Entity recognition method, device, equipment and medium for medical examination text

Abstract

The present application discloses a method, apparatus, device and medium for entity recognition of medical examination texts. The method comprises: obtaining medical examination texts containing a variety of expression formats; performing word segmentation processing on the medical examination texts through a professional dictionary corresponding to the medical examination texts to obtain text word segmentation information; performing component extraction on the word segmentation information to obtain component word segmentation information; performing strategy matching on the component word segmentation information according to an automaton matching strategy to obtain an entity matching sequence. The present method can be used to well identify medical examination texts containing a variety of expression formats, and accurately identify entities in the medical examination texts and corresponding entity components.

Description

Technical Field

[0001] The present application relates to the field of medical data processing technology, and in particular to an entity recognition method, device, equipment and medium for medical examination texts. Background Art

[0002] The study of medical record data is an important part of medical research, and most medical record data exists in the form of unstructured data, especially text. Among these unstructured medical texts, there are a large number of test data such as testing, immunohistochemistry, microbiology, and genetics. The formats of these data are diverse, and there are various professional terms in the expression, and there are mixed Chinese and English, misspellings, abbreviations, etc. The content has multiple attributes such as quantitative attributes, qualitative attributes, and descriptive attributes. In addition, the writing formats of different doctors in different hospitals are also different, so it is very difficult to identify the entities of medical texts. Summary of the invention

[0003] In order to solve the above-mentioned problems existing in the background technology, the embodiments of the present application creatively provide a method, device, equipment and medium for entity recognition of medical examination texts.

[0004] According to the first aspect of the embodiment of the present application, a method for entity recognition of medical examination text is provided, the method comprising: obtaining a medical examination text containing multiple expression formats; performing word segmentation processing on the medical examination text through a professional dictionary corresponding to the medical examination text to obtain text segmentation information; performing component extraction on the word segmentation information to obtain component segmentation information; performing strategy matching on the component segmentation information according to an automaton matching strategy to obtain an entity matching sequence.

[0005] According to one embodiment of the present application, the expression format includes at least one of the following formats: Chinese format, English format, numerical format, symbol format; the professional dictionary includes at least one of the following word categories: core words, unit words, trend words, relationship words, trigger words, symbol words, modified descriptive words, quantitative descriptive words, qualitative descriptive words and useless words.

[0006] According to one embodiment of the present application, the medical examination text is segmented using a professional dictionary corresponding to the medical examination text to obtain text segmentation information, including: text segmentation of the medical examination text based on a maximum length segmentation pattern and the professional dictionary to obtain text segmentation information.

[0007] According to an implementation manner of the present application, the strategy matching of the component word segmentation information according to the automaton matching strategy to obtain the entity matching sequence includes: obtaining a first word segmentation of the component word segmentation information in a specified order; matching the first word segmentation with the start state of multiple automaton matching strategies to determine that the first word segmentation satisfies at least one first matching strategy of the start state; obtaining a second word segmentation adjacent to the first word segmentation, matching the second word segmentation with the transition state of the first matching strategy to determine that the second word segmentation satisfies the second matching strategy of the transition state; judging whether the second word segmentation satisfies the stop state of the second matching strategy; if it is judged that the second word segmentation satisfies the stop state of the first matching strategy, determining the first word segmentation and the second word segmentation as an entity matching sequence.

[0008] According to an embodiment of the present application, the method further includes: if the first participle does not have a first matching strategy that satisfies the starting state, obtaining a second participle adjacent to the first participle; matching the second participle with the starting state of the automaton matching strategy to determine a third matching strategy that satisfies the starting state.

[0009] According to an embodiment of the present application, the method further includes: if it is determined that the second participle does not satisfy the stop state of the first matching strategy, obtaining a third participle adjacent to the second participle; matching the third participle with the transition state of the second matching strategy, and determining that the third participle satisfies a fourth matching strategy of the transition state.

[0010] According to one embodiment of the present application, the method also includes: determining a corresponding segmentation strategy and a data filtering strategy based on an automaton matching strategy; segmenting the entity matching sequence according to the segmentation strategy to obtain an entity segmentation sequence; and filtering data on the entity segmentation sequence according to the data filtering strategy to obtain an entity recognition sequence.

[0011] According to the second aspect of the embodiment of the present application, there is also provided an entity recognition device for medical examination texts, the device comprising: an acquisition module for acquiring medical examination texts containing multiple expression formats; a word segmentation module for performing word segmentation processing on the medical examination texts through a professional dictionary corresponding to the medical examination texts to obtain text segmentation information; an extraction module for performing component extraction on the word segmentation information to obtain component segmentation information; and a matching module for performing strategy matching on the component segmentation information according to an automaton matching strategy to obtain an entity matching sequence.

[0012] According to one embodiment of the present application, the word segmentation module includes: performing text segmentation on the medical test text based on the maximum length word segmentation mode and the professional dictionary to obtain text segmentation information.

[0013] According to an embodiment of the present application, the matching module includes: an acquisition submodule, which is used to acquire a first participle of the component participle information in a specified order; a matching submodule, which is used to match the first participle with a start state of multiple automaton matching strategies, and determine that the first participle satisfies at least one first matching strategy of the start state; the acquisition submodule is also used to acquire a second participle adjacent to the first participle, match the second participle with a transition state of the first matching strategy, and determine that the second participle satisfies the second matching strategy of the transition state; a judgment submodule, which is used to judge whether the second participle satisfies the stop state of the second matching strategy; and a determination submodule, which is used to determine the first participle and the second participle as an entity matching sequence if it is judged that the second participle satisfies the stop state of the first matching strategy.

[0014] According to one embodiment of the present application, the acquisition submodule is used to acquire a second participle adjacent to the first participle if the first participle does not have a first matching strategy that satisfies the starting state; the matching submodule is used to match the second participle with the starting state of the automaton matching strategy to determine that the second participle satisfies a third matching strategy for the starting state.

[0015] According to one embodiment of the present application, the acquisition submodule is further used to acquire a third participle adjacent to the second participle if it is determined that the second participle does not satisfy the stop state of the first matching strategy; the matching submodule is further used to match the third participle with the transition state of the second matching strategy to determine that the third participle satisfies the fourth matching strategy of the transition state.

[0016] According to one embodiment of the present application, the device also includes: a determination module, which is used to determine the corresponding segmentation strategy and data filtering strategy according to the automaton matching strategy; a segmentation module, which is used to segment the entity matching sequence according to the segmentation strategy to obtain an entity segmentation sequence; and a filtering module, which is used to filter the entity segmentation sequence according to the data filtering strategy to obtain an entity recognition sequence.

[0017] According to the third aspect of the present application, an electronic device is provided, which includes: one or more processors; a storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement any of the methods described above.

[0018] According to a fourth aspect of the present application, a computer-readable storage medium is provided, wherein the storage medium comprises a set of computer-executable instructions, and when the instructions are executed, the instructions are used to execute any of the methods described above.

[0019] The entity recognition method, device and computer-readable storage medium for medical examination text provided in the embodiments of the present application perform word segmentation processing on the medical examination text through a professional dictionary, and perform component extraction on the text segmentation information to obtain more accurate and detailed component segmentation information, and then perform strategy matching on the component segmentation information through an automatic machine matching strategy to obtain an entity matching sequence. The method realizes fine-grained splitting of the medical examination text through professional dictionaries and component extraction, and then matches it through an automatic machine matching strategy. It can well identify medical examination texts containing various expression formats, accurately identify entities in the medical examination text and corresponding entity components, and is conducive to the subsequent standardized output of the obtained entities and the research of entity data.

[0020] It should be understood that the teachings of the present application are not required to achieve all of the beneficial effects described above, but specific technical solutions can achieve specific technical effects, and other embodiments of the present application can also achieve beneficial effects not mentioned above. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present application will become easily understood. In the accompanying drawings, several embodiments of the present application are shown in an exemplary and non-limiting manner, wherein:

[0022] In the drawings, the same or corresponding reference numerals represent the same or corresponding parts.

[0023] Figure 1 A schematic diagram of the implementation process of an entity recognition method for medical examination text according to an embodiment of the present application is shown;

[0024] Figure 2 The schematic diagram shows the automatic matching process of an entity recognition method for medical examination text in an embodiment of the present application. Figure 1 ;

[0025] Figure 3 The schematic diagram shows the automatic matching process of an entity recognition method for medical examination text in an embodiment of the present application. Figure 2 ;

[0026] Figure 4 A schematic diagram of implementation modules of an entity recognition device for medical examination texts according to an embodiment of the present application is shown;

[0027] Figure 5 A schematic block diagram of an exemplary electronic device according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0028] The principles and spirit of the present application will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the present application, and are not intended to limit the scope of the present application in any way. On the contrary, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art.

[0029] The technical solution of the present application is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0030] Figure 1 A schematic diagram of the implementation flow of an entity recognition method for medical examination text according to an embodiment of the present application is shown.

[0031] See also Figure 1 According to the first aspect of the embodiment of the present application, a method for entity recognition of medical examination text is provided, and the method includes: operation 101, obtaining a medical examination text containing multiple expression formats; operation 102, performing word segmentation processing on the medical examination text through a professional dictionary corresponding to the medical examination text to obtain text segmentation information; operation 103, performing component extraction on the word segmentation information to obtain component word segmentation information; operation 104, performing strategy matching on the component word segmentation information according to an automaton matching strategy to obtain an entity matching sequence.

[0032] The entity recognition method for medical test text provided in the embodiment of the present application performs word segmentation processing on medical test text containing multiple expression formats through a professional dictionary to obtain text segmentation information, and then further extracts the text segmentation information to obtain more accurate and detailed component segmentation information, and then performs strategy matching on the component segmentation information through an automaton matching strategy, thereby obtaining an entity matching sequence. This method realizes fine-grained splitting of medical test texts through professional dictionaries and component extraction, and then matches through an automaton matching strategy, which can well identify medical test texts containing various expression formats. By further processing the entity matching sequence, standardized entities and corresponding entity components can be obtained, which is conducive to standardized output of the obtained entities, and is conducive to output entity data and subsequent data research. It can improve the generalization and expansion capabilities of entity recognition and ensure the recall rate.

[0033] In operation 101 of the method, the electronic medical record of at least one designated patient may be obtained through the electronic medical record database. The electronic medical record contains medical examination texts. The medical examination texts may specifically be descriptive texts in the examination report issued after the medical examination of the designated patient. The medical examination texts contain a large amount of data such as examination, immunohistochemistry, microbiology, and gene, such as the report modules such as "doctor's advice", "preliminary diagnosis", and "description" in the examination report. The medical examination texts are unstructured data, which usually contain one or several sentences. There is at least one expression format in the medical examination texts, such as Chinese format, English format, numerical format, and symbol format.

[0034] Among them, the Chinese format is used to represent the presence of Chinese expressions in medical test texts, such as "leukocytosis", "hemoglobin concentration", etc. The English format is used to represent the presence of English expressions in medical test texts, such as "diabetes", "DNA", etc. The numerical format is used to represent the presence of numerical expressions in medical test texts, such as "84", "90", etc. The symbol format is used to represent the presence of symbolic expressions in medical test texts, such as "β", "μm", etc. It can be understood that in conventional medical test texts, there are usually multiple expression formats.

[0035] In operation 102 of the method, the professional dictionary is a data dictionary obtained by sorting data in the field of medical examination text. The professional dictionary of the method includes at least one of the following word classes: core words, unit words, trend words, relational words, trigger words, symbol words, modified descriptive words, quantitative descriptive words, qualitative descriptive words and useless words.

[0036] Core words are the core words that need to be identified for medical test texts. An entity without core words does not constitute a test entity. Furthermore, for different test types, core words can be further subdivided, for example: β2 microglobulin, leukocytouria, hemoglobin concentration, etc. are test core words; CK8 / 18, EBERS, etc. are immunohistochemistry core words; HER2, K-RAS, SLC7A13, etc. are gene core words; respiratory viruses, Filoviridae, Poxviridae, etc. are microbial core words.

[0037] Unit words are data units that need to be identified for medical test texts. Most test entities have corresponding data units. For example, test data units include but are not limited to: CFU / ML, MG / 24H urine volume, cell count / UL, MG / GCR, etc.; but some test entities may not have units, such as genes.

[0038] Trend words are usually used in medical examination texts to describe the comparative trend between the current examination value and the normal value range. This comparative trend is usually represented by "↓" or "↑". "↓" is used to represent a downward trend, and "↑" is used to represent an upward trend. This type of description is a trend word.

[0039] Relational words. When doctors are writing medical test texts, they will sometimes compare the test results of the current test items with those of the previous round of similar test items. In this case, descriptions such as "significantly higher than before", "increased than before", "gradually decreased" will appear. These descriptions are relational words.

[0040] Trigger words. In the entity recognition of the inspection category, there are many obvious trigger words, such as immunohistochemistry trigger words: immunohistochemistry, immunohistochemistry results, immunohistochemistry indication, etc., gene trigger words: fusion gene, gene, mutation detection, etc.

[0041] Symbol words are symbolic expressions that need to be connected to identify the entire inspection entity, such as "space", "(", ")", etc.

[0042] Modified descriptors are some modifiers in test identification, for example, some modifiers in immunohistochemistry include "local", "small focus", etc., and gene modifiers in genes include "abundance", "mutation frequency", etc.

[0043] Quantitative descriptors are numerical descriptors for test entities. For example, when the description "Hepatitis B surface antibody: 535.19mIU / ml" exists in the medical test text, "535.19" is a quantitative descriptor. Quantitative descriptors can be identified through simple quantity regularization.

[0044] Qualitative descriptors are qualitative descriptors for test entities, such as "negative", "positive", "gene mutation", "no mutation", etc.

[0045] Useless words are words that are used to connect the context but have no actual meaning, such as: "equal", "equal inspection", "equal auxiliary inspection", etc.

[0046] By using word segmentation tools in conjunction with professional dictionaries to segment medical test texts, text segmentation information can be obtained.

[0047] In operation 103 of the present method, since there are too many types of expression formats of medical test texts, it is usually impossible to accurately segment each type of words in one word segmentation. Based on this, the present method extracts components from the word segmentation information to obtain component word segmentation information. The specific component extraction method can be determined according to the word type of the professional dictionary and the characteristics of the word segmentation tool. For example, when the word segmentation tool has a poor effect on the word segmentation of English and numerical values, the present method can design a component extraction method for core words and quantitative descriptive words to extract the text word segmentation information, that is, to extract core words and quantitative descriptive words from the text word segmentation information to obtain component word segmentation information, so that the component word segmentation information has a higher granularity than the text word segmentation information, which is beneficial to improving the accuracy of subsequent automatic machine matching.

[0048] In operation 104 of the method, each automaton matching strategy is composed of a two-dimensional array, and a single array represents the state transition of the automaton. The method is provided with multiple automaton matching strategies, and the automaton matching strategy is a preset strategy. Specifically, the automaton matching strategy can be obtained by analyzing the description rules of the historical medical examination text and pre-writing and designing. The historical medical examination text is matched using the automaton matching strategy to obtain an entity matching sequence, and the entity matching sequence includes words and corresponding word attributes.

[0049] In one implementation scenario, the automaton matching strategy can be expressed as: Strategy 1: ["key_state", "num_desc_state", "num_state", "unit_state"], Strategy 2: ["key_state", "num_desc_state", "relation_state", "num_state", "unit_state"].

[0050] In strategy 1 and strategy 2, "key_state" indicates the core word state, which identifies the core word component in the component segmentation information; "num_desc_state" is the qualitative description word state, which identifies the qualitative description word; "num_state" is the quantitative description word state, which identifies the quantitative description word; "unit_state" is the unit word state, which identifies the unit word; "relation_state" is the relation word state, which identifies the relation word. In order to accurately complete the recognition, the design of the automaton strategy will be designed into an overall strategy pattern according to the context of the entity core word. For example: when the medical test text is "hemoglobin 84g / L↓", the corresponding automaton matching strategy is: ["key_state", "num_state", "unit_state", "trend_state"], and the entity matching sequence obtained by matching is: key_state (hemoglobin, core word) = > num_state (84, quantitative description) = > unit_state (g / L, unit) = > trend_state (↓, trend).

[0051] According to one embodiment of the present application, operation 102, the medical examination text is segmented by a professional dictionary corresponding to the medical examination text to obtain text segmentation information, including: text segmentation of the medical examination text based on a maximum length segmentation mode and a professional dictionary to obtain text segmentation information.

[0052] Specifically, this method obtains the medical inspection text and loads the corresponding professional dictionary. The medical inspection text is segmented through the professional dictionary. This application can use an open source word segmentation engine as a word segmentation tool, such as the jieba word segmentation engine. Specifically, this method first loads the professional dictionary into the word segmentation engine, and sets the weight of the loaded professional dictionary to be higher than the weight of the default dictionary of the word segmentation engine to ensure that the word segmentation engine can segment the medical inspection text according to the professional dictionary. The word segmentation mode set by the word segmentation engine is selected as the maximum length word segmentation mode, so as to obtain the maximum length of the text segmentation information. For example, when a text contains "very good looking", the professional dictionary corresponding to the word segmentation engine loads two words, namely: "good looking" and "very good looking". The word segmentation engine divides the "very good looking" in the text into one word, that is, the word segmentation engine preferentially cuts out the word with the maximum length in the corresponding professional dictionary in the medical inspection text, and performs text segmentation on the medical inspection text through the word segmentation engine and the professional dictionary to obtain the text segmentation information.

[0053] It should be added that when the word engine is the Jieba word segmentation engine, since the Jieba word segmentation engine has the advantage of performing Chinese word segmentation, it performs poorly on the word segmentation of English words, data words, symbol words, etc., so in the obtained text word segmentation information, there will be continuous English words, quantitative descriptive words, and symbol words merged together without word segmentation. However, there are usually a lot of English and quantities and some special non-Chinese characters in medical examination texts. Based on this, this method needs to perform operation 103 to extract components from the text word segmentation information to solve the problem that the Jieba word segmentation engine has insufficient processing capabilities for English words, quantitative descriptive words, and some symbol words.

[0054] In the operation of component extraction, it is necessary to perform fine-grained segmentation of text segmentation information, accurately identify the component information of entities, and process the combination formed by continuous quantitative descriptors, English words, symbol words, etc. The combination formed by continuous English words, quantitative descriptors, and symbol words often contains English core words, unit words, quantitative descriptors, trend words, etc. For example, when the medical test text is "Total Prostate Specific Antigen (T-PSA) 9.89NG / ML↑", the jieba segmentation engine is used for segmentation, and the text segmentation information obtained is "Total Prostate Specific Antigen" and "(T-PSA) 9.89NG / ML↑", where "Total Prostate Specific Antigen" is the Chinese core word of the test, "T-PSA" in "(T-PSA) 9.89NG / ML↑" is the English core word of the test, "9.89" is a quantitative descriptor, "NG / ML" is a unit word, and "↑" is a trend word. By extracting components from text segmentation information, "total prostate-specific antigen (T-PSA) 9.89NG / ML↑" is processed into "total prostate-specific antigen", "T-PSA", "9.89", "NG / ML", "↑", thereby achieving highly fine-grained segmentation of medical test texts and obtaining component segmentation information.

[0055] According to the needs, the specific component extraction methods include but are not limited to core word extraction, quantitative description word extraction, unit word extraction, custom component extraction, etc. Among them, core word extraction is used to extract core words from continuous words composed of English words, symbol words, unit words and trend words. For example, the text segmentation information is "(T-PSA)9.89NG / ML↑". By extracting the core words in the text segmentation information, extracting the English words in the brackets, and then matching whether there is a test English core word corresponding to "T-PSA" in the professional dictionary, the component of "T-PSA" is determined as the test English core word according to the matching result.

[0056] Quantitative descriptive word extraction is used to extract quantitative descriptive words from continuous words composed of English words, symbol words, unit words and trend words. For example, the text segmentation information is "(T-PSA)9.89NG / ML↑". Based on the regular expression, quantitative descriptive word extraction is performed on "(T-PSA)9.89NG / ML↑" to obtain the quantitative descriptive word "9.89".

[0057] Unit word extraction is used to extract unit words from consecutive words composed of English words, symbol words, unit words and trend words. The unit words can be matched by professional dictionaries or unit dictionaries. The matching strategy is based on the longest unit matching principle. For example, if the text segmentation information is "(T-PSA)9.89NG / ML↑", all units in the unit dictionary are loaded and matched with "(T-PSA)9.89NG / ML↑". According to the longest unit matching principle, the unit word "NG / ML" is obtained.

[0058] Custom component extraction: For some special cases, a custom function can be used to form an extraction method for custom components, so as to extract custom components from text segmentation information.

[0059] According to an embodiment of the present application, operation 104, strategy matching is performed on the component word segmentation information according to the automaton matching strategy to obtain an entity matching sequence, including: first, obtaining a first word segmentation of the component word segmentation information in a specified order; then, matching the first word segmentation with the start state of multiple automaton matching strategies, and determining that the first word segmentation satisfies at least one first matching strategy of the start state; then, obtaining a second word segmentation adjacent to the first word segmentation, matching the second word segmentation with the transition state of the first matching strategy, and determining that the second word segmentation satisfies the second matching strategy of the transition state; then, determining whether the second word segmentation satisfies the stop state of the second matching strategy; finally, if it is determined that the second word segmentation satisfies the stop state of the first matching strategy, the first word segmentation and the second word segmentation are determined as an entity matching sequence.

[0060] According to an embodiment of the present application, the method further includes: if it is determined that the second participle does not satisfy the stop state of the first matching strategy, obtaining a third participle adjacent to the second participle; matching the third participle according to the transition state of the second matching strategy, and determining a fourth matching strategy in which the three participles satisfy the transition state.

[0061] Figure 2 The schematic diagram shows the automatic matching process of an entity recognition method for medical examination text in an embodiment of the present application. Figure 1 . Figure 3 The schematic diagram shows the automatic matching process of an entity recognition method for medical examination text in an embodiment of the present application. Figure 2 .

[0062] See also Figure 2 and Figure 3 , the matching operation of the automaton matching strategy is to load the component word segmentation information and the automaton matching strategy into the automaton engine, and then match the component word segmentation information with the automaton matching strategy. The automaton engine strategy mainly includes the start state, the transfer state, and the stop state.

[0063] The start state is the state where the automaton matching strategy starts matching. The start state is a judgment strategy for judging whether the specified word segmentation meets the set strategy. According to actual needs, in general, this method can use the core word strategy, trigger word strategy, etc. as the start state. Therefore, the start state is composed of one or more different types of transfer states. In the automaton engine, there can be multiple automaton matching strategies. In the calculation process of the automaton engine, the specified order to obtain the first word segmentation can be to select the word segmentation in sequence, that is, first obtain the first word segmentation in sequence, and then match the start strategy of each automaton matching strategy with the first word segmentation. If the match is successful, the start state is obtained and the next step, that is, the matching of the transfer state, is executed; if the match is unsuccessful, the next automaton matching strategy is selected in sequence, and the start strategy of each automaton matching strategy is matched with the first word segmentation. If the match is unsuccessful, the automaton directly exits and the recognition fails.

[0064] For example, the automaton matching strategy is ["key_state", "num_state", "unit_state", "trend_state"], and the component segmentation information is [("hemoglobin", "core word"), ("84", "quantitative description word"), ("g / L", "unit word"), ("↓", "trend word")].

[0065] When the starting state of the automaton matching strategy is the core word strategy, the match with the first word in the component word segmentation information is successful. When the starting state of the automaton matching strategy is not the core word strategy, the match with the first word in the component word segmentation information fails, and the starting state of other automaton matching strategies is selected to match the first word.

[0066] The transition state is composed of multiple different types of state strategies, and is usually also a judgment strategy. When the first word is matched to the start state, the transition state is mainly used to judge whether the second word after the first word satisfies the strategy set by the transition state. The specific judgment process is similar to the judgment process of the start state.

[0067] First, according to the starting state of the automaton matching strategy, the corresponding set of transition states is obtained, and a transition state is selected in order to match the second participle. If the match is successful, the next state set is directly obtained according to the previous part of the state, and the transition state is re-matched according to the third participle located after the second participle; if the match fails, the next automaton matching strategy is selected in order to match the second participle. If no match is successful, the automaton exits directly and the recognition fails.

[0068] The following is an example. In a specific implementation scenario, the automaton matching strategy of the present application can be set to: ["key_state", "num_state", "unit_state", "trend_state"], and the component segmentation information is: [("hemoglobin", "core word"), ("84", "quantitative descriptor"), ("g / L", "unit word"), ("↓", "trend word")], when the starting state "key_state" successfully matches the first segmentation [("hemoglobin", "core word"), the set of transition states obtained is ["num_state"], and the transition state is matched with the second segmentation ("84", "quantitative descriptor"). If the match is successful, and so on, the set of transition states ["unit_state"] is matched with the third segmentation. If unsuccessful, other starting states that can successfully match the first segmentation are continued to select states of other automaton matching strategies to match the second segmentation.

[0069] Stop state: When the automaton matches the transfer state and satisfies the stop transfer strategy, the automaton will stop transferring. The specific strategy operation method is similar to the start state and transfer state. Usually, running to the stop transfer strategy means that an entity recognition is completed. For example, when the stop state is set to a trend word, when the automaton matches the segmentation of the trend word, it will automatically exit. It can be understood that the automaton is set with a stop state to avoid falling into an infinite loop.

[0070] According to an embodiment of the present application, the method further includes: if the first participle does not have a first matching strategy that satisfies the starting state, obtaining a second participle adjacent to the first participle; matching the second participle with the starting state of the automaton matching strategy, and determining a third matching strategy that satisfies the starting state of the second participle.

[0071] If after matching the first word with all the automaton matching strategies, it is determined that none of the automaton matching strategies contains the first matching strategy that satisfies the start state, the second word adjacent to the first word is obtained. Then, the second word is matched with the start state of all the automaton matching strategies to determine the third matching strategy that satisfies the start state of the second word. The remaining operations are similar to the above matching operations and will not be described in detail below.

[0072] According to one embodiment of the present application, the method also includes: first, determining the corresponding segmentation strategy and data filtering strategy according to the automaton matching strategy; then, segmenting the entity matching sequence according to the segmentation strategy to obtain an entity segmentation sequence; and then, filtering the entity segmentation sequence according to the data filtering strategy to obtain an entity recognition sequence.

[0073] It should be added that, since the automaton recognition is in greedy mode, the longest length match will be performed according to the defined automaton matching strategy. For example, the medical test text is "Immunohistochemistry Tips: CD3 (+), Ki-67 (25%)". Through operations 102 to 104, the entity matching sequence obtained is: [("immunohistochemistry", "guide word"), ("tip": "useless word"), ("cd3", "core word"), ("(", "symbol word"), ("+", "qualitative description word"), (")", "symbol word"), ("ki-67", "core word"), ("(", "symbol word"), ("25", "quantitative" description word), ("%", "unit word"), ("(", "symbol word")], the entity matching sequence contains multiple immunohistochemistry results and some useless special symbols, which need to be processed to facilitate the standardized output of the subsequent entity recognition sequence and subsequent further use.

[0074] Specifically, this method first needs to determine the corresponding segmentation strategy and data filtering strategy according to the automaton matching strategy. The segmentation strategy is used to segment the entity matching sequence. This method uses the starting state of the automaton matching strategy to determine the segmentation strategy. Specifically, this method sets the segmentation position before the first word in the text that does not match the starting state. For example: the entity matching sequence is [("immunohistochemistry", "guide word"), ("prompt": "other useless words"), ("cd3", "core word"), ("(", "symbol"), ("+", "qualitative description"), (")", "symbol"), ("ki-67", "core word"), ("(", "symbol"), ("25", "quantitative"), ("%", "unit"), ("(", "symbol")]. The starting state is the core word strategy and the guide word strategy. The starting states in the above text are ("immunohistochemistry", "guide word"), ("cd3", "core word"), ("(", "symbol"), ("25", "quantitative"), ("%", "unit"), ("(", "symbol")]. ") and ("ki-67", "core word"), among which ("immunohistochemistry", "guide word") belongs to the first segmentation word that matches the start state, ("cd3", "core word") belongs to the second segmentation word that matches the start state, and ("ki-67", "core word") belongs to the third segmentation word that matches the start state. According to the segmentation strategy of "setting the segmentation position before the segmentation word that does not match the start state in the text", the segmentation position is before ("cd3", "core word") and before ("ki-67", "core word"), and there are three entity segmentation sequences obtained, namely:

[0075] Sequence 1: [[("immunohistochemistry", "guide word"), ("hint": "useless word")];

[0076] Sequence 2: [("cd3", "core word"), ("(", "symbol word"), ("+", "qualitative description word"), (")", "symbol word")];

[0077] Sequence three: [("ki-67", "core word"), ("(", "symbol word"), ("25", "quantitative description word"), ("%", "unit word"), ("(", "symbol word")]].

[0078] By segmenting the entity matching sequence, it is helpful to subsequently filter the data for each entity segment sequence, thereby improving the accuracy of data filtering.

[0079] The data filtering strategies of this method include a non-core word segment filtering strategy and a useless word filtering strategy.

[0080] The strategy for filtering out non-core words is: if there is no core word in the entity segmentation sequence, the entity segmentation sequence is discarded. The strategy for filtering out useless words is: filtering out the meaningless words in the entity segmentation sequence according to the components corresponding to the words, wherein the meaningless components can be preset.

[0081] For example, the above sequence 1: [[("immunohistochemistry", "guide word"), ("prompt": "useless word")] does not contain a core word, which meets the filtering conditions of the no-core word segment filtering strategy, and the sequence 1 is discarded.

[0082] In sequence 2: the symbol words are participles without actual meaning, so they are discarded, and the entity filtering sequence obtained is: [("cd3", "core word"), ("+", "qualitative description word")].

[0083] Similarly, in sequence three, the symbol word is a word without actual meaning, so it is discarded, and the entity filtering sequence obtained is: [("ki-67", "core word"), ("25", "quantitative word"), ("%", "unit word")].

[0084] All entity filtering sequences were integrated to obtain the entity recognition sequence: [[("cd3", "core word"), ("+", "qualitative description")], [("ki-67", "core word"), ("25", "quantitative"), ("%", "unit")]]. The entity recognition sequence thus obtained has the characteristics of standardization, which is conducive to the formation of structured data, and is conducive to subsequent data query, data analysis, report generation and other operations, providing accurate and standardized data basis for the research of medical record data.

[0085] Figure 4 A schematic diagram of implementation modules of an entity recognition device for medical examination texts according to an embodiment of the present application is shown.

[0086] See also Figure 4 According to the second aspect of the embodiment of the present application, there is also provided an entity recognition device for medical examination texts, the device comprising: an acquisition module 401, used to acquire medical examination texts containing multiple expression formats; a word segmentation module 402, used to perform word segmentation processing on the medical examination texts through the professional dictionary corresponding to the medical examination texts, and obtain text segmentation information; an extraction module 403, used to extract components from the word segmentation information, and obtain component word segmentation information; a matching module 404, used to perform strategy matching on the component word segmentation information according to the automaton matching strategy, and obtain an entity matching sequence.

[0087] According to an embodiment of the present application, the word segmentation module 402 includes: performing text segmentation on the medical test text based on the maximum length word segmentation mode and the professional dictionary to obtain text segmentation information.

[0088] According to an embodiment of the present application, the matching module 404 includes: an acquisition submodule 4041, which is used to acquire a first participle of the component participle information in a specified order; a matching submodule 4042, which is used to match the first participle with the start state of multiple automaton matching strategies, and determine that the first participle satisfies at least one first matching strategy of the start state; the acquisition submodule 4041 is also used to acquire a second participle adjacent to the first participle, match the second participle with the transition state of the first matching strategy, and determine that the second participle satisfies the second matching strategy of the transition state; a judgment submodule 4043, which is used to judge whether the second participle satisfies the stop state of the second matching strategy; and a determination submodule 4044, which is used to determine the first participle and the second participle as an entity matching sequence if it is judged that the second participle satisfies the stop state of the first matching strategy.

[0089] According to one embodiment of the present application, the acquisition submodule 4041 is used to obtain the second participle adjacent to the first participle if the first participle does not exist in the first matching strategy that satisfies the starting state; the matching submodule 4042 is used to match the second participle with the starting state of the automaton matching strategy to determine that the second participle satisfies the third matching strategy of the starting state.

[0090] According to one embodiment of the present application, the acquisition submodule 4041 is also used to obtain the third participle adjacent to the second participle if it is determined that the second participle does not satisfy the stop state of the first matching strategy; the matching submodule 4042 is also used to match the third participle with the transition state of the second matching strategy to determine that the third participle satisfies the fourth matching strategy of the transition state.

[0091] According to one embodiment of the present application, the device also includes: a determination module 405, which is used to determine the corresponding segmentation strategy and data filtering strategy according to the automaton matching strategy; a segmentation module 406, which is used to segment the entity matching sequence according to the segmentation strategy to obtain an entity segmentation sequence; a filtering module 407, which is used to filter the entity segmentation sequence according to the data filtering strategy to obtain an entity recognition sequence.

[0092] It should be noted that the above description of an embodiment of an entity recognition device for medical examination text is different from the above description of Figures 1 to 3 The description of the method embodiment shown is similar, with the same Figures 1 to 3 For technical details not disclosed in the embodiment of the entity recognition device for medical examination text of this application, please refer to the aforementioned Figures 1 to 3 The method embodiment shown in the figure can be understood by referring to the description, and will not be described in detail for the sake of saving space.

[0093] According to the third aspect of the present application, an electronic device is provided, comprising: one or more processors; a storage device for storing one or more programs, and when the one or more programs are executed by one or more processors, the one or more processors implement any of the above methods.

[0094] According to a fourth aspect of the present application, a computer-readable storage medium is provided, the storage medium comprising a set of computer-executable instructions, which are used to execute any of the above methods when the instructions are executed.

[0095] According to an embodiment of the present application, the present application also provides an electronic device and a readable storage medium.

[0096] Figure 5 A schematic block diagram of an example electronic device that can be used to implement embodiments of the present application is shown.

[0097] Electronic devices are intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or claimed herein.

[0098] like Figure 5 As shown, the device 500 includes a computing unit 501, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 502 or a computer program loaded from a storage unit 508 into a random access memory (RAM) 503. In the RAM 503, various programs and data required for the operation of the device 500 can also be stored. The computing unit 501, the ROM 502, and the RAM 503 are connected to each other via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0099] A number of components in the device 500 are connected to the I / O interface 505, including: an input unit 506, such as a keyboard, a mouse, etc.; an output unit 507, such as various types of displays, speakers, etc.; a storage unit 508, such as a disk, an optical disk, etc.; and a communication unit 509, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 509 allows the device 500 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0100] The computing unit 501 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 501 performs the various methods and processes described above, such as an entity recognition method for medical inspection texts. For example, in some embodiments, an entity recognition method for medical inspection texts may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 508. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 500 via the ROM 502 and / or the communication unit 509. When the computer program is loaded into the RAM 503 and executed by the computing unit 501, one or more steps of the entity recognition method for medical inspection texts described above may be performed. Alternatively, in other embodiments, the computing unit 501 may be configured in any other appropriate manner (eg, by means of firmware) to execute an entity recognition method for medical examination text.

[0101] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0102] The program code for implementing the method of the present application can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, implements the functions / operations specified in the flow chart and / or block diagram. The program code can be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0103] In the context of the present application, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0104] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0105] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0106] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0107] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution disclosed in this application can be achieved, and this document is not limited here.

[0108] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0109] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for entity recognition of medical examination text, characterized in that: The method comprises: Obtain medical examination texts in various expression formats; Performing word segmentation processing on the medical examination text using a professional dictionary corresponding to the medical examination text to obtain text word segmentation information; Performing component extraction on the word segmentation information to obtain component word segmentation information; Performing strategy matching on the component word segmentation information according to the automaton matching strategy to obtain an entity matching sequence; The step of performing strategy matching on the component word segmentation information according to the automaton matching strategy to obtain the entity matching sequence includes: obtaining a first word segmentation of the component word segmentation information in a specified order; matching the first word segmentation with the start state of multiple automaton matching strategies to determine that the first word segmentation satisfies at least one first matching strategy of the start state; obtaining a second word segmentation adjacent to the first word segmentation, matching the second word segmentation with the transition state of the first matching strategy to determine that the second word segmentation satisfies the second matching strategy of the transition state; judging whether the second word segmentation satisfies the stop state of the second matching strategy; and if it is judged that the second word segmentation satisfies the stop state of the first matching strategy, determining the first word segmentation and the second word segmentation as the entity matching sequence.

2. The method according to claim 1, characterized in that: The expression format includes at least one of the following formats: Chinese format, English format, numerical format, and symbol format; The professional dictionary contains at least one of the following word classes: core words, unit words, trend words, relational words, trigger words, symbolic words, modified descriptive words, quantitative descriptive words, qualitative descriptive words and useless words.

3. The method according to claim 1, characterized in that The step of performing word segmentation processing on the medical examination text by using a professional dictionary corresponding to the medical examination text to obtain text word segmentation information includes: The medical examination text is segmented based on the maximum length segmentation mode and the professional dictionary to obtain text segmentation information.

4. The method according to claim 1, characterized in that The method further comprises: If the first participle does not have a first matching strategy that satisfies the start state, obtaining a second participle adjacent to the first participle; According to matching the second word segmentation with the start state of the automaton matching strategy, it is determined that the second word segmentation satisfies a third matching strategy of the start state.

5. The method according to claim 1, characterized in that: The method further comprises: If it is determined that the second participle does not satisfy the stop state of the first matching strategy, obtaining a third participle adjacent to the second participle; According to the matching of the third participle with the transition state of the second matching strategy, it is determined that the third participle satisfies a fourth matching strategy of the transition state.

6. The method according to claim 1, characterized in that The method further comprises: Determine the corresponding segmentation strategy and data filtering strategy according to the automaton matching strategy; Segment the entity matching sequence according to the segmentation strategy to obtain an entity segmentation sequence; The entity segmentation sequence is subjected to data filtering according to the data filtering strategy to obtain an entity recognition sequence.

7. An entity recognition device for medical examination text, characterized in that: The device comprises: An acquisition module, used for acquiring medical examination texts containing multiple expression formats; A word segmentation module, used to perform word segmentation processing on the medical examination text through a professional dictionary corresponding to the medical examination text to obtain text word segmentation information; An extraction module, used to extract components from the word segmentation information to obtain component word segmentation information; A matching module, used for performing strategy matching on the component word segmentation information according to an automaton matching strategy to obtain an entity matching sequence; Wherein, the matching module includes: an acquisition submodule, which is used to acquire the first participle of the component participle information in a specified order; a matching submodule, which is used to match the first participle with the starting state of multiple automaton matching strategies, and determine that the first participle satisfies at least one first matching strategy of the starting state; the acquisition submodule is also used to acquire the second participle adjacent to the first participle, match the second participle with the transition state of the first matching strategy, and determine that the second participle satisfies the second matching strategy of the transition state; a judgment submodule, which is used to judge whether the second participle satisfies the stop state of the second matching strategy; and a determination submodule, which is used to judge that the second participle satisfies the stop state of the first matching strategy, and determine the first participle and the second participle as an entity matching sequence.

8. An electronic device, characterized in that: The device comprises: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images