CN108932223A - It is not logged in attribute extraction method and apparatus - Google Patents
It is not logged in attribute extraction method and apparatus Download PDFInfo
- Publication number
- CN108932223A CN108932223A CN201710374631.3A CN201710374631A CN108932223A CN 108932223 A CN108932223 A CN 108932223A CN 201710374631 A CN201710374631 A CN 201710374631A CN 108932223 A CN108932223 A CN 108932223A
- Authority
- CN
- China
- Prior art keywords
- attribute
- logged
- center object
- candidate
- word
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/253—Grammatical analysis; Style critique
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
- G06F40/295—Named entity recognition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Machine Translation (AREA)
Abstract
The present invention provide it is a kind of extracted from medical files be not logged in attribute be not logged in attribute extraction method and apparatus.The attribute extraction device that is not logged in includes:Acquiring unit is configured to obtain text sentence from text document;Attribute extraction unit, is configured to extract from the text sentence and has logged in attribute, candidate be not logged in attribute, the relationship logged between attribute and described logged in attribute and the candidate is not logged in relationship between attribute;Be not logged in Attribute Recognition unit, be configured to the candidate be not logged in the case that attribute meets a part of general-purpose attribute structure by the candidate be not logged in Attribute Recognition be not logged in attribute.
Description
Technical field
The present invention relates to text analyzing and the field of data mining, more particularly to extract the attribute that is not logged in text document
Method and device.
Background technique
When carrying out text analyzing and data mining, in order to extract different objects from text document (such as medical files)
Attribute, the word for belonging to a different category and showing the one aspect of object is extracted using natural language processing technique.Citing
For, in medical files, it can be an object extremely, position, size, shape are abnormal attributes.Currently, new attribute
It just quickly generates and occurs, so, predefined attribute classification can not cover all categories in the medical files newly encountered before
Property.Therefore, it is necessary to a kind of technologies for identifying attribute that is newly-generated or being not logged in.The technology of attribute, people are not logged in based on identification
Class can be with the attribute of one specific area of fast understanding, and the semantic classes of defined attribute, to realize attribute extraction technology.
In general, Attribute Recognition technology includes the attribute extraction processing and attribute verification processing of text document.United States Patent (USP)
US8311807B2 discloses a kind of example technique, which includes:It is not logged in word decimation rule according to pre-determining, from
Candidate is extracted in text document and is not logged in word, is then based on the appearance that the candidate retrieved from text document is not logged in word
The information of frequency is not logged in verifying in word in candidate and is not logged in word.Japan Patent JP3743204B2 discloses another kind and shows
Example property technology, a kind of data analysis supporting method.According to the method for the Japanese Patent Publication, can be taken out based on the frequency of occurrences or position
It takes and is not logged in attribute.
But in medical files, center object is usually medical discovery or medical diagnosis, and different medical discoveries has
Not generally existing particular community.That is, most of particular communities are to be not logged in attribute and have the lower frequency of occurrences.It is another
Aspect, some is not attributes but the everyday expressions with the higher frequency of occurrences may be identified as attribute.Therefore, based on appearance
Frequency or position identify that these attributes are very difficult.
Summary of the invention
Therefore, in view of the record in background technique above, the disclosure aims to solve the problem that the above problem.
According to an aspect of the present invention, one kind is provided and is not logged in attribute extraction device, described device includes:It obtains single
Member is configured to obtain text sentence from text document;Attribute extraction unit is configured to extract from the text sentence and step on
Record attribute, it is candidate be not logged in attribute, the relationship logged between attribute and it is described logged in attribute and it is described it is candidate not
Log in the relationship between attribute;It is not logged in Attribute Recognition unit, is configured to be not logged in attribute in the candidate and meets general-purpose attribute
In the case where a part of structure, it is to be not logged in attribute that the candidate, which is not logged in Attribute Recognition,.
Using the present invention, it is not logged in attribute by be modified to text analyzing and data mining, to improve attribute extraction
Precision.
According to description referring to the drawings, other property features of the invention and advantage be will be evident.
Detailed description of the invention
It is incorporated in this specification and the attached drawing for constituting this specification a part illustrates the embodiment of the present invention by way of illustration,
And the principle used to explain the present invention together with verbal description.
Figure 1A, which is illustrated, extracts attribute from medical files to obtain the example of structured message.
Figure 1B illustrates the general-purpose attribute structure recognition used according to the invention obtained from structured message and is not logged in attribute
Example.
Fig. 2 is the block diagram for schematically showing the hardware configuration that technology according to an embodiment of the present invention can be achieved.
Fig. 3 is to illustrate the block diagram of the configuration for being not logged in property recognition means according to a first embodiment of the present invention.
Fig. 4 schematically shows the flow chart according to an embodiment of the present invention for being not logged in attribute recognition process.
Fig. 5 schematically shows the flow chart of step S430 as shown in Figure 4 according to the present invention.
Fig. 6 is to illustrate the block diagram of the configuration of particular community structure generating device according to a second embodiment of the present invention.
Specific embodiment
Describe exemplary embodiment of the present invention in detail below with reference to accompanying drawings.It should be noted that following description is substantial
It is only illustrative and exemplary of, and it is in no way intended to limit the present invention and its application or purposes.Unless otherwise expressly specified, no
Then the positioned opposite of component and step described in embodiment, numerical expression and numerical value are not limit the scope of the invention.Separately
Outside, technology known to those skilled in the art, method and apparatus may not be discussed in detail, but in situation appropriate
It should be a part of this specification.
It note that similar appended drawing reference and letter refer to similar project in attached drawing, therefore, once a project is one
It is defined, then need not discuss in following attached drawing to it in a attached drawing.
In the disclosure, term " first ", " second " etc. are only used to distinguish element or step, and are not intended to mean that the time
Sequentially, priority or importance.
Text sentence can be extracted from image document, text document or medical files.In the disclosure, from medical files
The text sentence of middle extraction is only the example for being used to illustrate, it is not intended to be limited the scope of the invention.
Medical files can be partially any related to diagnosis to impression (or diagnosis) part including observation (or discovery)
Document.Medical files may include radiological report, such as computer tomography (CT) diagnosis report, nuclear magnetic resonance
(NMR) diagnosis report etc. and other kinds of report, such as clinical report, preoperative and postoperative report, admission records, discharge
Brief summary etc..Exception means the Novel presentation of human body.Disease means the illness or lesion of human body.
It was found by the inventors of the present invention that writer always uses one object of natural language description, such as film, song, people
Object or tissue;Record some attributes of the object, the one aspect of one of attribute description object.So natural language
Must interosculate (for example, a center object and its attribute) on structural level.
By taking medical files as an example, for writing the various pieces of medical diagnosis report, " it was found that and diagnosis " is medical diagnosis
The major part of report, because doctor, which needs will be seen that, is shown to related reader (for example, patient and/or other phases to diagnosis
Guan doctor).In each sentence, doctor should allow reader to understand that center object is, what the attribute of the object is, institute
With even if each sentence is mutually dissimilar in form on surface layer, but most of sentence has similar structure, and therefore, we can catch
It catches these structured messages and is not logged in attribute to identify in medical files.That is, attribute extraction can be concentrated on from text
Drawing-out structure information in this document, the structured message is usually by a center object and the set of properties of the center object
At.
Therefore, the present inventor thinks, can extract big portion according to the structured message obtained from medical files
That divides is not logged in attribute.
Figure 1A, which is illustrated, extracts attribute from medical files to obtain the example of structured message.For example, from medicine text
The text sentence that shelves extract can be that " right lung S4 To, (right lung S4 is visible always for diameter 2.5cm Knot Festival が Recognize め ら れ ま The
The tubercle of diameter 2.5cm) ".The attribute extracted from text sentence includes position attribution (for example, right lung S4 (right lung S4)), size
Attribute (for example, diameter 2.5cm (diameter 2.5cm)), abnormal attribute (Li such as , Knot Festival (tubercle)) and assert attribute (Li such as , Recognize
め ら れ ま The (visible)).Abnormal attribute is the core information of sentence, and writing other attributes is to show abnormal details.
In addition it is also possible to extract the relationship between attribute.Relationship between center object and attribute can be one-to-one, one-to-many or more
To one corresponding relationship.In the case where extracting relationship between center object and attribute, it is meant that the center object with it is described
Attribute is relevant, that is to say, that the attribute can describe the center object.Therefore, between center object and attribute
Relationship refers to incidence relation between the two.Herein, attribute includes to have logged in attribute and be not logged in attribute.Furthermore, it is possible to take up a job as a doctor
Learn the relationship extracted between attribute and attribute in other sentences of document.
It is assumed that each attribute (either logged in attribute and be still not logged in attribute) can be mapped as structuring element;Belong to
Relationship between property can also be mapped as structuring element to generate attribute structure.Each structuring element can be a structure
Change type, for example, type, property content type, attribute compare type, connection part of speech in terms of center object type, attribute
Type, grammatical term for the character type and auto-correlation type.
In addition, property content type can also include several subtypes, for example, judgement, number are (for example, percentage, day
Phase), replaceable part (for example, type A, type B, early stage, advanced stage).
As shown in Figure 1A, tubercle belongs to center object type, position attribution, size attribute and asserts that attribute is attribute side
Noodles type, right lung S4, diameter 2.5cm and is visible as property content type.
Attribute can be mapped as limited structured type, such as have a center object and display centre object detail
Other attributes, in terms of they can be same attribute or in terms of different attribute, so, a center object has several attributes
The same attribute aspect of aspect, different center objects has different property contents.In medical files, doctor can not record one
In terms of all properties of a object.For example, as shown in Figure 1A, there are three center object (example such as, Knot Festival (tubercle)) tools
In terms of attribute, include position, size and judgement (Li such as , Recognize め ら れ ま The (asserting)).
In addition, the inventors found that there is the center object with attribute in a large amount of sentences in medical files,
And it was found that can be according to structured message and some predefined most of structured messages for covering document doctor writing
General-purpose attribute structure is not logged in attribute to extract.Herein, the general-purpose attribute structure can be made a reservation for based on structured message
Justice, and can be concluded with manually identifying or from training data.The general-purpose attribute structure includes at least three related knots of tool
Structure element;One of structuring element is to be not logged in attribute, and described be not logged in attribute and other two structuring members
Element has relationship;At least two structuring elements of the general-purpose attribute structure can be center object or one is
Heart object and another be describe center object attribute.
In addition, they may have different particular community structures, but in higher structured layer for different attributes
On face, they may share identical general-purpose attribute structure.Although doctor can write different sentences with different statement structures
One object is described, but according to the observation of the attribute in a large amount of sentences to medical files, main general-purpose attribute structure is simultaneously
Seldom, there is example of the following four kinds predefined general-purpose attribute structures as the disclosure:
A) the general-purpose attribute structure as shown in the S1 in Figure 1B, including two center objects and one it is related to center object
Position mark word;Then it is set as the position mark word to be not logged in attribute.
B) the general-purpose attribute structure as shown in the S4 in Figure 1B, including a center object, center object attribute, one
A unique attribute value relevant to center object and center object attribute (for example, " unique value " as shown in fig. 1b);Then
It is set as the unique attribute value to be not logged in attribute.
C) the general-purpose attribute structure as shown in the S2 in Figure 1B, including a center object, center object attribute, one
A word (for example, " unknown " as shown in fig. 1b) and an enumerated value relevant to the word for indicating particular category
Word (for example, " replaceable value " as shown in fig. 1b), the enumerated value word and center object and center object attribute
With relationship;Then it is set as the word for indicating particular category and the enumerated value word to be not logged in attribute.
D) the general-purpose attribute structure as shown in the S3 in Figure 1B, including a center object, center object attribute, one
A word (for example, " unknown " as shown in fig. 1b) for indicating particular category and modification value relevant to the word
Word (for example, " modification parameter " as shown in fig. 1b), the modification value word and center object and center object attribute
With relationship;Then it is set as the word for indicating particular category and the modification value word to be not logged in attribute.
Figure 1B illustrates the general-purpose attribute structure recognition used according to the invention obtained from structured message and is not logged in attribute
Example.As shown in fig. 1b, there are four types of general-purpose attribute structure (S1, S2, S3, S4), and every kind of general-purpose attribute structure is by three
A related structuring element composition of tool.The structuring element includes different structure type, for example, center object type
(such as center object), property content type (such as judgement).The line of connection attribute represents between center object and different attribute
Relationship.
As shown in fig. 1b, every kind of general-purpose attribute structure is not logged in attribute comprising one.For example, general-purpose attribute structure
S1 includes that two center objects and a position mark, the position mark are arranged to be not logged in attribute.If from medicine text
The structured message extracted in shelves can be matched with general-purpose attribute structure S1, then the candidate in the structured message is not logged in category
Property can be identified as being not logged in attribute.Therefore, other three kinds of general-purpose attribute structures S2, S3 and S4 also include that its respective is not logged in
Attribute, such as being not logged in as shown in attribute column for Figure 1B, respectively unknown replaceable value, unknown modifications parameter and unique value.
Therefore, according to the present invention, during being not logged in Attribute Recognition processing, the different sentences of analysis and shared medical files
Between similar structured message can improve and be not logged in Attribute Recognition.
(hardware configuration)
First by referring to Fig. 2 description can be achieved hereafter described in technology hardware configuration.Fig. 2 is that schematically show can
Realize the block diagram of the hardware configuration of technology according to an embodiment of the present invention.
Hardware configuration 200 for example including central processing unit (CPU) 210, random access memory (RAM) 220, read-only deposit
Reservoir (ROM) 230, hard disk 240, input equipment 250, output equipment 260, network interface 270 and system bus 280.In addition, hard
Part configuration 200 can pass through such as work station, server, tablet computer, laptop, desktop computer or other suitable electronics
Equipment is realized.
In the first implementation, the process for being not logged in attribute in medical files is identified according to the present invention by hardware or is consolidated
Part configures and is used as the module or component of hardware configuration 200.For example, hereinafter with reference to the device 300 of Fig. 3 detailed description
Module or component as hardware configuration 200.In the second implementation, being not logged in medical files is identified according to the present invention
The process of attribute is by being stored in the software configuration executed in ROM 230 or hard disk 240 and by CPU 210.For example, will hereinafter
The process 400 being described in detail referring to Fig. 4 is used as the program being stored in ROM 230 or hard disk 240.
CPU 210 is any suitable programmable control device (such as, processor), and may be implemented within ROM
230 or hard disk 240 (such as, memory) in various application programs execute the various functions being described hereinafter.RAM 220
It is used to temporarily store the program or data loaded from ROM 230 or hard disk 240, and is held wherein used also as CPU 210
The sky of row various processes (such as, implementing the technology being described in detail hereinafter with reference to Fig. 4 and Fig. 5) and other available functions
Between.Hard disk 240 stores much information, and such as operating system (OS), control program, is pre-stored by manufacturer or in advance various applications
It the data of definition and is pre-stored by manufacturer or the model and/or classifier of pre-generatmg.
In one implementation, input equipment 250 is for allowing user to interact with hardware configuration 200.In an example
In, user can input text document by input equipment 250.In another example, user can be touched by input equipment 250
The corresponding process for sending out of the invention.In addition, various forms, such as button, keyboard or touch screen can be used in input equipment 250.Another
In a kind of implementation, input equipment 250 is used to receive the spy from such as digital camera and/or electron medicine document file management system
Text/attribute of different electronic equipment output.
In one implementation, output equipment 260 is not logged in Attribute Recognition result (for example, new for showing to user
Exception, the center object of new type attribute in terms of etc.).Moreover, various forms, such as cathode can be used in output equipment 260
Ray tube (CRT) or liquid crystal display and/or printer.In another implementation, output equipment 260 be used for text/
Attributive analysis and the subsequent process of identification (for example, diagnostic analysis, patient's tracking, medical discovery, abnormality detection, attribute confirmation and/
Or identification, etc.) export the recognition result for being not logged in attribute.
Network interface 270 provides the interface for hardware configuration 200 to be connected to network.For example, hardware configuration 200 can be through
Data communication is carried out by network interface 270 and other electronic equipments connected via a network.Alternatively, can be hardware configuration 200
Wireless interface is provided, to carry out wireless data communication.System bus 280 can be provided in CPU 210, RAM 220, ROM
230, the data of mutual data transmission are transmitted between hard disk 240, input equipment 250, output equipment 260 and network interface 270 etc.
Path.Although referred to as bus, system bus 280 is not limited to any specific data transmission technology.
Above-mentioned hardware configuration 200 is merely illustrative, and is in no way intended to limit invention, its application, or uses.And
And for brevity, a hardware configuration is only shown in Fig. 2.But it is also possible to be matched as needed using multiple hardware
It sets.
(being not logged in Attribute Recognition processing)
(first embodiment)
The main object of the present invention is as described above according to the structured message and predefined general-purpose attribute in medical files
Structure recognition is not logged in attribute.Below with reference to Fig. 3 to Fig. 6 description, identification is not logged in attribute from medical files according to the present invention
Process.
Fig. 3 is to illustrate the block diagram of the configuration for being not logged in Attribute Recognition processing unit 300 according to a first embodiment of the present invention.
Wherein, part or all of module shown in Fig. 3 can be realized by specialized hardware.Flow chart 400 shown in Fig. 4 is institute in Fig. 3
The corresponding process of the device 300 shown.
As shown in Figure 3, device 300 includes medical files acquiring unit 310, attribute extraction unit 320, is not logged in attribute
Recognition unit 330 and it is not logged in attribute acquiring unit 340.
Firstly, input equipment 250 shown in Fig. 2 receives medical files from special electrical devices or user.Then, it inputs
Received medical files are transferred to medical files acquiring unit 310 via system bus 280 by equipment 250.Next, medicine is literary
Shelves acquiring unit 310 obtains medical files from input equipment 250 via system bus 280.
In addition, medical files acquiring unit 310 executes step S410 as shown in Figure 4, from received medical files
Obtain text sentence.As shown in Figure 4, in obtaining step S410, medical files acquiring unit 310 is from received medical files
Middle acquisition text sentence.
In step S410, medical files acquiring unit 310 obtains text sentence from medical files.In a kind of realization side
In formula, medical files acquiring unit 310 obtains medical files and extracts a text sentence from the medical files.Citing comes
It says, text sentence can be " right lung S4 To, diameter 2.5cm Knot Festival Ga Recognize め ら れ ま The (the visible diameter of right lung S4
The tubercle of 2.5cm) ", as described in above in conjunction with Figure 1A.Another text sentence can be that " scaphoid Herbert divides Class type A
It fractures possibility Ga あ り ま The (it is possible that scaphoid Herbert A type is fractured) ".
Next, attribute extraction unit 320 obtains text language from medical files acquiring unit 310 via system bus 280
Sentence.
Attribute extraction unit 320 extracts attribute from text sentence, does not step on comprising having logged in attribute and at least one candidate
Record attribute.Attribute extraction unit 320 also extracts the relationship between attribute.The natural language processing of further investigation attribute can be used
(natural language processing, NLP) method extracts the relationship between attribute and/or attribute.It is usually used to be based on
The method or machine learning method of rule.The relationship can be predefined type or represent center object attribute and dependent attributes
Simple modified relationship.In the present invention, simple relation or predefined relationship type can be used.
As described above, medical files acquiring unit 310 extracts many text sentences, therefore, attribute extraction from medical files
Attribute extraction step S420 as shown in Figure 4 can be performed in unit 320, to extract attribute, the attribute packet from text sentence
Attribute is not logged in containing attribute and at least one candidate has been logged in.
In the step s 420, attribute extraction unit 320 is extracted from text sentence has logged in attribute and candidate is not logged in category
Property.In one implementation, attribute extraction and name entity extraction (named entity extraction, NEE) are similar,
But standard NEE only handles the certain types of entity such as personage, tissue or country.In specific area, attribute extraction can be with
Attribute is extracted using the technology of similar NEE.The technology of similar NEE includes rule-based method and the side based on machine learning
Method.
It has logged in attribute extraction to define with the attribute tags of pre-determining, example as shown in Table 1 below is with mark
The attribute of label:
Table 1
Attribute extraction method is typically based on annotating data and is trained, and definition can be extracted from the text sentence of acquisition
In attribute.Attribute extraction unit 320 extracts attribute from text sentence:" scaphoid Herbert divides Class type A to fracture
Possibility Ga あ り ま The (it is possible that scaphoid Herbert A type is fractured) ".Herein, as shown in table 2, " scaphoid is (navicular
Bone) " it is the attribute for having logged in attribute and having belonged to human body." fracture " (fracture) is to have logged in attribute and belonged to the category of disease
Property." possibility Ga あ り ま The (it is possible that) " it is to have logged in attribute and belonged to the attribute asserted." Herbert divides Class type A
Attribute of (the Herbert A type) " due to being not belonging to any pre-determining is not logged in attribute for candidate.
Table 2
In addition, attribute extraction unit 320 extracts relationship from text sentence:" scaphoid Herbert divides Class type A bone
Roll over possibility Ga あ り ま The (it is possible that scaphoid Herbert A type is fractured) ".Centered on disease (for example, fracture (fracture))
The attribute of object type, human body (for example, scaphoid (scaphoid)) are asserted (for example, it may be possible to which (having can for property Ga あ り ま The
Can)) and candidate be not logged in attribute (for example, Herbert divides Class type A (Herbert A type)) be attribute in terms of type, i.e.,
The dependent attributes of center object.Therefore, all of which and center object attribute have relationship.That is, all dependent attributes are all
For describing the attribute of center object from different aspect.
In addition, in this step, stopping word list can be used for identifying that candidate does not step on from text sentence as an option
Record attribute.Neither the attribute with label, which is also not, stops word (for example, " " (), " The " (constitute helping for object component
Word)) word can be not logged in attribute for candidate.
Therefore, according to the above, attribute extraction unit 320 is taken out from medical files from medical files acquiring unit 310
The relationship between attribute and attribute is extracted in each text sentence taken.
Attribute is obtained via 280 dependence extracting unit 320 of system bus next, being not logged in Attribute Recognition unit 330
And the relationship between attribute.
Be not logged in Attribute Recognition unit 330 based on above in conjunction with general-purpose attribute structure described in Figure 1B from from text sentence
The candidate of extraction is not logged in identification in attribute and is not logged in attribute.It is executed shown in Fig. 4 so being not logged in Attribute Recognition unit 330
Be not logged in Attribute Recognition step S430, determine that candidate is not logged in whether attribute is to be not logged in attribute.
In step S430, Attribute Recognition unit 330 is not logged in using one group of general-purpose attribute structure recognition and is not logged in attribute.
In one implementation, Attribute Recognition unit 330 is not logged in general above in conjunction with one group described in Figure 1B
Attribute structure is not logged in identification in attribute from candidate and is not logged in attribute.Four kinds of general-purpose attribute knots are used below in conjunction with Fig. 5 description
Structure identifies the example for being not logged in attribute.
Flow chart 500 shown in Fig. 5 is that Attribute Recognition step S430 is not logged in shown in Fig. 4 according to the present invention
Corresponding process.
Turning now to Fig. 5, in step S510, attribute will have been logged in and candidate has not stepped on by being not logged in Attribute Recognition unit 330
Record attribute is mapped as structuring element.Due to having logged in the attribute type of attribute it is known that being therefore not logged in Attribute Recognition unit 330
Attribute has been logged according to the mapping of predefined map listing.It can be mapped according to the feature that candidate is not logged in attribute,
The feature includes distribution in sentence, frequency, location information, relationship, morphological feature and field with other attributes
Knowledge etc..These features can be extracted from text sentence and medical files.Next, being not logged in Attribute Recognition unit 330
The structuring element for being not logged in attribute is determined according to the feature.For example, in sentence listed above " Herbert divides Class
Type A (Herbert A type) " includes English glossary and replaceable label " A ".It can be mapped as shown in Table 3 below
" unknown-replaceable " content.It is structuring element that Attribute Recognition unit 330, which is not logged in, by the relationship map between attribute, therefore
There is relationship between structuring element.
Table 3
As described above, structuring element includes such as Types Below:Type, property content class in terms of center object type, attribute
Type, attribute compare type, conjunction type, grammatical term for the character type and auto-correlation type.Property content type can be further
Being divided into two-value property type, number attribute type, limited replaceable attribute type, (wherein each replaceable attribute is in pre-determining
Be defined in dictionary), (wherein each replaceable attribute is defined in the dictionary of pre-determining to unlimited replaceable attribute type
The example of specific concept), co-occurrence attribute type (usually occurring simultaneously in terms of corresponding attribute).
Attribute will have been logged in and candidate is not logged in attribute and is mapped as after having related structuring element, in step S520
In, it is not logged in Attribute Recognition unit 330 and determines whether the related structuring element of tool meets predefined general-purpose attribute
Structure.Herein, it is not logged in Attribute Recognition unit 330 and determines whether the related structuring element of tool meets at least one and lead to
With attribute structure, i.e., in the case where the related structuring element of the tool and a general-purpose attribute structure matching, general category
The candidate in position that property structure defines, which is not logged in attribute and is confirmed as one, is not logged in attribute.
It therefore meets having the general-purpose attribute structure being alternatively worth as shown in S2 in Figure 1B.The general-purpose attribute structure
Structuring meaning is:One center object is modified by a certain unknown content, which has replaceable value, and doctor's root
Subjective judgement is provided according to modifier and center object.Therefore, having the unknown content being alternatively worth must be the display center
The attribute in a certain respect of object, even if not being defined to it currently.
Based on having the general-purpose attribute structure being alternatively worth, it is not logged in Attribute Recognition unit 330 and candidate is not logged in attribute
It is determined as being not logged in attribute.
It is obtained via system bus 280 from Attribute Recognition unit 330 is not logged in next, being not logged in attribute acquiring unit 340
It takes and is not logged in attribute.
Therefore, it is not logged in attribute acquiring unit 340 and is obtained from the text sentence of medical files and be not logged in attribute, and via
What system bus 280 will acquire, which be not logged in attribute, is output to output equipment 260 and is further processed.For example, it is not logged in
Attribute acquiring unit 340 exports word " Herbert divides Class type A (Herbert A type) " as one and is not logged in attribute.This
Outside, attribute storage can will be not logged in into RAM 220, ROM 230 or hard disk 240 by being not logged in attribute acquiring unit 340.(second
Embodiment)
After being not logged in the acquisition of attribute acquiring unit 340 and being not logged in attribute, output equipment 260 can search for medicine text in user
Shelves and while wanting to check patient health status will be not logged in attribute display to user.
In another implementation, in order to improve the retrieval precision of attribute extraction, the present invention provides according to combining
The second embodiment of the device 600 of journey 400.
As shown in Figure 6, the present invention can generate particular community structure according to attribute is not logged in.Then, user can make
New attribute is extracted from medical files with the particular community structure, to improve the retrieval precision of attribute extraction.
Fig. 6 is to illustrate the block diagram of the configuration of attribute extraction device 600 according to a second embodiment of the present invention.Wherein, in Fig. 6
Some or all of shown module can be realized by specialized hardware.As shown in Figure 6, device 600 includes device 300 and particular community
Structural generation unit 610.
Device 300 obtains from medical files be not logged in attribute as described above.Next, particular community structural generation unit
610 execute particular community structural generation step S450 shown in Fig. 4, are not logged in attribute with basis and generate particular community structure.
Particular community structure can by structuring element and be not logged in community-internal one or more specific parts and replaceable part
Composition;It also may include the relationship between specific part, replaceable part and structuring element.
In one implementation, particular community structural generation unit 610 will be not logged in by least one following steps
Attribute is divided into specific part and replaceable part, to generate particular community structure:
Firstly, particular community structural generation unit 610 is not logged in the core that Attribute Recognition is not logged in attribute by analysis
Point and modifier part, the core is then appointed as specific part, the modifier part is appointed as alternatively
Part.
Secondly, particular community structural generation unit 610 identifies the numerical portion for being not logged in attribute, then by the digital section
Divide and is appointed as replaceable part.
Third, particular community structural generation unit 610 identify be not logged in attribute and at least one other attribute identical in
Hold, the content is then appointed as specific part.
Finally, particular community structural generation unit 610 identifies the word being not logged in the infinite set of attribute, it then will be described
Word in infinite set is appointed as replaceable part.
According to the attribute that is not logged in of identification, particular community structural generation unit 610 finds corresponding general-purpose attribute structure
One specific example, to generate particular community structure.It is specific in the case where encountering the text sentence that another has this attribute
Attribute structure can only extract this attribute.Because natural language is very free in form on surface layer, got over according to attribute generation is not logged in
More particular community structures can identify more specific parts for being not logged in attribute and replaceable part from medical files;
Therefore, attribute extraction can reach higher precision.
(3rd embodiment)
Device 600 can further comprise attribute updating unit, pre- really with the new particular community topology update according to generation
Fixed attribute extraction method.According to new particular community structure, it will be extracted from medical files and more be not logged in attribute, thus
Reach higher precision.
(invention application)
According to it is above-described for be not logged in Attribute Recognition processing device and method, the present invention can recognize it is different not
Attribute is logged in, moreover, generating particular community structure according to the attribute that is not logged in.
In one implementation, medical files acquiring unit 310 executes step S410 to obtain text from medical files
This sentence, for example, the text sentence can be the intentional big To of な Swollen of " Knot Festival は connect て い ま The (visit and tubercle obviously swell
Greatly) ".
Attribute extraction unit 320 executes step S420 to extract from text sentence and log in attribute and at least one candidate
It is not logged in attribute, the also relationship between extraction attribute.That is, " Knot Festival (tubercle) " He " Swollen is big (enlargement) " it is abnormal, it is to have logged in
Attribute, " connecing て い ま The (visit and) " is that candidate is not logged in attribute.
Next, being not logged in Attribute Recognition unit 330 executes step S430, not stepped on by will log in attribute and candidate
Record attribute is mapped as having related structuring element and determines whether the related structuring element of tool meets at least one
A general-purpose attribute structure, to identify that the candidate is not logged in whether attribute is to be not logged in attribute.That is, " Knot Festival (tubercle) " He " Swollen
(enlargement) greatly " is exception, so belonging to center object attribute type.Candidate is not logged in attribute " connecing て い ま The (visit and) "
Position mark between two center objects, belongs to connection relationship attribute type.Therefore, being not logged in Attribute Recognition unit 330 will
It has logged in attribute and candidate to be not logged in attribute and logged in attribute and the candidate relationship map being not logged between attribute is structure
Change element.Therefore, in the case where having related structuring element and one of general-purpose attribute structure matching, candidate is not stepped on
Record attribute can be identified as being not logged in attribute.Structuring meaning is that two center objects are connected by a certain unknown content, this is unknown
Content representation position, therefore, the unknown content must show a certain positional relationship of the center object, even if currently not
It is defined.Based on replaceable value, candidate is not logged in attribute " connecing て い ま The (visit and) " and is identified as being not logged in category
Property.It is matched that is, candidate is not logged in attribute " connecing て い ま The (visit and) " with the position mark of general-purpose attribute structure, therefore, institute
It states candidate and is not logged in attribute and be identified as being not logged in attribute.
It is not logged in the execution of attribute acquiring unit 340 step S440 and is not logged in attribute to obtain, the attribute that is not logged in can quilt
Output or for further processing.
Particular community structural generation unit 610 executes step S450 and is not logged in attribute generation particular community structure with basis.
That is, particular community structure includes two center objects and is not logged in attribute " connecing The Ru (visit and) ".In addition, particular community structure can
For attribute extraction.
In another implementation, medical files acquiring unit 310 executes step S410 with from received medical files
Middle acquisition text sentence, for example, text sentence can be common To See ら れ Ru institute See In あ Ru (the hemotoncus art of " Xue Swollen が Intraoperative Hou
It is typically seen afterwards) ".
Attribute extraction unit 320 executes step S420 to extract from text sentence and log in attribute and at least one candidate
It is not logged in attribute, and also extracts the relationship between attribute.That is, " Xue Swollen (hemotoncus) " it is abnormal, " See ら れ Ru institute See In あ Ru
(visible) " is to assert, both for having logged in attribute." Intraoperative Hou is common (postoperative usual) " it is that candidate is not logged in attribute.
Next, being not logged in Attribute Recognition unit 330 executes step S430, not stepped on by will log in attribute and candidate
Record attribute is mapped as having related structuring element and determines whether the related structuring element of tool meets at least one
A general-purpose attribute structure, to identify that the candidate is not logged in whether attribute is to be not logged in attribute.That is, " Xue Swollen (hemotoncus) " it is different
Often, so belonging to center object attribute type." See ら れ Ru institute See In あ Ru (visible) " is to assert, so belonging to contents attribute
Type.It is common (postoperative usually) that candidate is not logged in attribute " Intraoperative Hou " it is unique value in text sentence.Unique value is meant will not
It is found in having logged in attribute.Therefore, be not logged in Attribute Recognition unit 330 will log in attribute and candidate be not logged in attribute and
Having logged in attribute and the candidate relationship map being not logged between attribute is structuring element.Therefore, having related structuring
In the case where element and one of general-purpose attribute structure matching, candidate is not logged in attribute and can be identified as being not logged in attribute.Knot
Structure meaning is that a center object is modified by unique modifier, so unique modifier must be by doctor's writing this is right
A certain attribute as being different from every other object.Based on the unique value, it is common (postoperative logical that candidate is not logged in attribute " Intraoperative Hou
Often) " it is identified as being not logged in attribute.That is, candidate is not logged in attribute " Intraoperative Hou commonly (postoperative usual) " and general-purpose attribute structure
Unique value matching, therefore, the candidate is not logged in attribute and is identified as being not logged in attribute.
It is not logged in the execution of attribute acquiring unit 340 step S440 and is not logged in attribute to obtain, the attribute that is not logged in can quilt
Output or for further processing.
Particular community structural generation unit 610 executes step S450 and is not logged in attribute generation particular community structure with basis.
That is, particular community structure includes a center object and is not logged in attribute " Intraoperative Hou commonly (postoperative usual) ".In addition, particular community
Structure can be used for attribute extraction.
In another implementation, medical files acquiring unit 310 executes step S410 with from received medical files
Middle acquisition text sentence, for example, the text sentence can be, " the more い translocation of メ ラ ニ Application amount move The and doubt い ま The (bosom
Doubt the transfer more than melanin content) ".
Attribute extraction unit 320 executes step S420 to extract from text sentence and log in attribute and at least one candidate
It is not logged in attribute, and also extracts the relationship between attribute.That is, " translocation moves (transfer) " is exception, " doubting い ま The (suspection) " is
It asserts, both for having logged in attribute." the メ ラ more い of ニ Application amount (melanin content is more) " are that candidate is not logged in attribute.
Next, being not logged in Attribute Recognition unit 330 executes step S430, not stepped on by will log in attribute and candidate
Record attribute is mapped as having related structuring element and determines whether the related structuring element of tool meets at least one
A general-purpose attribute structure, to identify that the candidate is not logged in whether attribute is to be not logged in attribute.That is, " translocation moves (transfer) " is different
Often, so belonging to center object attribute type." doubting い ま The (suspection) " is to assert, so belonging to contents attribute type.It is candidate
Being not logged in attribute " the メ ラ more い of ニ Application amount (melanin content is more) " is to have the unknown of modification parameter in text sentence
Content.Therefore, it is not logged in that Attribute Recognition unit 330 will log in attribute and candidate is not logged in attribute and has logged in attribute and time
It is structuring element that choosing, which is not logged in the relationship map between attribute,.Therefore, having related structuring element and one of them
In the case where general-purpose attribute structure matching, candidate is not logged in attribute and can be identified as being not logged in attribute.Structuring meaning is one
Center object is related to unknown content, and the unknown content is modified by special parameter, so described with special parameter
Unknown content must be attribute.Based on modifier, candidate is not logged in the attribute " more い (melanin contents of メ ラ ニ Application amount
It is more) " it is identified as being not logged in attribute.That is, candidate be not logged in attribute " the メ ラ more い of ニ Application amount (melanin content is more) " with
The modifier of general-purpose attribute structure matches, and therefore, the candidate is not logged in attribute and is identified as being not logged in attribute.
It is not logged in the execution of attribute acquiring unit 340 step S440 and is not logged in attribute to obtain, the attribute that is not logged in can quilt
Output or for further processing.
Particular community structural generation unit 610 executes step S450 and is not logged in attribute generation particular community structure with basis.
That is, particular community structure includes a center object and is not logged in attribute " メ ラ ニ Application amount (melanin content) ".In addition,
Particular community structure can be used for attribute extraction.
In another implementation, particular community structural generation unit 610 generates spy after generating particular community structure
Determining attribute facilitates attribute extraction.This embodiment will store entire particular community structure, to establish particular community structure example number
According to library.In this embodiment, general-purpose attribute structure refers to the abstract form of many specific structures.
In another embodiment, it is other to generate Attribute class can be not logged in attribute by cluster for attribute extraction device;And
It sets the attribute classification to the attribute type of center object, and sets institute for the attribute that is not logged in clustered in the classification
State the property content of attribute type.Attribute extraction device can be not logged in attribute by executing following steps to cluster:Calculating is not stepped on
It records attribute and has logged in first degree of correlation between the attribute in property set, be not logged in attribute described in calculating and be not logged in category with other
Second degree of correlation between property;Attribute is not logged in based on first degree of correlation and second degree of correlation cluster.
In another implementation, in the case where user encounters frontier, for the drawing-out structure from this field
Information is needed to be carried out defined attribute type using the present invention, while extracting attribute from text.
Above-mentioned all units are the exemplary and/or preferred modules for realizing processing described in the disclosure.These units
Can be hardware cell (such as, field programmable gate array (FPGA), digital signal processor, specific integrated circuit etc.) and/
Or software module (such as, computer-readable program).It does not describe at large for realizing the unit of each step above.However,
In the case where there is the step of executing particular procedure, there may be the corresponding function modules or list for realizing the same process
First (passing through hardware and/or software realization).All combined skills of the step of passing through description and the unit corresponding to these steps
Art scheme is included in disclosure herein, as long as the technical solution that they are constituted is complete, applicable.
The process and apparatus of the present invention can be implemented in various ways.For example, can by software, hardware, firmware or
Any combination thereof implements methods and apparatus of the present invention.Unless otherwise expressly specified, otherwise this method the step of it is above-mentioned suitable
Sequence is only intended to be illustrative, and the step of method of the invention is not limited to the sequence of above-mentioned specific descriptions.In addition, one
In a little embodiments, the present invention can also be implemented as the program recorded in the recording medium comprising for realizing according to this hair
The machine readable instructions of bright method.Therefore, the present invention covers storage also for realizing program according to the method for the present invention
Recording medium.
Although some specific embodiments of the present invention, those skilled in the art has been shown in detail by example
Member is it should be understood that above-mentioned example is only intended to be illustrative, and does not limit the scope of the invention.Those skilled in the art should
Understand, above-described embodiment can be modified without departing from the scope and spirit of the present invention.The scope of the present invention is by institute
Attached claim limits.
Claims (17)
1. one kind is not logged in attribute extraction device, the attribute extraction device that is not logged in includes:
Acquiring unit is configured to obtain text sentence from text document;
Attribute extraction unit, be configured to extract from the text sentence logged in attribute, candidate is not logged in attribute, described has stepped on
Relationship between record attribute and described attribute is logged in and the candidate is not logged in relationship between attribute;
It is not logged in Attribute Recognition unit, is configured to be not logged in the feelings that attribute meets a part of general-purpose attribute structure in the candidate
Under condition, it is to be not logged in attribute that the candidate, which is not logged in Attribute Recognition,.
2. according to claim 1 be not logged in attribute extraction device, wherein the general-purpose attribute structure be it is predefined,
It and include to be not logged in attribute, at least two structuring elements, the relationship between at least two structuring element and institute
State the relationship being not logged between attribute and at least two structuring element.
3. according to claim 1 be not logged in attribute extraction device, wherein the Attribute Recognition unit that is not logged in executes such as
Lower operation:
Attribute has been logged in, the candidate is not logged in attribute and described has logged in attribute and the candidate is not logged in attribute by described
Between relationship map be structuring element;
Determine whether the related structuring element of tool meets the general-purpose attribute structure, in the related structure of tool
In the case where changing element and the general-purpose attribute structure matching, it is mapped as the time of the related structuring element of tool
It is to be not logged in attribute that choosing, which is not logged in attribute,.
4. according to claim 2 or 3 be not logged in attribute extraction device, the attribute extraction device that is not logged in further includes:
Particular community structural generation unit is configured to be not logged in attribute generation particular community structure according to, wherein the spy
Attribute structure is determined by the structuring element and described at least one specific part for being not logged in community-internal and replaceable portion
It is grouped as, and also comprising the relationship between the specific part, the replaceable part and the structuring element.
5. according to claim 4 be not logged in attribute extraction device, wherein
The attribute that is not logged in is divided into the specific part and the replaceable part by the particular community structural generation unit,
And the core and modifier part for being not logged in attribute is identified by being not logged in attribute described in analysis, and will be described
Core is appointed as the specific part, and the modifier part is appointed as the replaceable part.
6. according to claim 4 be not logged in attribute extraction device, wherein
It is not logged in the numerical portion of attribute described in the particular community structural generation unit identification, then refers to the numerical portion
It is set to the replaceable part.
7. according to claim 4 be not logged in attribute extraction device, wherein
The identical content of attribute and at least one other attribute is not logged in described in the particular community structural generation unit identification, so
The content is appointed as the specific part afterwards.
8. according to claim 4 be not logged in attribute extraction device, wherein
The word being not logged in the infinite set of attribute described in the particular community structural generation unit identification, then will be described unlimited
The word concentrated is appointed as replaceable part.
9. according to claim 2 be not logged in attribute extraction device, wherein the structuring element includes from center object
Type, property content type, attribute compare type, conjunction type, grammatical term for the character type or auto-correlation in terms of type, attribute
At least two structured types selected in type.
10. according to claim 3 be not logged in attribute extraction device, wherein the candidate is not logged in attribute and is mapped as
The structuring element, including:
The feature that candidate is not logged in attribute is extracted from the text sentence and medical files;
Determine that the candidate is not logged in the structuring element of attribute according to the feature.
11. according to claim 2 be not logged in attribute extraction device, wherein the structure of the general-purpose attribute structure
Changing element is one of following items:
It a) include two center objects and at least one position mark word relevant to center object;
It b) include a center object, a center object attribute and one and the center object and the center object
The relevant unique attribute value of attribute;
C) include a center object, a center object attribute, one expression particular category word and one with it is described
The relevant enumerated value word of word, the enumerated value word and the center object and the center object attribute, which have, to close
System;Or
D) include a center object, a center object attribute, one expression particular category word and one with it is described
The relevant modification value word of word, the modification value word and the center object and the center object attribute, which have, to close
System.
12. one kind is not logged in attribute extraction method, the attribute extraction method that is not logged in includes:
Text sentence obtaining step, for obtaining text sentence from text document;
Attribute extraction step, for extracted from the text sentence logged in attribute, candidate is not logged in attribute, described has logged in
Relationship between attribute and described attribute is logged in and the candidate is not logged in relationship between attribute;
It is not logged in Attribute Recognition step, for being not logged in the situation that attribute meets a part of general-purpose attribute structure in the candidate
Under, it is to be not logged in attribute that the candidate, which is not logged in Attribute Recognition,.
13. according to claim 12 be not logged in attribute extraction method, wherein the general-purpose attribute structure is predefined
, and include be not logged in attribute, at least two structuring elements, the relationship between at least two structuring element and
The relationship being not logged between attribute and at least two structuring element.
14. according to claim 12 be not logged in attribute extraction method, wherein described to be not logged in Attribute Recognition step packet
It includes:
Attribute has been logged in and the candidate is not logged in attribute and described has logged in attribute and the candidate is not logged in category for described
Property between relationship map be structuring element;
Determine whether the related structuring element of tool meets the general-purpose attribute structure, in the related structure of tool
In the case where changing element and the general-purpose attribute structure matching, it is mapped as the time of the related structuring element of tool
It is to be not logged in attribute that choosing, which is not logged in attribute,.
15. being not logged in attribute extraction method described in 3 or 14 according to claim 1, the attribute extraction method that is not logged in also is wrapped
It includes:
Particular community structural generation step generates particular community structure for being not logged in attribute according to, wherein described specific
Attribute structure is by the structuring element and described at least one specific part for being not logged in community-internal and replaceable part group
At, and also comprising the relationship between the specific part, the replaceable part and the structuring element.
16. according to claim 15 be not logged in attribute extraction method, wherein
In the particular community structural generation step, the attribute that is not logged in is divided into the specific part and described replaceable
Part, and the core and modifier part for being not logged in attribute is identified by being not logged in attribute described in analysis, and
The core is appointed as the specific part, the modifier part is appointed as the replaceable part.
17. according to claim 13 be not logged in attribute extraction method, wherein the structure of the general-purpose attribute structure
Changing element is one of following items:
It a) include two center objects and at least one position mark word relevant to center object;
It b) include a center object, a center object attribute and one and the center object and the center object
The relevant unique attribute value of attribute;
C) include a center object, a center object attribute, one expression particular category word and with described in one
The relevant enumerated value word of word, the enumerated value word and the center object and the center object attribute, which have, to close
System;Or
D) include a center object, an attribute of center object, expression particular category a word and
One modification value word relevant to the word, the modification value word and the center object and the center pair
The attribute of elephant has relationship.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710374631.3A CN108932223A (en) | 2017-05-24 | 2017-05-24 | It is not logged in attribute extraction method and apparatus |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710374631.3A CN108932223A (en) | 2017-05-24 | 2017-05-24 | It is not logged in attribute extraction method and apparatus |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108932223A true CN108932223A (en) | 2018-12-04 |
Family
ID=64450510
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710374631.3A Pending CN108932223A (en) | 2017-05-24 | 2017-05-24 | It is not logged in attribute extraction method and apparatus |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108932223A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109933672A (en) * | 2019-02-12 | 2019-06-25 | 北京百度网讯科技有限公司 | Handle method, apparatus, electronic equipment and the computer readable storage medium of inquiry |
-
2017
- 2017-05-24 CN CN201710374631.3A patent/CN108932223A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109933672A (en) * | 2019-02-12 | 2019-06-25 | 北京百度网讯科技有限公司 | Handle method, apparatus, electronic equipment and the computer readable storage medium of inquiry |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Olczak et al. | Artificial intelligence for analyzing orthopedic trauma radiographs: deep learning algorithms—are they on par with humans for diagnosing fractures? | |
CN105940401B (en) | System and method for providing executable annotations | |
Magge et al. | Overview of the sixth social media mining for health applications (# smm4h) shared tasks at naacl 2021 | |
US8700589B2 (en) | System for linking medical terms for a medical knowledge base | |
CN106415555B (en) | Associated system and method for pathologists report and radiological report | |
Marelli et al. | Compounding as abstract operation in semantic space: Investigating relational effects through a large-scale, data-driven computational model | |
RU2686627C1 (en) | Automatic development of a longitudinal indicator-oriented area for viewing patient's parameters | |
De Lusignan et al. | Defining datasets and creating data dictionaries for quality improvement and research in chronic disease using routinely collected data: an ontology-driven approach | |
CN106233289A (en) | Visualization method and system for patient history | |
CN107077528A (en) | Picture archiving system with the text image link based on text identification | |
Gøeg et al. | Clustering clinical models from local electronic health records based on semantic similarity | |
US20100010806A1 (en) | Storage system for symptom information of Traditional Chinese Medicine (TCM) and method for storing TCM symptom information | |
CN110427994A (en) | Digestive endoscope image processing method, device, storage medium, equipment and system | |
CN113656706A (en) | Information pushing method and device based on multi-mode deep learning model | |
JP2017134694A (en) | Attribute assignment control program, information processor and attribute assignment control method | |
US9881004B2 (en) | Gender and name translation from a first to a second language | |
CN108932223A (en) | It is not logged in attribute extraction method and apparatus | |
Nair et al. | Automated clinical concept-value pair extraction from discharge summary of pituitary adenoma patients | |
JP2017134693A (en) | Meaning information registration support program, information processor and meaning information registration support method | |
US20230377697A1 (en) | System and a way to automatically monitor clinical trials - virtual monitor (vm) and a way to record medical history | |
CN113553840A (en) | Text information processing method, device, equipment and storage medium | |
WO2021157718A1 (en) | Document creation assistance device, document creation assistance method, and program | |
CN112101034B (en) | Method and device for judging attribute of medical entity and related product | |
Keloth et al. | Extending import detection algorithms for concept import from two to three biomedical terminologies | |
CN108009157A (en) | A kind of sentence classifying method and device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20181204 |