Relation Extraction: Perspective from Convolutional Neural Networks

03-02

模型由4部分組成，就是經典的embedding，conv，pooling，fc

(i) the look-up tables to encode words in sentences by real-valued

vectors,

(ii) the convolutional layer to recognize n-grams,

(iii) the pooling layer to determine the most relevant features and

(iv) a logistic regression layer (a fully connected neural network with a softmax at

the end) to perform classification

CNN輸入：the word embeddings e_i and the position embeddings d_i_1 and d_i_2 are concatenated into a single vector x_i = [e_i, d_i_1, d_i_2]，再拼成一整句

CNN輸出：is a vector, the dimension of which is equal to the number of predefined relation types. The value of each dimension is the confidence score of the corresponding relation.

數據示例：