Font Size: a A A

Design And Application Of Feature-based Medical Synonyms Algorithm

Posted on:2021-06-03Degree:MasterType:Thesis
Country:ChinaCandidate:M GaoFull Text:PDF
GTID:2504306503473884Subject:Software engineering
Abstract/Summary:PDF Full Text Request
Since the development of artificial intelligence technology,various fields have been looking for favorable integration.In recent years,artificial intelligence technology has also emerged a trend of combining with the medical field.At the same time,the maturity and innovation of basic preparations such as a large amount of data resources,fast computing capabilities,and continuous optimization of algorithms have also become important cornerstones for the development of intelligent technology in the medical field.In this trend,the development of intelligent healthcare in China is facing unprecedented opportunities and challenges.The maturity of image and speech technologies has brought new opportunities to the medical industry.However,due to the complexity of language,the application of natural language processing in the medical field still faces many challenges.For example,the Chinese medical standard terminology set is incomplete,Chinese word segmentation is difficult,and Chinese medical clinical data is not standard.When integrating medical texts and other data resources,we often encounter the situation of multiple words have the same meaning,which makes it difficult for us to build a richer medical knowledge base.The difference between terminology and colloquialization of medical entities leads to a large amount of communication costs between patients and the medical system,which hinders the development and popularization of intelligent medical systems.It can be seen from the above questions that medical synonym extraction is a key technology for applying natural language processing to the medical field.The existing thesaurus is not complete,especially in the medical field.Relying on manpower to augment and enrich medical synonym data is a very labor and financially consuming task.To solve the above problems,this paper constructs a Chinese medical synonym dataset,compares the synonym classification algorithms based on cosine similarity and support vector machine,and then proposes a attentionbased medical synonym classification algorithm and applies it to online synonym editing and retrieval platform.This method not only makes full use of the context information of medical texts to learn the word vector representation of each medical entity,but also combines medical vocabulary features and search engine information mining feature vectors.Based on the comparison of two synonym classification algorithms,the attention-based synonym classification algorithm proposed in this paper reasonably learns appropriate weights for each feature and improves the effect of synonym classification.The online synonym editing platform satisfies the various needs of different users,and provides convenience for obtaining medical synonyms.The main research contents of this article are as follows:(1)Feature miningUse medical vocabulary features and search engine information to capture global context information,and add it to word vectors which focus on local context information.This approach gives more meaning to word vectors.Not only learned the semantic characteristics of language and context,but also improved the knowledge representation ability and limitations of information learning.(2)Compare the cosine similarity based medical synonyms classification algorithmThis paper makes full use of the word vectors learned from the context information,and the features mined from medical vocabulary and search engine information.The cosine similarity algorithm is used to obtain synonyms classification results,and the performance of different vector representation methods are compared.(3)Compare the machine learning based medical synonyms classification algorithmIn this paper,the obtained cosine similarity and the mined features as input of the machine learning classification algorithm(SVM),achieving better synonym classification results.(4)Design a attention mechanism based medical synonyms classification algorithmWhen using machine learning for classification,we found that SVM cannot fully capture the impact of different features on classification results,so we proposed an attention mechanism to assign more appropriate weights to different features,and finally achieved the best classification results.(5)Construct a medical synonym dataset and design an online synonym editing and retrieval platformThis article uses web crawler technology and massive Internet resources to build a Chinese medical synonyms dataset.This dataset and the synonym classification algorithm proposed in this paper are embedded in the synonym platform for users to use.
Keywords/Search Tags:medical synonyms, word embedding, cosine similarity, SVM, attention mechanism
PDF Full Text Request
Related items