1:4032 (VBD#trade$v:-1:0 traded)(PP#barrel$n:-1:4032 (IN#at$i:-1:0 at)(NP#barrel$n:-1:4032 (NP#37.80$c:-1:448 ($#$$$:-1:0 $)(CD#37.80$c:3:448 37.80))(NP#barrel$n:-1:3584 (DT#a$d:-1:0 a)(NN#barrel$n:4:3584 barrel))))))"
ss = "(S (@S (NP (NP (@NP (DT The) (NN wait)) (NN time)) (PP (IN for) (NP (@NP (DT a) (JJ green)) (NN card)))) (VP (AUX has) (VP (@VP (VBN risen) (PP (IN from) (NP (@NP (NP (CD 21) (NNS months)) (TO to)) (NP (CD 33) (NNS months))))) (PP (IN in) (NP (@NP (DT those) (JJ same)) (NNS regions)))))) (. .))"
ss = '(NOTYPE##ROOT(NOTYPE##NP(NOTYPE##S(NOTYPE##NP(NOTYPE##NNP(NOTYPE##Children)))(NOTYPE##VP-REL(NOTYPE##VBG-REL(NOTYPE##W))(NOTYPE##CC(NOTYPE##and))(NOTYPE##VBG(NOTYPE##waving))(NOTYPE##PP(NOTYPE##IN(NOTYPE##W))(NOTYPE##NP(NOTYPE##NN(NOTYPE##camera))))))))'
ss = ss.replace(")", ") ").replace("(", " (")
t = Tree(string=ss)
kernel = partialTreeKernel(dimension=8192, LAMBDA= 0.6, operation=op.fast_shuffled_convolution)
v = kernel.sRecursive(t)
w = kernel.dt(t)
print (v)
print (w)
#print(kernel.kernel(t,frag))
outStrings.append(
treeWithActivationsToString(tree, i,
(minmax[0][i], minmax[1][i])))
else:
outStrings.append(
treeWithActivationsToString(tree, i, minmax_global))
return outStrings
if __name__ == "__main__":
cf = CanvasFrame()
#t = Tree.fromstring('(S (NP:red this pippo) (VP (V is) (AdjP pretty)))')
t = Tree(string='(S (NP this pippo) (VP (V is) (AdjP pretty)))')
(_, subtrees) = DT.subtrees(t)
r.seed(10)
swa = [(st, [r.random(), r.random(), r.random()]) for st in subtrees]
treeout = activationsCalculator(t, swa)
stringout = treesWithActivationsToString(treeout)
print(stringout)
#activationsVisualizer(t,swa,cf)
#cf.mainloop()
#tree.demo()
