HRED and internal/external memory network unit-based emotional dialogue generation method
A network unit and emotion technology, applied in neural learning methods, biological neural network models, character and pattern recognition, etc., can solve problems such as unable to reply intelligently, dialogue system unable to use additional information, grammatical errors, etc., to improve automatic balance Effect
Inactive Publication Date: 2018-09-21
SUN YAT SEN UNIV
View PDF8 Cites 4 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
For the model based on retrieval technology, because the knowledge base is used and the data is pre-defined, the content of the reply is grammatically smooth and there are fewer grammatical errors; however, the model based on retrieval technology does not have the concept of conversation and cannot be combined with context Give Smarter Responses
The generative model is smarter, it can use context information more effectively to know what you are talking about; however, the generative model is more difficult to train, and the output often has some grammatical errors (especially for long sentences) language), and model training requires large-scale data
Retrieval-based dialogue systems cannot utilize additional information, so it is difficult to solve the problem of embedding emotional factors in dialogue systems
[0005] As far as we know, some people have used neural network-based generative models to generate answers in dialogue systems, but no one has yet embedded emotional factors into network models to solve the problem of emotional embedding in dialogue systems
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 Construction
[0019] The accompanying drawings are for illustrative purposes only and cannot be construed as limiting the patent;
[0020] In order to better illustrate this embodiment, some parts in the drawings will be omitted, enlarged or reduced, and do not represent the size of the actual product;
[0021] For those skilled in the art, it is understandable that some well-known structures and descriptions thereof may be omitted in the drawings.
[0022] The technical solutions of the present invention will be further described below in conjunction with the accompanying drawings and embodiments.
[0023] like figure 1 As shown, an emotional dialogue generation method based on HRED and internal and external memory network units, including the following steps:
[0024] S1: Prepare emotional dialogue system data;
[0025] S2: If the data set obtained in S1 is a single-round dialogue, use the single-round dialogue data set to train the encoding-decoding model, in which the pre-trained wor...
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 provides a HRED and internal / external memory network unit-based emotional dialogue generation method. According to the method, special emotional chat dialogue corpuses are constructed and a hierarchical recurrent neural network and internal / external memory network unit-based emotional dialogue generation model is put forward to solve embedding and application of emotional factors inlarge-scale multiple rounds of dialogues. Through an internal memory network, automatic balance between emotion expression and gramma expression is improved, and through an external memory network, explicit expression of emotions is increased, so that the answers generated by the model have more emotional information.
Description
technical field [0001] The invention relates to the field of automatic dialogue generation, and more specifically, relates to an emotional dialogue generation method based on HRED and internal and external memory network units. Background technique [0002] Encoder-decoder model: Also known as the encoding-decoding model, this is a model applied to the seq2seq problem. Simply put, seq2seq is to generate another output sequence y based on an input sequence x. seq2seq has many applications, such as translation, document extraction, question answering system and so on. In translation, the input sequence is the text to be translated, and the output sequence is the translated text; in question answering systems, the input sequence is the question asked, and the output sequence is the answer. In order to solve the seq2seq problem, someone proposed an encoder-decoder model, which is an encoding-decoding model. The so-called encoding is to convert the input sequence into a fixed-...
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 IPC IPC(8): G06F17/27G06F17/30G06K9/62G06N3/08
CPCG06N3/084G06F40/205G06F18/214
Inventor 卓汉逵纪登林
Owner SUN YAT SEN UNIV