WO2021179688A1 - Medical literature retrieval method and apparatus, electronic device, and storage medium - Google Patents

Abstract

A medical literature retrieval method and apparatus, an electronic device, and a storage medium, which are applied in the technical field of medical technology. Said method comprises: acquiring a first query statement; translating the first query statement to obtain a second query statement; according to the first query statement and the title and abstract of each piece of first medical literature, determining a first target similarity corresponding to the first query statement and each piece of the first medical literature; according to the second query statement and the title and abstract of each piece of second medical literature, determining a second target similarity corresponding to the second query statement and each piece of the second medical literature; and according to the first target similarity and the second target similarity, determining a piece of target medical literature corresponding to the first query statement. Said method facilitates increasing the retrieval efficiency of medical literature.

Description

Medical literature retrieval method, device, electronic equipment and storage medium

This application claims the priority of a Chinese patent application filed with the Chinese Patent Office on October 23, 2020, the application number is 202011152153.X, and the invention title is "Medical Document Retrieval Methods, Apparatus, Electronic Equipment, and Storage Media", and its entire contents Incorporated in this application by reference.

Technical field

This application relates to the technical field of information recommendation, in particular to a medical document retrieval method, device, electronic equipment and storage medium.

Background technique

The public medicine (PUBMED) database contains a large amount of medical literature, and the mass medical literature often contains the development trend of a certain medical field. By reading the medical literature in the medical field, researchers in related fields can be improved The efficiency and accuracy of decision-making by our and relevant public health policy makers. However, with the rapid increase in the number of publications in the medical field, the PUBMED medical database has also collected more and more medical literature.

The inventor realizes that at present, in order to improve the retrieval efficiency from the PUBMED medical database, it is generally necessary for experienced medical workers to mark each medical document in the PUBMED medical database with a related topic, and obtain the medical subject headings (Medical Subject Headings, MeSH), so that in the subsequent retrieval process, the query sentence can be matched with MeSH keywords to retrieve the relevant medical literature.

However, due to the large number of medical documents in the PUBMED database, manual annotation requires a lot of labor costs, and the retrieval results rely on manual annotation results, resulting in low retrieval efficiency and retrieval accuracy.

Summary of the invention

The embodiments of the present application provide a medical document retrieval method, device, electronic equipment, and storage medium, so as to improve the retrieval efficiency and retrieval accuracy of medical documents.

In the first aspect, an embodiment of the present application provides a medical document retrieval method, including:

Get the first query statement;

Translating the first query sentence to obtain a second query sentence, wherein the language types of the first query sentence and the second query sentence are different;

According to the first query sentence and the title and abstract of each first medical document, the first target similarity corresponding to the first query sentence and each first medical document is determined, wherein the first target similarity of each first medical document is determined. The language type of a medical document is the same as the first query sentence;

According to the second query sentence and the title and abstract of each second medical document, the second target similarity corresponding to the second query sentence and each second medical document is determined. 2. The language type of the medical document is the same as the second query sentence;

According to the first target similarity and the second target similarity, a target medical document corresponding to the first query sentence is determined.

In the second aspect, an embodiment of the present application provides a medical document retrieval device, including:

The transceiver unit is used to obtain the first query sentence;

A processing unit, configured to translate the first query sentence to obtain a second query sentence, wherein the language types of the first query sentence and the second query sentence are different;

The processing unit is further configured to determine the first target similarity corresponding to the first query sentence and each first medical document according to the first query sentence and the title and abstract of each first medical document , Wherein the language type of each first medical document is the same as the first query sentence;

The processing unit is further configured to determine the second target similarity corresponding to the second query sentence and each second medical document according to the second query sentence and the title and abstract of each second medical document , Wherein the language type of each second medical document is the same as the second query sentence;

The processing unit is further configured to determine the target medical document corresponding to the first query sentence according to the first target similarity and the second target similarity.

In a third aspect, an embodiment of the present application provides an electronic device, including a processor, the processor is connected to a memory, the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory , So that the electronic device executes the following method:

Get the first query statement;

Translating the first query sentence to obtain a second query sentence, wherein the language types of the first query sentence and the second query sentence are different;

According to the first query sentence and the title and abstract of each first medical document, the first target similarity corresponding to the first query sentence and each first medical document is determined, wherein the first target similarity of each first medical document is determined. The language type of a medical document is the same as the first query sentence;

According to the second query sentence and the title and abstract of each second medical document, the second target similarity corresponding to the second query sentence and each second medical document is determined. 2. The language type of the medical document is the same as the second query sentence;

According to the first target similarity and the second target similarity, a target medical document corresponding to the first query sentence is determined.

In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, the computer-readable storage medium stores a computer program, and the computer program causes a computer to execute the following method:

Get the first query statement;

Translating the first query sentence to obtain a second query sentence, wherein the language types of the first query sentence and the second query sentence are different;

According to the first query sentence and the title and abstract of each first medical document, the first target similarity corresponding to the first query sentence and each first medical document is determined, wherein the first target similarity of each first medical document is determined. The language type of a medical document is the same as the first query sentence;

According to the second query sentence and the title and abstract of each second medical document, the second target similarity corresponding to the second query sentence and each second medical document is determined. 2. The language type of the medical document is the same as the second query sentence;

According to the first target similarity and the second target similarity, a target medical document corresponding to the first query sentence is determined.

In a fifth aspect, embodiments of the present application provide a computer program product, the computer program product includes a non-transitory computer-readable storage medium storing a computer program, and the computer is operable to cause the computer to execute the computer program as described in the first aspect Methods.

The implementation of the embodiments of this application eliminates the need to mark the medical documents in the medical database in advance in the process of retrieving medical documents, which reduces labor costs and improves retrieval efficiency; in addition, in the process of retrieving medical documents, it can be retrieved at the same time Medical literature in different languages is published, and the accuracy of medical literature retrieval is improved.

Description of the drawings

In order to more clearly describe the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed in the description of the embodiments. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, without creative work, other drawings can be obtained from these drawings.

FIG. 1 is a schematic flowchart of a medical document retrieval method provided by an embodiment of the application;

FIG. 2 is a schematic structural diagram of a network model provided by an embodiment of this application;

3 is a block diagram of functional units of a medical document retrieval device provided by an embodiment of the application;

FIG. 4 is a schematic structural diagram of a medical document retrieval device provided by an embodiment of the application.

Detailed ways

The technical solutions in the embodiments of the present application will be described clearly and completely in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of them. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of this application.

The terms "first", "second", "third" and "fourth" in the specification and claims of this application and the drawings are used to distinguish different objects, not to describe a specific order . In addition, the terms "including" and "having" and any variations of them are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes unlisted steps or units, or optionally also includes Other steps or units inherent to these processes, methods, products or equipment.

Reference to "embodiments" herein means that specific features, results or characteristics described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art clearly and implicitly understand that the embodiments described herein can be combined with other embodiments.

The technical solution of this application can be applied to the fields of artificial intelligence, smart city, digital healthcare, blockchain and/or big data technology to realize document retrieval. Optionally, the data involved in this application, such as medical documents and/or vectors, can be stored in a database, or can be stored in a blockchain, which is not limited in this application.

Refer to FIG. 1, which is a schematic flowchart of a medical literature retrieval method provided by this application. The medical document sorting method is applied to a medical document retrieval device, and the method includes the following steps:

101: The medical document retrieval device acquires the first query sentence.

The first query sentence may be input by the user in the information input box of the medical document retrieval device, or may be obtained by performing voice recognition on the user's voice, for example, the user inputs the user's voice through a voice assistant. This application does not limit the method of obtaining the first query sentence.

102: The medical document retrieval device translates the first query sentence to obtain a second query sentence.

Wherein, the language types of the first query statement and the second query statement are different.

Exemplarily, the first query sentence and the second language type are Chinese or English. Therefore, when the first query sentence is in Chinese, the first query sentence can be translated into English to obtain the second query sentence; when the first query sentence is in English, the first query sentence can be translated into English. The query sentence is translated into Chinese, and the second query sentence is obtained.

It should be understood that in this application, two language types, Chinese and English, are mainly used as examples. In practical applications, the first query sentence can also be translated into other types of languages, such as Korean, Japanese, and so on.

Exemplarily, in the process of translating the first query sentence in this application, the first query sentence may be translated in combination with the medical knowledge graph. Specifically, the first medical knowledge graph corresponding to the first query sentence can be obtained from the medical knowledge graph database. For example, the first medical knowledge graph can be obtained by keyword matching, and then combined with the first medical knowledge graph. The medical knowledge graph translates the first query sentence to obtain a second query sentence corresponding to the first query sentence. Exemplarily, the first medical knowledge graph can be vectorized to obtain the first feature vector corresponding to the first medical knowledge graph; each word in the first query sentence is processed by word embedding to obtain each word The corresponding word vector; then, semantic feature extraction is performed on the word vector corresponding to each word to obtain the second feature vector corresponding to the first query sentence; finally, the second feature vector corresponding to the first query sentence and the first A first feature vector corresponding to a medical knowledge graph is spliced to obtain a first target feature vector; translation is performed according to the first target feature vector to obtain a second query sentence.

Wherein, the translation is performed based on the first target feature vector, and the existing coding network can be used to translate the second query sentence in an iterative manner, which will not be described in detail.

It can be seen that in the process of translating the first query sentence, the medical knowledge graph can be combined, that is, the prior knowledge is combined, thereby improving the accuracy of the translation of the first query sentence.

103: The medical document retrieval device determines the first target similarity corresponding to the first query sentence and each first medical document according to the first query sentence and the title and abstract of each first medical document.

Among them, the language type of each first medical document is the same as the first query sentence.

The first medical document may be a medical document with the same language type as the first query sentence in a medical database. For example, if the first query sentence is in English, then the English medical document in the medical database is regarded as the first medical document. Literature, where the medical database may be a PUBMED database.

Exemplarily, the first similarity and the second similarity between the first query sentence and the title and abstract of each first medical document can be determined respectively. For example, the BM25 algorithm can be used to determine the first similarity and the second similarity between the first query sentence and the title and abstract of each first medical document.

To determine the first similarity, for example, determine the first correlation between each word in the first query sentence and the title of each first medical document, and the second correlation between each word and the first query sentence Correlation, and determine the weight of each word; finally, according to the first correlation between each word and each first medical document, the second correlation between each word and the first query sentence, and each word The weight of, determines the first degree of similarity.

Further, the entity recognition of the first query sentence is performed to obtain the entity in the first query sentence; and the title and abstract of each first medical document are respectively recognized, and the title of each first medical document is obtained. Entity and the entity in the abstract; then, determine the third similarity between the entity in the first query sentence and the entity in the title of each first medical document, and determine the entity in the first query sentence and The fourth degree of similarity between entities in the abstract of each first medical document.

Exemplarily, the third degree of similarity and the fourth degree of similarity may be characterized by a Jaccard coefficient. Therefore, the third degree of similarity can be expressed by formula (1):

It should be understood that the method of determining the fourth degree of similarity is similar to that of determining the third degree of similarity, and will not be described again.

Among them, S3 is the third degree of similarity, A is the set of entities in the first query sentence, B is the set of entities in the title of each first medical document, 丨A∩B丨 can be set A and set The number of elements in the intersection of B, 丨A∪B丨 can be the number of elements in the union of set A and set B, 丨A丨, 丨B丨 can be the number of elements in set A and set B .

Exemplarily, determining the entity in the first query sentence, the title of each first medical document, and the entity in the abstract can be achieved by completing the trained bert model.

The entity recognition process is described below by taking the entity in the first query sentence as an example. The subsequent entity recognition is similar to this and will not be described again.

Perform word segmentation on the first query sentence through the trained bert model; perform word embedding processing on each word in the first query sentence through the bert model to obtain the word vector corresponding to each word; The word vector performs semantic feature extraction to obtain the target word vector of each word; according to the target word vector of each word, the type of each word is determined, and the entity in the first query sentence is obtained.

In addition, in the process of identifying entities in the first query sentence, the entity in the first query sentence can be determined in combination with the first medical knowledge graph, and it can be excluded that the first query sentence is not corresponding to the first medical knowledge graph. For the entities in the medical field, the entities that are more conducive to medical literature retrieval. Specifically, firstly, each entity in the first medical knowledge graph is vectorized to obtain the feature vector corresponding to each entity; then, based on the attention mechanism, the weight coefficient between each word and each entity is determined, that is, each entity is determined separately. The similarity between the word vector corresponding to each word and the feature vector of each entity, and then the similarity corresponding to each entity is normalized to obtain the weight coefficient between each word and each entity; according to the corresponding entity The weight coefficient weights the feature vector of each entity to obtain the target word vector corresponding to each word; finally, according to the target word vector corresponding to each word, the type of each word is determined, and the entity in the first query sentence is obtained .

It can be seen that in combination with the medical knowledge graph, some entities in the first query sentence that are not in the medical field corresponding to the first medical knowledge graph can be excluded, such as some, name entities, name entities, and so on. Since these entities that are of little help to medical literature retrieval are excluded, the accuracy of the third similarity degree and the fourth similarity degree can be improved, and the retrieval efficiency and accuracy of the medical literature can be improved.

In some possible implementations, before the entity in the first query sentence is determined, the bert model can be used to identify the subject and abstract of each first medical document in the medical database to obtain each first medical document. The subject and the entities in the abstract. In other words, the subject of medical literature and the entity in the abstract can be identified in advance, so that in the process of medical literature retrieval, there is no need to identify the title and entity in the abstract of each first medical literature, which improves The retrieval efficiency of medical literature.

It should be understood that the entities involved in the embodiments of the present application (for example, the entities in the abstracts and titles of medical documents, or the entities in the first query sentence) may include diseases, drugs, surgery, genes, laboratory tests, etc. Wait.

Finally, weighting is performed on the first similarity, the second similarity, the third similarity, and the fourth similarity to obtain the first target similarity. Exemplarily, the first target similarity can be expressed by formula (2):

S _m1 =α ₁ *S ₁ +α ₂ *S ₂ +α ₃ *S ₃ +α ₄ *S ₄ (Equation 2);

Among them, S _m1 is the first target similarity, S ₁ , S ₂ , S ₃ and S ₄ respectively correspond to the first similarity, the second similarity, the third similarity, and the first similarity corresponding to the first query sentence. Four similarities, α ₁ , α ₂ , α ₃ and α ₄ are preset weight coefficients, and α ₁ +α ₂ +α ₃ +α ₄ =1.

104: The medical document retrieval device determines the second target similarity corresponding to the second query sentence and each second medical document according to the second query sentence and the title and abstract of each second medical document.

Wherein, the language type of each second medical document is the same as the second query sentence.

Similarly, a medical document with the same language type as the second query sentence in a public medical database can be used as the second medical document.

Exemplarily, similar to the method for determining the first target similarity corresponding to the first query sentence, the first similarity and the second similarity between the second query sentence and the title and abstract of the second medical document are respectively determined; Then, the third similarity and the fourth similarity between the entities in the second query sentence and the entities in the title and abstract of the second medical document are respectively determined; finally, the second query sentence corresponds to the third similarity and the fourth similarity. The first similarity, the second similarity, the third similarity, and the fourth similarity are weighted to obtain the second target similarity corresponding to the second query vector.

105: The medical document retrieval device according to the first target similarity between the first query sentence and each of the first medical documents and the second target corresponding to the second query sentence and each of the second medical documents The similarity determines the target medical document corresponding to the first query sentence.

Exemplarily, according to the first target similarity corresponding to the first query sentence and each first medical document, the first medical document corresponding to the largest first target similarity is taken as a target medical document; according to the first query The sentence corresponds to the second target similarity of each second medical document, and the second medical document corresponding to the largest second target similarity is regarded as the other target medical document; then, the two target medical documents are regarded as the second target medical document. A target medical literature corresponding to the query sentence.

Exemplarily, the medical document with the greatest similarity among the two target medical documents may also be used as the target medical document corresponding to the first query sentence.

It can be seen that in the implementation of this application, medical documents of different language types can be retrieved at the same time, and there is no need to mark the medical documents in the medical database in advance during the retrieval process, which reduces the labor cost; moreover, through The entity matching method determines the similarity between the entity in the query sentence and the abstract of each medical document and the entity in the title, and further improves the accuracy of medical document retrieval.

In one embodiment of this application, the medical literature retrieval method of this application can also be applied to smart medical scenarios. For example, a doctor can quickly retrieve the medical literature that the doctor wants to obtain through the medical literature retrieval method of this application, and then Quickly find historical cases or documents, so as to improve the data reference for the doctor's diagnosis, improve the doctor's diagnosis efficiency and accuracy, and push the development of medical technology.

In some possible implementation manners, the foregoing determination of the first target similarity and the second target similarity may be achieved by completing the training of the network model. The following describes the implementation process of the medical document retrieval method of the present application by combining the structure of the network model and determining the similarity of the first target as an example. Wherein, the method of determining the similarity of the second target is similar to the method of determining the similarity of the first target, and will not be described again.

Referring to Fig. 2, the network model includes an embedding layer 1, an embedding layer 2, a feature extraction network 1, a feature extraction network 2, a feature extraction network 3, a feature extraction network 4, a decoding network, and an atlas conversion network. Among them, the embedding layer 1, the embedding layer 2, the embedding layer 3, and the graph conversion network can be a network based on the bert model for word embedding; the feature extraction network 1, the feature extraction network 2, the feature extraction network 3, and the feature extraction network 4 can be It is a general feature extraction network used for semantic feature extraction, for example, it can be a long short-term memory network LSTM, a recurrent neural network RNN, etc.; a decoder network (Decoder) can include multiple stack layers.

Exemplarily, the embedding layer 1 is used to embed each word in the first query sentence to obtain the word vector corresponding to each word, and the feature extraction network 1 is used to perform semantic feature extraction on the word vector of each word. The second feature vector corresponding to the first query sentence is obtained; then, the first medical knowledge graph matching the first query sentence is obtained from the medical knowledge graph database; the embedding layer 2 is used to separately store the first medical knowledge graph Word embedding of medical knowledge (such as disease names, precautions, entities, etc.) to obtain the feature vector corresponding to the medical knowledge; feature extraction network 2 is used to perform feature extraction on the feature vector corresponding to the medical knowledge to obtain the first The first feature vector corresponding to the medical knowledge graph;

Then, the first feature vector and the second feature vector are spliced to obtain the first target feature vector corresponding to the first query sentence; the decoding network is used for translating according to the first target feature vector to obtain the The second query statement corresponding to the query statement;

Then, according to the self-attention mechanism, determine the similarity between the word vector of each word and the feature vector of each entity in the first medical knowledge graph, and obtain the weight coefficient between each word and each entity; use each word The weight coefficient between each entity and the feature vector corresponding to each entity are weighted to obtain the target word vector corresponding to each word;

Further, the feature extraction network 3 is used to perform feature extraction on the target word vector of each word to obtain the target feature vector corresponding to the first query sentence, and to obtain the target feature vector corresponding to the first query sentence according to the target feature vector corresponding to the first query sentence. The words in the query sentence are classified, and the entity 3 in the first query sentence is obtained;

Exemplarily, based on the BM25 algorithm, the first similarity and the second similarity between the first query sentence and the title and abstract of the first medical document are respectively calculated;

Exemplarily, the embedding layer 2 is used to embed the title and abstract of the first medical document respectively to obtain the word vector corresponding to the title and the abstract; then, the feature extraction network 4 is used to feature the word vector corresponding to the title and the abstract Extract and classify to obtain entity 1 in the title and entity 2 in the abstract; then, determine the third similarity between entity 1 and entity 3, and the fourth similarity between entity 2 and entity 3;

Finally, the first similarity, the second similarity, the third similarity, and the fourth similarity are weighted to obtain the first target similarity corresponding to each first medical document.

Referring to FIG. 3, FIG. 3 is a block diagram of the functional unit composition of a medical document retrieval device provided by an embodiment of the application. The medical document retrieval device 300 includes: a transceiver unit 301 and a processing unit 302, wherein:

The transceiver unit 301 is configured to obtain the first query sentence;

The processing unit 302 is configured to translate the first query sentence to obtain a second query sentence, wherein the language types of the first query sentence and the second query sentence are different;

The processing unit 302 is further configured to determine the first target similarity corresponding to the first query sentence and each first medical document according to the first query sentence and the title and abstract of each first medical document, Wherein, the language type of each first medical document is the same as the first query sentence;

The processing unit 302 is further configured to determine the second target similarity corresponding to the second query sentence and each second medical document according to the second query sentence and the title and abstract of each second medical document, Wherein, the language type of each second medical document is the same as the second query sentence;

The processing unit 302 is further configured to determine the target medical document corresponding to the first query sentence according to the first target similarity and the second target similarity.

In some possible implementation manners, in terms of translating the first query sentence to obtain the second query sentence, the processing unit 302 is specifically configured to:

Acquiring a first medical knowledge graph corresponding to the first query sentence in the medical knowledge graph database;

According to the first medical knowledge graph, the first query sentence is translated to obtain the second query sentence.

In some possible implementation manners, in terms of translating the first query sentence according to the first medical knowledge graph to obtain the second query sentence, the processing unit 302 is specifically configured to:

Vectorizing the first medical knowledge graph to obtain a first feature vector corresponding to the first medical knowledge graph;

Performing word embedding processing on each word in the first query sentence to obtain a word vector corresponding to each word;

Performing semantic feature extraction on the word vector corresponding to each word to obtain a second feature vector corresponding to the first query sentence;

Splicing a second feature vector corresponding to the first query sentence with the first feature vector to obtain a first target feature vector;

Translate according to the first target feature vector to obtain the second query sentence.

In some possible implementation manners, according to the first query sentence and the title and abstract of each first medical document, it is determined that the first query sentence is similar to the first target corresponding to each first medical document. In terms of degree, the processing unit 302 is specifically used for:

Determine the first similarity and the second similarity between the first query sentence and the title and abstract of each first medical document respectively

Performing entity recognition on the first query sentence to obtain the entity in the first query sentence;

Recognizing the title and abstract of each first medical document, respectively, to obtain the entity in the title of each first medical document and the entity in the abstract;

Determine the third degree of similarity between the entity in the first query sentence and the entity in the title of each first medical document, and between the entity in the abstract of each first medical document The fourth degree of similarity;

According to the first degree of similarity, the second degree of similarity, the third degree of similarity, and the fourth degree of similarity, determine the first target corresponding to the first query sentence and each of the first medical documents Similarity.

In some possible implementations, the third degree of similarity between the entity in the first query sentence and the entity in the title of each first medical document is determined, and the degree of similarity between the entity in the first query sentence and the entity in the title of each first medical document is determined. Regarding the fourth degree of similarity between entities in the abstract of medical documents, the processing unit 302 is specifically configured to:

Determine the first Jaccard coefficient between the entity in the first query sentence and the entity in the title of each first medical document, and use the first Jaccard coefficient as the third Similarity

Determine the second Jaccard coefficient between the entity in the first query sentence and the entity in the abstract of each first medical document, and use the second Jaccard coefficient as the fourth similarity Spend.

In some possible implementation manners, in terms of performing entity recognition on the first query sentence to obtain the entity in the first query sentence, the processing unit 302 is specifically configured to:

Performing word embedding on each word in the first query sentence to obtain a word vector corresponding to each word;

Vectorize each entity in the first medical knowledge graph to obtain a feature vector of each entity in the first medical knowledge graph;

Determining the weight coefficient between the word vector corresponding to each word and the feature vector of each entity in the first medical knowledge graph;

Weighting the feature vector of each entity in the first medical knowledge graph according to the weight coefficient to obtain the target word vector corresponding to each word;

Classify each word according to the target word vector corresponding to each word to obtain the entity in the first query sentence.

In some possible implementation manners, in terms of determining the target medical document corresponding to the first query sentence according to the first target similarity and the second target similarity, the processing unit 302 is specifically configured to:

Take the first medical document corresponding to the largest first target similarity as a target medical document;

Take the second medical document corresponding to the largest second target similarity as another target medical document;

Two target medical documents are taken as target medical documents corresponding to the first query sentence.

Refer to FIG. 4, which is a schematic structural diagram of an electronic device according to an embodiment of the application. The electronic device includes a processor, the processor is connected to a memory, the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the electronic device executes the above method. For example, as shown in FIG. 4, the electronic device 400 includes a transceiver 401, a processor 402, and a memory 403. They are connected by a bus 404 between them. The memory 403 is used to store computer programs and data, and can transmit the data stored in the memory 403 to the processor 402.

The processor 402 is configured to read the computer program in the memory 403 and perform the following operations:

Get the first query statement;

Translating the first query sentence to obtain a second query sentence, wherein the language types of the first query sentence and the second query sentence are different;

According to the first query sentence and the title and abstract of each first medical document, the first target similarity corresponding to the first query sentence and each first medical document is determined, wherein the first target similarity of each first medical document is determined. The language type of a medical document is the same as the first query sentence;

According to the second query sentence and the title and abstract of each second medical document, the second target similarity corresponding to the second query sentence and each second medical document is determined. 2. The language type of the medical document is the same as the second query sentence;

According to the first target similarity and the second target similarity, a target medical document corresponding to the first query sentence is determined.

In some possible implementation manners, in terms of translating the first query sentence to obtain the second query sentence, the processor 402 is specifically configured to perform the following operations:

Acquiring a first medical knowledge graph corresponding to the first query sentence in the medical knowledge graph database;

According to the first medical knowledge graph, the first query sentence is translated to obtain the second query sentence.

In some possible implementation manners, in terms of translating the first query sentence according to the first medical knowledge graph to obtain the second query sentence, the processor 402 is specifically configured to perform the following operations:

Vectorizing the first medical knowledge graph to obtain a first feature vector corresponding to the first medical knowledge graph;

Performing word embedding processing on each word in the first query sentence to obtain a word vector corresponding to each word;

Performing semantic feature extraction on the word vector corresponding to each word to obtain a second feature vector corresponding to the first query sentence;

Splicing a second feature vector corresponding to the first query sentence with the first feature vector to obtain a first target feature vector;

Translate according to the first target feature vector to obtain the second query sentence.

In some possible implementation manners, according to the first query sentence and the title and abstract of each first medical document, it is determined that the first query sentence is similar to the first target corresponding to each first medical document. In terms of degrees, the processor 402 is specifically configured to perform the following operations:

Determine the first similarity and the second similarity between the first query sentence and the title and abstract of each first medical document respectively

Performing entity recognition on the first query sentence to obtain the entity in the first query sentence;

Recognizing the title and abstract of each first medical document, respectively, to obtain the entity in the title of each first medical document and the entity in the abstract;

Determine the third degree of similarity between the entity in the first query sentence and the entity in the title of each first medical document, and between the entity in the abstract of each first medical document The fourth degree of similarity;

According to the first degree of similarity, the second degree of similarity, the third degree of similarity, and the fourth degree of similarity, determine the first target corresponding to the first query sentence and each of the first medical documents Similarity.

In some possible implementation manners, the third degree of similarity between the entity in the first query sentence and the entity in the title of each first medical document is determined, and the third similarity between the entity in the first query sentence and the entity in the title of each first medical document is Regarding the fourth degree of similarity between entities in the abstract of the medical document, the processor 402 is specifically configured to perform the following operations:

Determine the first Jaccard coefficient between the entity in the first query sentence and the entity in the title of each first medical document, and use the first Jaccard coefficient as the third Similarity

Determine the second Jaccard coefficient between the entity in the first query sentence and the entity in the abstract of each first medical document, and use the second Jaccard coefficient as the fourth similarity Spend.

In some possible implementation manners, in terms of performing entity recognition on the first query sentence to obtain the entity in the first query sentence, the processor 402 is specifically configured to perform the following operations:

Performing word embedding on each word in the first query sentence to obtain a word vector corresponding to each word;

Vectorize each entity in the first medical knowledge graph to obtain a feature vector of each entity in the first medical knowledge graph;

Determining a weight coefficient between the word vector corresponding to each word and the feature vector of each entity in the first medical knowledge graph;

Weighting the feature vector of each entity in the first medical knowledge graph according to the weight coefficient to obtain the target word vector corresponding to each word;

Classify each word according to the target word vector corresponding to each word to obtain the entity in the first query sentence.

In some possible implementation manners, in terms of determining the target medical document corresponding to the first query sentence according to the first target similarity and the second target similarity, the processor 402 is specifically configured to perform the following operations :

Take the first medical document corresponding to the largest first target similarity as a target medical document;

Take the second medical document corresponding to the largest second target similarity as another target medical document;

Two target medical documents are taken as target medical documents corresponding to the first query sentence.

Specifically, the transceiver 401 may be the transceiver unit 301 of the medical document retrieval device 300 in the embodiment shown in FIG. 3, and the processor 402 may be the processing unit 302 of the medical document retrieval device 300 in the embodiment shown in FIG. .

It should be understood that the medical document retrieval device in this application may include smart phones (such as Android phones, iOS phones, Windows Phone phones, etc.), tablet computers, handheld computers, notebook computers, and mobile Internet Devices (MID). ) Or wearable devices, etc. The aforementioned medical document retrieval device is only an example, not an exhaustive list, and includes but not limited to the aforementioned medical document retrieval device. In practical applications, the aforementioned medical document retrieval device may also include: intelligent vehicle-mounted terminals, computer equipment, and so on.

The embodiment of the present application also provides a computer-readable storage medium, the computer-readable storage medium stores a computer program, and the computer program is executed by a processor to realize any medical document retrieval as recorded in the above method embodiment Part or all of the steps of the method.

Optionally, the storage medium involved in this application, such as a computer-readable storage medium, may be non-volatile or volatile.

The embodiments of the present application also provide a computer program product, the computer program product includes a non-transitory computer-readable storage medium storing a computer program, the computer program is operable to cause a computer to execute the method described in the above method embodiment Part or all of the steps of any kind of medical literature retrieval method.

It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations, but those skilled in the art should know that this application is not limited by the described sequence of actions. Because according to this application, some steps can be performed in other order or at the same time. Secondly, those skilled in the art should also know that the embodiments described in the specification are all optional embodiments, and the involved actions and modules are not necessarily required by this application.

In the above-mentioned embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in an embodiment, reference may be made to related descriptions of other embodiments.

In the several embodiments provided in this application, it should be understood that the disclosed device may be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other divisions in actual implementation, for example, multiple units or components may be combined or may be Integrate into another system, or some features can be ignored or not implemented. In addition, the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may be in electrical or other forms.

The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.

In addition, the functional units in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software program modules.

If the integrated unit is implemented in the form of a software program module and sold or used as an independent product, it can be stored in a computer readable memory. Based on this understanding, the technical solution of the present application essentially or the part that contributes to the existing technology or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a memory. A number of instructions are included to enable a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned memory includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above-mentioned embodiments can be completed by a program instructing relevant hardware. The program can be stored in a computer-readable memory, and the memory can include: a flash disk , Read-only memory (English: Read-Only Memory, abbreviation: ROM), random access device (English: Random Access Memory, abbreviation: RAM), magnetic disk or optical disc, etc.

The embodiments of the application are described in detail above, and specific examples are used in this article to illustrate the principles and implementation of the application. The descriptions of the above embodiments are only used to help understand the methods and core ideas of the application; at the same time, for Those of ordinary skill in the art, based on the idea of the application, will have changes in the specific implementation and the scope of application. In summary, the content of this specification should not be construed as a limitation to the application.

Claims (20)

A medical literature retrieval method, including: Get the first query statement; Translating the first query sentence to obtain a second query sentence, wherein the language types of the first query sentence and the second query sentence are different; According to the first query sentence and the title and abstract of each first medical document, the first target similarity corresponding to the first query sentence and each first medical document is determined, wherein the first target similarity of each first medical document is determined. The language type of a medical document is the same as the first query sentence; According to the second query sentence and the title and abstract of each second medical document, the second target similarity corresponding to the second query sentence and each second medical document is determined. 2. The language type of the medical document is the same as the second query sentence; According to the first target similarity and the second target similarity, a target medical document corresponding to the first query sentence is determined. The method according to claim 1, wherein said translating said first query sentence to obtain a second query sentence comprises: Acquiring a first medical knowledge graph corresponding to the first query sentence in the medical knowledge graph database; According to the first medical knowledge graph, the first query sentence is translated to obtain the second query sentence. The method according to claim 2, wherein the translating the first query sentence according to the first medical knowledge graph to obtain the second query sentence comprises: Vectorizing the first medical knowledge graph to obtain a first feature vector corresponding to the first medical knowledge graph; Performing word embedding processing on each word in the first query sentence to obtain a word vector corresponding to each word; Performing semantic feature extraction on the word vector corresponding to each word to obtain a second feature vector corresponding to the first query sentence; Splicing a second feature vector corresponding to the first query sentence with the first feature vector to obtain a first target feature vector; Translate according to the first target feature vector to obtain the second query sentence. The method according to any one of claims 1 to 3, wherein the first query sentence and the each article are determined according to the first query sentence and the title and abstract of each first medical document. The first target similarity corresponding to the first medical document includes: Determine the first similarity and the second similarity between the first query sentence and the title and abstract of each first medical document respectively Performing entity recognition on the first query sentence to obtain the entity in the first query sentence; Recognizing the title and abstract of each first medical document, respectively, to obtain the entity in the title of each first medical document and the entity in the abstract; Determine the third degree of similarity between the entity in the first query sentence and the entity in the title of each first medical document, and between the entity in the abstract of each first medical document The fourth degree of similarity; According to the first degree of similarity, the second degree of similarity, the third degree of similarity, and the fourth degree of similarity, determine the first target corresponding to the first query sentence and each of the first medical documents Similarity. The method according to claim 4, wherein said determining a third degree of similarity between the entity in the first query sentence and the entity in the title of each first medical document, and the The fourth degree of similarity between entities in the abstract of each first medical document, including: Determine the first Jaccard coefficient between the entity in the first query sentence and the entity in the title of each first medical document, and use the first Jaccard coefficient as the third Similarity Determine the second Jaccard coefficient between the entity in the first query sentence and the entity in the abstract of each first medical document, and use the second Jaccard coefficient as the fourth similarity Spend. The method according to claim 4, wherein said performing entity recognition on said first query sentence to obtain an entity in said first query sentence comprises: Performing word embedding on each word in the first query sentence to obtain a word vector corresponding to each word; Vectorize each entity in the first medical knowledge graph to obtain a feature vector of each entity in the first medical knowledge graph; Determining a weight coefficient between the word vector corresponding to each word and the feature vector of each entity in the first medical knowledge graph; Weighting the feature vector of each entity in the first medical knowledge graph according to the weight coefficient to obtain the target word vector corresponding to each word; Classify each word according to the target word vector corresponding to each word to obtain the entity in the first query sentence. The method according to any one of claims 1 to 3, wherein the determining the target medical document corresponding to the first query sentence according to the first target similarity and the second target similarity, include: Take the first medical document corresponding to the largest first target similarity as a target medical document; Take the second medical document corresponding to the largest second target similarity as another target medical document; Two target medical documents are taken as target medical documents corresponding to the first query sentence. A medical document retrieval device, including: The transceiver unit is used to obtain the first query sentence; A processing unit, configured to translate the first query sentence to obtain a second query sentence, wherein the language types of the first query sentence and the second query sentence are different; The processing unit is further configured to determine the first target similarity corresponding to the first query sentence and each first medical document according to the first query sentence and the title and abstract of each first medical document , Wherein the language type of each first medical document is the same as the first query sentence; The processing unit is further configured to determine the second target similarity corresponding to the second query sentence and each second medical document according to the second query sentence and the title and abstract of each second medical document , Wherein the language type of each second medical document is the same as the second query sentence; The processing unit is further configured to determine the target medical document corresponding to the first query sentence according to the first target similarity and the second target similarity. An electronic device, comprising: a processor connected to a memory, the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the electronic device executes the following method: Get the first query statement; Translating the first query sentence to obtain a second query sentence, wherein the language types of the first query sentence and the second query sentence are different; According to the first query sentence and the title and abstract of each first medical document, the first target similarity corresponding to the first query sentence and each first medical document is determined, wherein the first target similarity of each first medical document is determined. The language type of a medical document is the same as the first query sentence; According to the second query sentence and the title and abstract of each second medical document, the second target similarity corresponding to the second query sentence and each second medical document is determined. 2. The language type of the medical document is the same as the second query sentence; According to the first target similarity and the second target similarity, a target medical document corresponding to the first query sentence is determined. 9. The electronic device according to claim 9, wherein when the first query sentence is translated to obtain the second query sentence, the following is specifically executed: Acquiring a first medical knowledge graph corresponding to the first query sentence in the medical knowledge graph database; According to the first medical knowledge graph, the first query sentence is translated to obtain the second query sentence. The electronic device according to claim 10, wherein the first query sentence is translated according to the first medical knowledge graph, and when the second query sentence is obtained, the following is specifically executed: Vectorizing the first medical knowledge graph to obtain a first feature vector corresponding to the first medical knowledge graph; Performing word embedding processing on each word in the first query sentence to obtain a word vector corresponding to each word; Performing semantic feature extraction on the word vector corresponding to each word to obtain a second feature vector corresponding to the first query sentence; Splicing a second feature vector corresponding to the first query sentence with the first feature vector to obtain a first target feature vector; Translate according to the first target feature vector to obtain the second query sentence. The electronic device according to any one of claims 9-11, wherein the first query sentence is determined to be related to each of the first query sentence and the title and abstract of each first medical document. When the first target similarity corresponding to the first medical document, the specific implementation is as follows: Determine the first similarity and the second similarity between the first query sentence and the title and abstract of each first medical document respectively Performing entity recognition on the first query sentence to obtain the entity in the first query sentence; Recognizing the title and abstract of each first medical document, respectively, to obtain the entity in the title of each first medical document and the entity in the abstract; Determine the third degree of similarity between the entity in the first query sentence and the entity in the title of each first medical document, and between the entity in the abstract of each first medical document The fourth degree of similarity; According to the first degree of similarity, the second degree of similarity, the third degree of similarity, and the fourth degree of similarity, determine the first target corresponding to the first query sentence and each of the first medical documents Similarity. The electronic device according to claim 12, wherein said determining a third degree of similarity between the entity in the first query sentence and the entity in the title of each first medical document, and the When describing the fourth degree of similarity between entities in the abstract of each first medical document, the specific implementation is as follows: Determine the first Jaccard coefficient between the entity in the first query sentence and the entity in the title of each first medical document, and use the first Jaccard coefficient as the third Similarity Determine the second Jaccard coefficient between the entity in the first query sentence and the entity in the abstract of each first medical document, and use the second Jaccard coefficient as the fourth similarity Spend. The electronic device according to claim 12, wherein, when the entity recognition is performed on the first query sentence and the entity in the first query sentence is obtained, the following is specifically executed: Performing word embedding on each word in the first query sentence to obtain a word vector corresponding to each word; Vectorize each entity in the first medical knowledge graph to obtain a feature vector of each entity in the first medical knowledge graph; Determining a weight coefficient between the word vector corresponding to each word and the feature vector of each entity in the first medical knowledge graph; Weighting the feature vector of each entity in the first medical knowledge graph according to the weight coefficient to obtain the target word vector corresponding to each word; Classify each word according to the target word vector corresponding to each word to obtain the entity in the first query sentence. A computer-readable storage medium in which a computer program is stored, and the computer program is executed by a processor to implement the following method: Get the first query statement; Translating the first query sentence to obtain a second query sentence, wherein the language types of the first query sentence and the second query sentence are different; According to the first query sentence and the title and abstract of each first medical document, the first target similarity corresponding to the first query sentence and each first medical document is determined, wherein the first target similarity of each first medical document is determined. The language type of a medical document is the same as the first query sentence; According to the second query sentence and the title and abstract of each second medical document, the second target similarity corresponding to the second query sentence and each second medical document is determined. 2. The language type of the medical document is the same as the second query sentence; According to the first target similarity and the second target similarity, a target medical document corresponding to the first query sentence is determined. The computer-readable storage medium according to claim 15, wherein when the first query sentence is translated to obtain the second query sentence, the specific implementation is as follows: Acquiring a first medical knowledge graph corresponding to the first query sentence in the medical knowledge graph database; According to the first medical knowledge graph, the first query sentence is translated to obtain the second query sentence. 16. The computer-readable storage medium according to claim 16, wherein when the first query sentence is translated according to the first medical knowledge graph to obtain the second query sentence, the specific implementation is implemented: Vectorizing the first medical knowledge graph to obtain a first feature vector corresponding to the first medical knowledge graph; Performing word embedding processing on each word in the first query sentence to obtain a word vector corresponding to each word; Performing semantic feature extraction on the word vector corresponding to each word to obtain a second feature vector corresponding to the first query sentence; Splicing a second feature vector corresponding to the first query sentence with the first feature vector to obtain a first target feature vector; Translate according to the first target feature vector to obtain the second query sentence. The computer-readable storage medium according to any one of claims 15-17, wherein the first query sentence and the title and abstract of each first medical document are determined according to the first query sentence and When the first target similarity corresponding to each first medical document is described, it is specifically realized: Determine the first similarity and the second similarity between the first query sentence and the title and abstract of each first medical document respectively Performing entity recognition on the first query sentence to obtain the entity in the first query sentence; Recognizing the title and abstract of each first medical document, respectively, to obtain the entity in the title of each first medical document and the entity in the abstract; Determine the third degree of similarity between the entity in the first query sentence and the entity in the title of each first medical document, and between the entity in the abstract of each first medical document The fourth degree of similarity; According to the first degree of similarity, the second degree of similarity, the third degree of similarity, and the fourth degree of similarity, determine the first target corresponding to the first query sentence and each of the first medical documents Similarity. 18. The computer-readable storage medium according to claim 18, wherein said determining a third degree of similarity between the entity in the first query sentence and the entity in the title of each first medical document, And the fourth degree of similarity with the entities in the abstract of each first medical document, the specific realization is: Determine the first Jaccard coefficient between the entity in the first query sentence and the entity in the title of each first medical document, and use the first Jaccard coefficient as the third Similarity Determine the second Jaccard coefficient between the entity in the first query sentence and the entity in the abstract of each first medical document, and use the second Jaccard coefficient as the fourth similarity Spend. 18. The computer-readable storage medium according to claim 18, wherein when the entity recognition is performed on the first query sentence to obtain the entity in the first query sentence, the specific implementation is: Performing word embedding on each word in the first query sentence to obtain a word vector corresponding to each word; Vectorize each entity in the first medical knowledge graph to obtain a feature vector of each entity in the first medical knowledge graph; Determining a weight coefficient between the word vector corresponding to each word and the feature vector of each entity in the first medical knowledge graph; Weighting the feature vector of each entity in the first medical knowledge graph according to the weight coefficient to obtain the target word vector corresponding to each word; Classify each word according to the target word vector corresponding to each word to obtain the entity in the first query sentence.

Images