Air traffic control instruction translation method capable of perfecting semantic information
A technology of instruction translation and semantic information, applied in the field of machine translation, can solve problems such as unsatisfactory performance, achieve highly specific scenarios, improve performance and stability, and solve language bottlenecks
Inactive Publication Date: 2019-11-22
北京悠数智能科技有限公司
View PDF3 Cites 2 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
However, due to the professionalism and particularity of the air traffic control field, there are great differences between the instructions in the sentence form, syntax, word order and other aspects of the general domain text, which makes the general translation system perform unsatisfactorily in the air traffic control field
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment approach
[0038] The structural framework of the sequence-to-sequence model is as follows image 3 shown. The model is divided into an encoder end (encoding end) and a decoder end (decoding end). The encoder acts as a feature extractor, and the decoder acts as a semantic parser; the data flow at the encoder end passes through the Dropout layer, the bidirectional LSTM layer, and the Concatenate layer successively, and the data flow at the decoder end passes through LSTM layer and Dropout layer. The Dropout layer is mainly used to prevent overfitting and improve the generalization ability of the model. At the same time, the Encoder and Decoder are characterized by the same dimension of output data. Its implementation is as follows:
[0039] Encoder side:
[0040] encoder_input = ks.layers.Input(shape=(90,))
[0041] embed1=ks.layers.embeddings.Embedding(input_dim=length, output_dim=512, input_length=90, mask_zero=True)
[0042] encoder_inputs = embed1(encoder_input)
[0043] encoder...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention discloses an air traffic control instruction translation method capable of perfecting semantic information. The method comprises the following steps: acquiring air traffic control instruction characters represented by two types of codes; dividing an instruction into a plurality of sentences, and creating one-hot codes; building a deep network model; training a deep neural network through training data to obtain an air traffic control instruction semantic perfection and translation engine; and finally, by applying the obtained engine, translating and converting the air traffic control instruction into a representation of the other type of codes from a representation of one type of codes. The deep neural network uses a sequence-to-sequence structure with an attention mechanismmodule and a residual network; an encoder serves as a feature extractor, and a decoder serves as a semantic parser; connection processing is performed by the attention mechanism module; classified learning is carried out through a softmax classifier; and finally an optimal result is obtained through beam search in semantic perfection and translation conversion. The method takes the artificial intelligence deep learning engine as a core, has the advantages of being extremely high in professional applicability and accuracy, and lower in data volume dependence degree, and is excellent in semanticperfection and translation conversion of air traffic control instructions.
Description
technical field [0001] The invention relates to the technical field of machine translation, in particular to an air traffic control instruction translation method capable of perfecting semantic information. Background technique [0002] Due to the regional differences of the air traffic control business, the complexity of personnel, and the fast pronunciation speed, it is difficult for a single voice recognition system to truly achieve 100% recognition accuracy, and there are often situations where words are missing and many words are unclear. . The existing deep learning machine translation systems (such as Google, Baidu, etc.) are highly versatile, and the translation quality and speed can meet people's translation needs for general domain texts to a certain extent. However, due to the specialization and particularity of the air traffic control field, there are large differences in the sentence form, syntax, and word order between the instructions and the general domain...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Applications(China)
IPC IPC(8): G10L15/04G10L15/06G10L15/22G10L15/26G06N3/04G06N3/08
CPCG10L15/22G10L15/04G10L15/063G06N3/084G10L2015/0638G10L2015/0633G10L15/26G06N3/045
Inventor 孔维国
Owner 北京悠数智能科技有限公司