A Deep Representation Learning Method Based on Controllable Fusion of Features

Abstract

The invention discloses a deep representation learning method based on controllable fusion of features. On the basis of obtaining contextual embedded representations of words in a multi-layer language model based on pre-training, feature representations of different scales are respectively obtained from local and sequence perspectives. , and proposed to use a multi-head interactive linear attention mechanism to extract context summaries to realize the contextual information representation of words. The present invention uses a pre-trained multi-layer language model to embed and represent words, obtains a more contextualized representation of words, and solves the problem that word embedding representations in previous methods are not rich enough to solve polysemous words; the present invention proposes context Abstract, use multi-head interactive linear attention to calculate the specific representation of the current word under the influence of the entire sentence to find the difference between words to assist in the evaluation object mining; finally, the present invention uses the gate mechanism to filter features and assign weights to different features , which strengthens the impact of useful features.

Description

【Technical field】 [0001] The invention relates to a deep representation learning method for product evaluation object mining with controllable fusion of multi-scale and multi-type features. 【Background technique】 [0002] With the rapid development of the Internet, online shopping has become an indispensable part of people's lives, and the online review data of online products generated by online shopping has also shown an exponential growth. Most of these data are the real feelings and objective evaluations of consumers after using the products, which can not only guide or promote the purchase interest of other consumers, but also help product providers find problems and deficiencies in products, and promote product design and service improvement. Optimization, which contains a lot of commercial value. Specifically, from the perspective of consumers, what consumers want to pay attention to for a certain product may be certain attributes and characteristics of the product, ...

AI Technical Summary

Problems solved by technology

Among them, the first type of method mainly uses frequent itemset mining or manually constructed rule templates to mine evaluation objects. Although this type of method has achieved certain results, it is difficult to adapt to flexible and changeable syntactic features and semantic information, resulting in recall. The rate is generally not high

The second type of method is to use syntactic dependency parsing to capture the semantic relationship between words, and then use these relationships as the input of conditional random field or deep neural network, so as to mine the serialized semantic information features between words, but this type of The method is still limited by syntactic features and simple sequence features, lacks the breadth of features and deep semantic representation capabilities, and is also affected by syntax-dependent parsing results

Unfortunately, the above-mentioned methods do not take into account the influence and contribution of the use of local features, sequence features, and contextual features on the evaluation object mining results. How to controllably screen and optimize different types of features, And how to solve problems such as semantic differences in different fields, which largely affects the accuracy of product evaluation object mining

Images