Skip to content

Nodes

This class is used to get the nodes of the word graph.

Methods:

Name Description
__init__

str, dataset: List[str], level: int, k: int, c: int, word2vec_model_path: str, mlm_model_path: str, mlm_model_type: str = 'roberta') The constructor of the Nodes class.
get_similar_nodes

str) -> List[str] This method is used to get the similar nodes of a word.
get_context_nodes

str) -> List[str] This method is used to get the context nodes of a word.
get_node_features

Dict[str, List[str]]) This method is used to get the features of the nodes of the word graph.
Source code in semantics/graphs/temporal_graph.py 
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
class Nodes:
    """
    This class is used to get the nodes of the word graph.

    methods:
        __init__(self, target_word: str, dataset: List[str], level: int, k: int, c: int, word2vec_model_path: str, mlm_model_path: str, mlm_model_type: str = 'roberta')
            The constructor of the Nodes class.
        get_similar_nodes(self, word: str) -> List[str]
            This method is used to get the similar nodes of a word.
        get_context_nodes(self, word: str) -> List[str]
            This method is used to get the context nodes of a word.
        get_nodes(self) -> Dict[str, List[str]]
            This method is used to get the nodes of the word graph.
        get_node_features(self, nodes: Dict[str, List[str]])
            This method is used to get the features of the nodes of the word graph.
    """
    def __init__(
            self,
            target_word: str,
            dataset: List[str],
            level: int,
            k: int,
            c: int,
            word2vec_model: Word2VecInference,
            mlm_model: Union[RobertaInference, BertInference]
            ):

        """
        Args:
            target_word (str): the word to get the nodes for
            dataset (List[str]): the sentences to get the nodes from
            level (int): the level of the graph to get
            k (int): the number of similar nodes to get for each occurrence of the target word
            c (int): the number of context nodes to get for the target word
            word2vec_model (Word2VecInference): the word2vec model's Inference class
            mlm_model (RobertaInference, BertInference): the MLM model's Inference class
        """

        self.target_word = target_word
        self.dataset = dataset
        self.k = k
        self.c = c
        self.level = level
        self.word2vec = word2vec_model
        self.mlm = mlm_model


    def get_similar_nodes(self, word: str, keep_k: int = 50) -> List[str]:
        """
        This method is used to get the similar nodes of a word using the MLM model.

        Args:
            word (str): the word to get the similar nodes for
            keep_k (int): the number of similar nodes to keep for each occurrence of the word. Default: 50.

        Returns:
            similar_nodes (List[str]): the list of similar nodes of the word
        """
        print(f'Getting the similar nodes of the word: ')
        progress_bar = tqdm.tqdm(
            range(len(self.dataset)), 
            desc=f"{word}", 
            dynamic_ncols=True
            )
        similar_nodes = []
        for sentence in self.dataset:
            similar_nodes += self.mlm.get_top_k_words(word, sentence, self.k)
            progress_bar.update(1)

        similar_nodes = list(set(similar_nodes))
        similar_nodes = most_frequent(similar_nodes, keep_k)
        print(f'{len(similar_nodes)} similar nodes found.', '\n')
        return similar_nodes

    def get_context_nodes(self, word: str) -> List[str]:
        """
        This method is used to get the context nodes of a word using the word2vec model.

        Args:
            word (str): the word to get the context nodes for

        Returns:
            context_nodes (List[str]): the list of context nodes of the word
        """
        print(f'Getting the context nodes of the word "{word}" ...')
        context_nodes, _ = self.word2vec.get_top_k_words(word, self.c)

        print(f'{len(list(set(context_nodes)))} context nodes found.', '\n')
        return list(set(context_nodes))

    def get_nodes(self) -> Dict[str, List[str]]:
        """
        This method is used to get the nodes of the word graph (similar nodes, context nodes, and target node).

        Returns:
            nodes (Dict[str, List[str]]): the nodes of the word graph
        """
        nodes = {'target_node': [], 'similar_nodes': [], 'context_nodes': []}
        for level in range(self.level):
            print(f'Getting the nodes of level {level} ...')
            if level == 0:
                similar_nodes = self.get_similar_nodes(self.target_word, keep_k= 5)
                context_nodes = self.get_context_nodes(self.target_word)

                nodes['similar_nodes'].append(similar_nodes)
                nodes['context_nodes'].append(context_nodes)
                nodes['target_node'].append([self.target_word])

            else:
                similar_nodes = []
                context_nodes = []
                for word in nodes['similar_nodes'][level-1]:
                    similar_nodes += self.get_similar_nodes(word, keep_k= 5)
                    context_nodes += self.get_context_nodes(word)


                for word in nodes['context_nodes'][level-1]:
                    similar_nodes += self.get_similar_nodes(word, keep_k= 5)
                    context_nodes += self.get_context_nodes(word)

                nodes['similar_nodes'].append(similar_nodes)
                nodes['context_nodes'].append(context_nodes)          
        return nodes

    def get_node_features(self, nodes: Dict[str, List[str]]):
        """
        This method is used to get the features of the nodes of the word graph.

        Args:
            nodes (Dict[str, List[str]]): the nodes of the word graph

        Returns:
            index (Dict[str, Dict[int, str]]): the index of the nodes of the word graph. The index contains the 'index_to_key' and 'key_to_index' mapping dictionaries. Example: in the index_to_key dictionary {0: target_word}, and in the key_to_index dictionary {target_word: 0}.
            node_features (np.ndarray): the features of the nodes of the word graph of shape (num_nodes, 3) where num_nodes is the number of nodes in the graph. The features are:

                - node_type: the type of the node (target: 0, similar: 1, context: 2).

                - node_level: the level of the node in the graph. The target node is level 0.

                - frequency: the frequency of the word node in the dataset.
            embeddings (np.ndarray): the embeddings of the nodes of the word graph from the MLM model, of shape (num_nodes, 768).

        Examples:
            >>> word2vec = Word2VecInference('word2vec.model')
            >>> mlm = RobertaInference('MLM_roberta')
            >>> n = Nodes(target_word='sentence', dataset=['this is a sentence', 'this is another sentence'], level=3, k=2, c=2, word2vec_model = word2vec, mlm_model = mlm)
            >>> nodes = n.get_nodes()
            >>> index, node_features, embeddings = n.get_node_features(nodes)
            >>> print(index)
            {'index_to_key': {0: 'sentence', 1: 'this', 2: 'is', 3: 'a', 4: 'another'}, 'key_to_index': {'sentence': 0, 'this': 1, 'is': 2, 'a': 3, 'another': 4}
            >>> print(node_features)
            [[0, 0, 2], [1, 1, 2], [1, 1, 2], [1, 1, 2], [2, 1, 2]]
            >>> print(embeddings.shape)
            (5, 768)
        """
        index_to_key = {}
        key_to_index = {}
        node_types = []
        node_levels = []
        frequencies = []
        embeddings = []
        count = 0
        for node_type in ['target_node', 'similar_nodes', 'context_nodes']:
            for level in range(len(nodes[node_type])):
                for node in nodes[node_type][level]:
                    index_to_key[count] = node
                    key_to_index[node] = count
                    count += 1 
                    if node_type == 'target_node':
                        node_types.append(0)
                    elif node_type == 'similar_nodes':
                        node_types.append(1)
                    else:
                        node_types.append(2)
                    node_levels.append(level)
                    frequencies.append(count_occurence(self.dataset, node))
                    embeddings.append(self.mlm.get_embedding(main_word=node).mean(axis=0))

        embeddings = np.array(embeddings)
        node_features = np.stack([node_types, node_levels, frequencies]).T
        # node_features = np.concatenate((node_features, embeddings), axis=1)

        index = {'index_to_key': index_to_key, 'key_to_index': key_to_index}
        return index, node_features, embeddings

__init__(target_word, dataset, level, k, c, word2vec_model, mlm_model)

Parameters:

Name Type Description Default
target_word str

the word to get the nodes for

 required
dataset List[str]

the sentences to get the nodes from

 required
level int

the level of the graph to get

 required
k int

the number of similar nodes to get for each occurrence of the target word

 required
c int

the number of context nodes to get for the target word

 required
word2vec_model Word2VecInference

the word2vec model's Inference class

 required
mlm_model (RobertaInference, BertInference)

the MLM model's Inference class

 required
Source code in semantics/graphs/temporal_graph.py 
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
def __init__(
        self,
        target_word: str,
        dataset: List[str],
        level: int,
        k: int,
        c: int,
        word2vec_model: Word2VecInference,
        mlm_model: Union[RobertaInference, BertInference]
        ):

    """
    Args:
        target_word (str): the word to get the nodes for
        dataset (List[str]): the sentences to get the nodes from
        level (int): the level of the graph to get
        k (int): the number of similar nodes to get for each occurrence of the target word
        c (int): the number of context nodes to get for the target word
        word2vec_model (Word2VecInference): the word2vec model's Inference class
        mlm_model (RobertaInference, BertInference): the MLM model's Inference class
    """

    self.target_word = target_word
    self.dataset = dataset
    self.k = k
    self.c = c
    self.level = level
    self.word2vec = word2vec_model
    self.mlm = mlm_model

get_context_nodes(word)

This method is used to get the context nodes of a word using the word2vec model.

Parameters:

Name Type Description Default
word str

the word to get the context nodes for

 required

Returns:

Name Type Description
context_nodes List[str]

the list of context nodes of the word
Source code in semantics/graphs/temporal_graph.py 
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
def get_context_nodes(self, word: str) -> List[str]:
    """
    This method is used to get the context nodes of a word using the word2vec model.

    Args:
        word (str): the word to get the context nodes for

    Returns:
        context_nodes (List[str]): the list of context nodes of the word
    """
    print(f'Getting the context nodes of the word "{word}" ...')
    context_nodes, _ = self.word2vec.get_top_k_words(word, self.c)

    print(f'{len(list(set(context_nodes)))} context nodes found.', '\n')
    return list(set(context_nodes))

get_node_features(nodes)

This method is used to get the features of the nodes of the word graph.

Parameters:

Name Type Description Default
nodes Dict[str, List[str]]

the nodes of the word graph

 required

Returns:

Name Type Description
index Dict[str, Dict[int, str]]

the index of the nodes of the word graph. The index contains the 'index_to_key' and 'key_to_index' mapping dictionaries. Example: in the index_to_key dictionary {0: target_word}, and in the key_to_index dictionary {target_word: 0}.
node_features ndarray

the features of the nodes of the word graph of shape (num_nodes, 3) where num_nodes is the number of nodes in the graph. The features are:

  • node_type: the type of the node (target: 0, similar: 1, context: 2).

  • node_level: the level of the node in the graph. The target node is level 0.

  • frequency: the frequency of the word node in the dataset.
embeddings ndarray

the embeddings of the nodes of the word graph from the MLM model, of shape (num_nodes, 768).

Examples:

>>> word2vec = Word2VecInference('word2vec.model')
>>> mlm = RobertaInference('MLM_roberta')
>>> n = Nodes(target_word='sentence', dataset=['this is a sentence', 'this is another sentence'], level=3, k=2, c=2, word2vec_model = word2vec, mlm_model = mlm)
>>> nodes = n.get_nodes()
>>> index, node_features, embeddings = n.get_node_features(nodes)
>>> print(index)
{'index_to_key': {0: 'sentence', 1: 'this', 2: 'is', 3: 'a', 4: 'another'}, 'key_to_index': {'sentence': 0, 'this': 1, 'is': 2, 'a': 3, 'another': 4}
>>> print(node_features)
[[0, 0, 2], [1, 1, 2], [1, 1, 2], [1, 1, 2], [2, 1, 2]]
>>> print(embeddings.shape)
(5, 768)
Source code in semantics/graphs/temporal_graph.py 
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
def get_node_features(self, nodes: Dict[str, List[str]]):
    """
    This method is used to get the features of the nodes of the word graph.

    Args:
        nodes (Dict[str, List[str]]): the nodes of the word graph

    Returns:
        index (Dict[str, Dict[int, str]]): the index of the nodes of the word graph. The index contains the 'index_to_key' and 'key_to_index' mapping dictionaries. Example: in the index_to_key dictionary {0: target_word}, and in the key_to_index dictionary {target_word: 0}.
        node_features (np.ndarray): the features of the nodes of the word graph of shape (num_nodes, 3) where num_nodes is the number of nodes in the graph. The features are:

            - node_type: the type of the node (target: 0, similar: 1, context: 2).

            - node_level: the level of the node in the graph. The target node is level 0.

            - frequency: the frequency of the word node in the dataset.
        embeddings (np.ndarray): the embeddings of the nodes of the word graph from the MLM model, of shape (num_nodes, 768).

    Examples:
        >>> word2vec = Word2VecInference('word2vec.model')
        >>> mlm = RobertaInference('MLM_roberta')
        >>> n = Nodes(target_word='sentence', dataset=['this is a sentence', 'this is another sentence'], level=3, k=2, c=2, word2vec_model = word2vec, mlm_model = mlm)
        >>> nodes = n.get_nodes()
        >>> index, node_features, embeddings = n.get_node_features(nodes)
        >>> print(index)
        {'index_to_key': {0: 'sentence', 1: 'this', 2: 'is', 3: 'a', 4: 'another'}, 'key_to_index': {'sentence': 0, 'this': 1, 'is': 2, 'a': 3, 'another': 4}
        >>> print(node_features)
        [[0, 0, 2], [1, 1, 2], [1, 1, 2], [1, 1, 2], [2, 1, 2]]
        >>> print(embeddings.shape)
        (5, 768)
    """
    index_to_key = {}
    key_to_index = {}
    node_types = []
    node_levels = []
    frequencies = []
    embeddings = []
    count = 0
    for node_type in ['target_node', 'similar_nodes', 'context_nodes']:
        for level in range(len(nodes[node_type])):
            for node in nodes[node_type][level]:
                index_to_key[count] = node
                key_to_index[node] = count
                count += 1 
                if node_type == 'target_node':
                    node_types.append(0)
                elif node_type == 'similar_nodes':
                    node_types.append(1)
                else:
                    node_types.append(2)
                node_levels.append(level)
                frequencies.append(count_occurence(self.dataset, node))
                embeddings.append(self.mlm.get_embedding(main_word=node).mean(axis=0))

    embeddings = np.array(embeddings)
    node_features = np.stack([node_types, node_levels, frequencies]).T
    # node_features = np.concatenate((node_features, embeddings), axis=1)

    index = {'index_to_key': index_to_key, 'key_to_index': key_to_index}
    return index, node_features, embeddings

get_nodes()

This method is used to get the nodes of the word graph (similar nodes, context nodes, and target node).

Returns:

Name Type Description
nodes Dict[str, List[str]]

the nodes of the word graph
Source code in semantics/graphs/temporal_graph.py 
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
def get_nodes(self) -> Dict[str, List[str]]:
    """
    This method is used to get the nodes of the word graph (similar nodes, context nodes, and target node).

    Returns:
        nodes (Dict[str, List[str]]): the nodes of the word graph
    """
    nodes = {'target_node': [], 'similar_nodes': [], 'context_nodes': []}
    for level in range(self.level):
        print(f'Getting the nodes of level {level} ...')
        if level == 0:
            similar_nodes = self.get_similar_nodes(self.target_word, keep_k= 5)
            context_nodes = self.get_context_nodes(self.target_word)

            nodes['similar_nodes'].append(similar_nodes)
            nodes['context_nodes'].append(context_nodes)
            nodes['target_node'].append([self.target_word])

        else:
            similar_nodes = []
            context_nodes = []
            for word in nodes['similar_nodes'][level-1]:
                similar_nodes += self.get_similar_nodes(word, keep_k= 5)
                context_nodes += self.get_context_nodes(word)


            for word in nodes['context_nodes'][level-1]:
                similar_nodes += self.get_similar_nodes(word, keep_k= 5)
                context_nodes += self.get_context_nodes(word)

            nodes['similar_nodes'].append(similar_nodes)
            nodes['context_nodes'].append(context_nodes)          
    return nodes

get_similar_nodes(word, keep_k=50)

This method is used to get the similar nodes of a word using the MLM model.

Parameters:

Name Type Description Default
word str

the word to get the similar nodes for

 required
keep_k int

the number of similar nodes to keep for each occurrence of the word. Default: 50.

 50

Returns:

Name Type Description
similar_nodes List[str]

the list of similar nodes of the word
Source code in semantics/graphs/temporal_graph.py 
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
def get_similar_nodes(self, word: str, keep_k: int = 50) -> List[str]:
    """
    This method is used to get the similar nodes of a word using the MLM model.

    Args:
        word (str): the word to get the similar nodes for
        keep_k (int): the number of similar nodes to keep for each occurrence of the word. Default: 50.

    Returns:
        similar_nodes (List[str]): the list of similar nodes of the word
    """
    print(f'Getting the similar nodes of the word: ')
    progress_bar = tqdm.tqdm(
        range(len(self.dataset)), 
        desc=f"{word}", 
        dynamic_ncols=True
        )
    similar_nodes = []
    for sentence in self.dataset:
        similar_nodes += self.mlm.get_top_k_words(word, sentence, self.k)
        progress_bar.update(1)

    similar_nodes = list(set(similar_nodes))
    similar_nodes = most_frequent(similar_nodes, keep_k)
    print(f'{len(similar_nodes)} similar nodes found.', '\n')
    return similar_nodes