A home disabled elderly safety care education content system construction method and system

Abstract

The present application relates to the technical field of health data processing for the elderly, in particular to a method and system for constructing a safety care education content system for homebound disabled elderly people. The method comprises: obtaining a first text of a questionnaire, extracting adverse events from the first text based on semantic features representing negative risks; retrieving associated literature with the adverse events as keywords; extracting evidence items from the associated literature, and extracting care topics based on semantic features representing care targets or intentions in the associated literature; constructing a data structure with chapter topics of a preset chapter template as keys and a set of multiple placeholder tags as values; replacing chapter topics with care topics based on the similarity between the chapter topics and the care topics; replacing tags with evidence items based on the similarity between the evidence items and the tags; and constructing a content system after replacing the tags in the chapter template with the evidence items in the data structure, thereby improving the efficiency of constructing the care education content system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of health data processing for the elderly, and in particular to a method and system for constructing a safety care education content system for homebound disabled elderly people. BACKGROUND

[0002] Currently, in the care process for homebound disabled elderly people, the caregivers are usually family members, who mainly rely on personal experience and lack detailed operation specifications for different care scenarios and disability levels, which can easily lead to redundancy of unnecessary care actions or omission of necessary care actions, resulting in potential risks in the care process.

[0003] In the prior art, in the medical field, a preliminary index and consultation table of an index system are prepared through literature review, semi-structured interview method and expert symposium, and after Delphi expert consultation is conducted on experts in the relevant research field, a professional nursing demand index system for disabled elderly people is obtained. For example, a method for constructing a sports exercise behavior evaluation index system during pregnancy is disclosed in Chinese Patent Application No. CN202310878490.4, which obtains evaluation selection indexes by combining the aforementioned Delphi expert consultation, determines the weights of the evaluation selection indexes by using the analytic hierarchy process, and then obtains an evaluation index system by verifying the evaluation selection indexes through a pre-trained model.

[0004] However, the construction method of the above-mentioned system is directly applied to construct a safety care education content system for disabled elderly people. After questionnaire survey and literature retrieval, a large amount of text data will be obtained. In the process of analyzing and summarizing key evidence, a large amount of semantic logic judgment depending on human subjective experience is involved, which not only consumes time and effort, but also is difficult to maintain the stability of semantic understanding for similar texts in the summary, resulting in low consistency of the collated text results and difficulty in efficiently constructing a safety care education content system.

[0005] Therefore, the present application provides a method and system for constructing a safety care education content system for homebound disabled elderly people, which improves the efficiency of standardized construction of the care education content system, improves the compatibility of the standardized system with individual actual care situations, and helps to improve the overall care level of homebound disabled elderly people. SUMMARY

[0006] The application aims to solve the technical problems of long time and low consistency in sorting and summarizing key evidence by researchers in the process of constructing a theoretical system, and provides a homebound disabled elderly safety care education content system construction method and system, which extracts key evidence items from literature and questionnaires by introducing a pre-trained semantic model, improving the efficiency of processing questionnaires and literature; after constructing the content system based on the evidence items, the content system is made into media works such as videos, the complete care process is reproduced with the help of media works, and the evaluation results of the media works are used as the optimization basis of the theoretical system, which improves the efficiency of feedback on the content system and shortens the time of constructing the theoretical system.

[0007] In order to achieve the above-mentioned application purposes, the application provides the following technical solutions:

[0008] A homebound disabled elderly safety care education content system construction method comprises the following steps:

[0009] A first text of a questionnaire is obtained, and adverse events are extracted from the first text based on semantic features representing negative risks by a pre-trained model; associated literature is retrieved by taking the adverse events as keywords; evidence items are extracted from the associated literature, and care themes are extracted based on semantic features representing care targets or intentions in the associated literature;

[0010] A data structure is constructed by taking chapter themes of a preset chapter template as keys and a set of multiple labels for placeholders as values; cosine similarity of the care themes and the chapter themes is calculated, and if the cosine similarity is greater than a theme threshold value, the chapter themes in the data structure are replaced by the care themes; cosine similarity of the evidence items and the labels is calculated, and if the cosine similarity exceeds a label threshold value, the labels in the data structure are replaced by the evidence items;

[0011] After the labels in the chapter template are replaced by the evidence items in the data structure, the content system is constructed.

[0012] Preferably, the construction of the content system further comprises the following steps:

[0013] The chapter themes are taken as script themes of a script template, the labels are taken as shot texts of the script template, and the chapter template is converted into the script template; if the cosine similarity calculated from the shot texts and the evidence items is greater than a label threshold value, the shot texts are replaced by the evidence items, media works are made based on the replaced script template, the media works are scored, and if the score result exceeds a preset score threshold value, the construction of the content system is completed.

[0014] Preferably, the extracting the evidence item from the associated literature comprises:

[0015] The evidence item is extracted from the same associated literature based on semantic features of an affected subject, a care action execution object, and a result generation of the adverse event, and based on context features of a cause analysis, a treatment measure, and a space environment of the care object of the adverse event.

[0016] Preferably, after the content system is constructed, the method further comprises:

[0017] The evidence item is decomposed into a plurality of first units representing basic care actions through syntactic analysis and semantic role labeling;

[0018] Feature phrases representing dysfunction and / or disease are extracted from the associated literature, matched with a preset disability level semantic library, and a first level range is determined; the first level range is a recommended range of disability level of a care object to which the first unit is applied;

[0019] A second level range is determined based on a matching result of the first unit and a preset care skill semantic library; the second level range represents a recommended range of skill level of a care worker applying the first unit to the care object;

[0020] The first level range and the second level range are associated with the first unit.

[0021] Preferably, the method further comprises:

[0022] A first disability level of the care object is evaluated, and if the first level range contains the first disability level, the first unit associated with the first level range is obtained; a plurality of the first units are combined to form a new evidence item, the data structure is updated based on the new evidence item, and the content system is constructed based on the data structure.

[0023] Preferably, the method further comprises:

[0024] A first skill level of the care worker is evaluated, and if the second level range contains the first skill level, the first unit associated with the first skill level in the second level range is obtained; a plurality of the first units are combined to form a new evidence item, the data structure is updated based on the new evidence item, and the content system is constructed based on the data structure.

[0025] Preferably, the method further comprises:

[0026] evaluating a second disability level of the care target; evaluating a second skill level of the caregiver; obtaining the first unit whose first level range contains the second disability level and whose second level range contains the second skill level, combining a plurality of the first units to form a new evidence item, updating the data structure based on the new evidence item, and constructing the content system based on the data structure.

[0027] Preferably, the second disability level is set as a disability level expected to be reached by the caregiver.

[0028] Preferably, the retrieving of the associated literature based on the adverse event as a keyword comprises:

[0029] Retrieving the associated literature based on the adverse event as a keyword from top to bottom in the databases with decreasing literature summary degree based on a 6S evidence pyramid model.

[0030] The application provides a home safety and care education content system for disabled elderly people, which comprises:

[0031] The obtaining module is configured to obtain a first text of a questionnaire, and extract an adverse event from the first text based on semantic features representing negative risks by using a pre-trained model.

[0032] The retrieving module is configured to retrieve associated literature based on the adverse event as a keyword, and extract an evidence item from the associated literature and a care theme based on semantic features representing care targets or intentions in the associated literature.

[0033] The constructing module is configured to construct a data structure based on chapter themes of a preset chapter template as keys and a set of a plurality of labels for placeholders as values, calculate a cosine similarity between the care theme and the chapter theme, replace the chapter theme in the data structure with the care theme if the cosine similarity is greater than a theme threshold, calculate a cosine similarity between the evidence item and the label, replace the label in the data structure with the evidence item if the cosine similarity exceeds a label threshold, and construct the content system after replacing the label in the chapter template with the evidence item in the data structure.

[0034] Compared with the prior art, the application has the following beneficial effects:

[0035] 1. By pre-training model, the adverse events are extracted from the questionnaire text data according to the learned various semantic features, and the evidence items are extracted from the related literature, so that the extracted results reduce the deviation of human experience, and are beneficial to improve the accuracy of text extraction and the consistency of similar texts; the data structure of the associated care theme and the evidence item is constructed by the associated structure of the chapter theme and the placeholder label in the chapter template, so that the large number of extracted evidence items can be integrated into a hierarchical structure, which is beneficial to efficiently and quickly fill the chapter template, improve the construction efficiency of the content system, and is beneficial to quickly form a safety care education content system with high credibility. The content system is made into a media work by the application, the media work can reflect the real care content, the optimization and verification of the content system are completed by the feedback of the media work, the efficiency of the feedback optimization of the content system is improved, and the content system is quickly constructed.

[0036] 2. By decomposing the evidence items into basic care actions and associating them with the disability level of the disabled elderly and / or the skill level of the caregivers, the content system is beneficial to realize reuse to generate an individualized new content system, reduce the construction cost of the individualized content system, and improve the reuse efficiency of the content system. Moreover, different basic care actions are extracted according to the disability level and / or the skill level, the content system is formed for different use purposes of the disabled elderly and / or the caregivers, a self-adaptive care standard system is constructed, and the expansibility of the content system is improved. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 Structure schematic diagram of the home disabled elderly safety care education content system construction system;

[0038] Figure 2 Process schematic diagram for constructing the home disabled elderly safety care education content system based on theory retrieval;

[0039] Figure 3 Process schematic diagram for decomposing the evidence items into basic care skill units and associating them with the disability level and the skill level;

[0040] Figure 4 Process schematic diagram for constructing an individualized content system based on the disability level and the skill level. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical scheme and advantages of the embodiments of the application clearer, the technical scheme in the embodiments of the application will be described clearly and completely below with reference to the drawings. Obviously, the described embodiments are part of the embodiments of the application, rather than all the embodiments.

[0042] Therefore, the following detailed description of the embodiments of the application is not intended to limit the scope of the application as claimed, but merely represents some embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the application.

[0043] It should be noted that the embodiments in the present application and the features and technical solutions in the embodiments can be combined with each other without conflict.

[0044] It should be noted that: similar reference numbers and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0045] In the description of the present application, it should be noted that the terms "first", "second" and the like are only used to distinguish description, and cannot be understood as indicating or implying relative importance.

[0046] For example Figure 1 As shown in the structural schematic diagram of the home disabled elderly safety care education content system construction system, the construction system is used to execute the specific process of the home disabled elderly safety care education content system construction method. The construction system comprises:

[0047] The acquisition module is configured to acquire a first text of a questionnaire, extract an adverse event from the first text based on semantic features representing negative risks through a pre-trained model;

[0048] The retrieval module is configured to retrieve associated literature with the adverse event as a keyword, extract evidence items from the associated literature, and extract care topics based on semantic features representing care targets or intentions in the associated literature.

[0049] The construction module is configured to construct a data structure with a chapter topic of a preset chapter template as a key and a set of multiple placeholder labels as a value, calculate the cosine similarity of the care topics and the chapter topics, replace the chapter topics in the data structure with the care topics if the cosine similarity is greater than a topic threshold, calculate the cosine similarity of the evidence items and the labels, replace the labels in the data structure with the evidence items if the cosine similarity exceeds a label threshold, and construct the content system after replacing the labels in the chapter template with the evidence items in the data structure.

[0050] Further, the method further comprises: an optimization module, configured to convert the chapter template into a script template by taking the chapter theme as a script theme of the script template and taking the label as a shot text of the script template; if a cosine similarity calculated from the shot text and the evidence entry is greater than a label threshold, replacing the shot text with the evidence entry; and producing a media work based on the replaced script template, scoring the media work, and if a score result exceeds a preset score threshold, completing the construction of the content system.

[0051] For example Figure 2 As shown in FIG. 1, a flowchart of a method for constructing a safety care education content system for homebound disabled elderly people is shown. The construction method comprises:

[0052] Text data obtained from the survey questionnaire of the surveyed object by scanning is denoted as a first text, and adverse events are extracted from the first text based on semantic features representing negative risks by using a pre-trained model; associated literature containing the adverse events in the text content is retrieved from a database by taking the adverse events as keywords; and evidence entries are extracted from the associated literature.

[0053] In an embodiment, the surveyed object is a care object and / or a caregiver thereof, and the care object includes disabled elderly people of different disability levels. A researcher selects a preset number of surveyed objects by using a convenient sampling method to conduct a questionnaire survey. Based on a cross-sectional study method, adverse events occurring in the process of home care are investigated and counted by using a “home care adverse event questionnaire for disabled elderly people” designed by experts in a preset time period; a ratio of the number of elderly people with adverse events to the total number of surveyed elderly people is calculated to count the incidence of adverse events. Adverse events with an incidence exceeding a preset threshold are selected, and the home care adverse events and their influencing factors are analyzed from the perspectives of the care object and the caregiver, for example, for the adverse event of choking, whether the care object eats too fast and whether the caregiver grinds the food sufficiently.

[0054] Specifically, since the manual processing of the questionnaire relies on subjectivity, there are differences in the measurement scales of objective events, and in the process of extracting adverse events, adverse events with the same semantics are divided into different topics, or similar adverse events repeatedly appear under a topic, which affects the accuracy and efficiency of subsequent retrieval of related documents and calculation of semantic feature similarity. Therefore, in this embodiment, based on the semantic features representing the negative risks directly to the disabled elderly, such as falls, choking, pressure ulcers, adverse events are extracted from the first text by a pre-trained artificial intelligence model. In this embodiment, the pre-trained model preferably uses a BERT semantic model. By inputting a large amount of literature text, the adverse event word groups representing negative risks, the word groups representing disabled elderly and their related attributes (such as disability level word groups), the word groups representing care actions, the word groups representing care actions that produce adverse events, the word groups representing emergency treatment, and other content are labeled, and the BERT semantic model is trained to learn different types of semantic features and context features, thereby using the pre-trained artificial intelligence model to perform efficiently and maintain the learned semantic features and context features unchanged. The characteristics of the key evidence items in the questionnaire survey and the literature content are extracted, and the efficiency of sorting the questionnaire survey and the literature content is improved.

[0055] In other embodiments, the pre-trained model can also be a general artificial intelligence model. The questionnaire is input into the general artificial intelligence model to extract adverse events, such as inputting the questionnaire into the pre-trained artificial intelligence model deepseek, GPT, etc. to interact, taking the word group representing the semantic feature of the negative risk as a reference, writing a prompt word to interact with the general artificial intelligence model, and making it extract adverse events from the questionnaire.

[0056] Further, since the questionnaire usually selects a preset number of respondents to fill out, the number of respondents covered by the respondents is small, and the number of adverse events and their treatment measures related content summarized from the questionnaire is limited, it is difficult to fully reflect all risk events that occur in the care process of the care object. Therefore, preferably, after summarizing the adverse events, the researchers develop a retrieval strategy using adverse events as keywords to obtain related literature related to adverse events from medical-related databases, such as research papers, diagnostic cases, and nursing home risk event records. These related literature at least includes one of the causes of adverse events, treatment processes, preventive measures, and correct care steps, in order to provide reliable references for subsequent production of reference care videos, manuals, and other dissemination media.

[0057] Further, when retrieving from medical related databases, based on the 6S evidence pyramid model, from top to bottom, the databases are retrieved in order of decreasing literature summarization degree, and multiple retrieval results are combined, i.e., the associated literature is obtained. Specifically, for processing adverse events, the integrated medical decision is directly retrieved from the clinical decision support module in the medical information system, the recommended care plan is retrieved from the care guidelines published by the authoritative department or association, the care guidance is retrieved from the abstracts related to the care content in medical journals or nursing journals, the care-related research literature is retrieved from the medical paper database, the recommended opinion is retrieved from the abstracts in the medical paper database, and the experimental results and experimental summary are retrieved from the randomized controlled trial reports published by the medical research team. Retrieving associated literature from databases of different summarization degrees is beneficial to enrich the literature basis of associated adverse events and improve the content richness of the content system formed in response to adverse events.

[0058] In other embodiments, after obtaining the associated literature, a Meta analysis method is used to synthesize all the associated literature for quality evaluation, and existing quality evaluation tools, such as the Newcastle-Ottawa Scale, are used to evaluate the retrieved associated literature. The consistency or conflict of multiple associated literature is processed by merging similar recommendations, tracing conflicts, or preferentially adopting high-quality literature evidence, and redundant information is removed.

[0059] Further, the associated literature is input into a pre-trained BERT semantic model or other pre-trained artificial intelligence model to extract the care theme and its associated evidence items. From the associated literature, based on the semantic features representing the care process goals or intentions, a care theme associated with the adverse event is extracted, such as the feeding process belonging to the nutrition support theme. From the same associated literature, based on the semantic features of the affected object, the care action execution object, and the result of the adverse event, and based on the cause analysis, treatment measures, and context features of the space environment of the care object, evidence items are extracted. Correspondingly, in this embodiment, the types of evidence items include incorrect care actions that cause adverse events, causes of adverse events, emergency treatment measures, preventive measures, and reference care actions.

[0060] Further, a data structure is constructed with the chapter theme of the preset chapter template as the key and the set of multiple placeholder labels in the chapter template as the value, and the care theme and evidence items in the same associated literature are associated through the data structure.

[0061] Specifically, before extracting adverse events from the questionnaire, a chapter template is created by the researcher, and the chapter template is input as a Word text document for storage. The chapter template is written by the researcher based on the standard guide for safe care of homebound disabled elderly, obtains text content of the title, care action, emergency treatment measure, and other related adverse events and care content, extracts semantic features representing the theme of the text content as tags from the text content for replacing and placeholders of the corresponding text content; other content used to link the complete text is retained, for example, the chapter template of the chapter theme "nutritional support" includes part of the content "in the process of {care action}, {adverse events} are prone to occur, the following measures can effectively reduce the risk: {emergency treatment measures}; {prevention measures}", wherein {care action}, {adverse events}, {emergency treatment measures}, and {prevention measures} are tags for placeholders.

[0062] The evidence items extracted from the related literature, journals, and other related literature become discrete text, losing the relevance to the theme of the same related literature and the context relationship. Moreover, due to the large amount of text data, multiple threads are usually used to extract evidence items, and evidence items of multiple literatures may be output at the same time. When the evidence items are directly used to construct the content system, the context order is disturbed, for example, emergency treatment measures should be after the care action that produces adverse events. At the same time, it should be noted that for the same adverse event, there may be multiple care actions that produce adverse events, for example, pipe slipping may occur in care actions containing body position change, including turning over, getting up, and bathing. Therefore, in this embodiment, by means of the concept of key-value pair, a key-value pair data structure is pre-set, which takes the content structure of the chapter template as a reference, takes the chapter theme as the key, takes the set formed by multiple tags as the value, and takes the order of the tags appearing in the chapter template as the index value order of the set. When extracting evidence items from the related literature, the tags in the data structure are replaced with evidence items with similar semantics according to the semantic similarity size of the evidence items and the tags in the data structure, realizing the association of care themes and evidence items in the same related literature.

[0063] Specifically, the care theme and the chapter theme in the data structure are converted into vector representations by a word vector embedding model, the cosine similarity of the care theme vector and the chapter theme vector is calculated, if the cosine similarity exceeds a preset theme threshold, it indicates that the care theme has a corresponding data structure, and the chapter theme in the data structure is replaced by the care theme. A plurality of evidence items and a plurality of labels in the data structure are converted into vector representations by a word vector embedding model, the cosine similarity of the evidence item vector and the label vector is calculated, if the cosine similarity exceeds a preset label threshold, it indicates that the evidence item has a corresponding label in the data structure, and the label in the set of data structures is replaced by the evidence item; the care theme and the evidence item in the same associated literature are associated through the data structure.

[0064] After extracting all the evidence items from the associated literature, the obtained plurality of data structures are used to fill in the corresponding chapter templates, the evidence items in the set of plurality of data structures are sorted according to the index value, and the labels in the plurality of corresponding chapter templates are replaced in turn, and the plurality of chapter templates after replacement are combined to construct the content system.

[0065] Further, the content system is evaluated by using the Delphi expert inquiry method to verify the reliability and effectiveness of the content system. 15-20 experts are invited to participate in the inquiry, and an expert inquiry questionnaire is prepared and distributed to conduct expert inquiry on the theoretical system. The inquiry questionnaire includes a letter to the expert (including research background, research purpose, pre-research basis, questionnaire filling method and matters needing attention); expert information table (including demographic information, expert familiarity with questionnaire content and judgment basis self-evaluation table); expert consultation table (including care theme, evidence item, importance score, modification suggestion). The questionnaire is distributed and collected by WeChat and email, and the filling time of each round is 1 week, and the interval between two rounds of inquiry is 2 weeks. The collected inquiry results are input into an EXCEL table, and the expert inquiry results are statistically analyzed by SPSS 22.0 software to obtain the evaluation results. Specifically, the expert activity coefficient is represented by the questionnaire return rate; the authority of expert opinion is represented by the authority coefficient; the coordination degree of expert opinion is represented by the full score ratio, Kendall harmony coefficient and variation coefficient. If the evaluation results of the content system meet the preset index screening standard, media works are made based on the content system. In an embodiment, the index screening standard is set to an importance score mean greater than 4.0; a full score ratio greater than 0.3; and a variation coefficient less than or equal to 0.2.

[0066] Further, the chapter theme is taken as the script theme of the preset script template, the label is taken as the shot text of the script template, and the chapter template is converted into a script template; if the cosine similarity calculated for the shot text and the evidence item is greater than the label threshold, the shot text is replaced with the evidence item, a media work is produced based on the replaced script template, the media work is scored, and if the score result exceeds the preset score threshold, the construction of the content system is completed.

[0067] Specifically, before extracting adverse events from the questionnaire, a script template is created by a media producer and saved as a Word text document. The script template includes multiple scripts, each of which includes but is not limited to scene description, care props, shot type, duration, and shot text. The shot text is used to represent the picture description and / or commentary script for a group of continuous care sub-actions, which corresponds to the label in the chapter template used to represent the adverse event, care action, emergency treatment measure, and other associated adverse event and care content placeholders. The time sequence of the script corresponds to the order of the labels in the chapter template.

[0068] Specifically, after the content system is evaluated by the Delphi expert consultation method, evidence items in the data structure are used to fill the preset script template to produce a media work. When filling the script template, the care theme and the script theme in the data structure are converted into vector representations by a word vector embedding model, and the cosine similarity of the care theme vector and the script theme vector is calculated. If the cosine similarity exceeds the theme threshold, it indicates that the data structure and the script template correspond to each other. Multiple evidence items in the data structure and multiple shot texts in the script template are converted into vector representations by a word vector embedding model, and the cosine similarity of the evidence item vector and the shot text vector is calculated. If the cosine similarity exceeds the label threshold, it indicates that the evidence item and the shot text correspond to each other, and the shot text in the script template is replaced with the evidence item. After the script template is filled, multiple scripts required by the media producer are obtained, and the scripts are used for design and shooting, or an AI artificial intelligence generation model is used to create a media work, forming a series of digital home safety care popular science media works, including videos, posters, audios, and promotional brochures.

[0069] Further, a plurality of subjects and experts are selected to evaluate the media work by means of questionnaire and option scoring, wherein the subjects include old people of different disability levels and their caregivers. After the subjects watch and / or listen to the media work, the subjects are invited to fill in the questionnaire. The questionnaire includes basic demographic survey (such as gender, age, education level), content evaluation (such as interest, understandability, practicality), design evaluation (such as layout, color matching, font and size, animation image, voice speed), overall evaluation and modification suggestions. The organization experts score the adaptability according to the preset evaluation standard of the popular science micro video, and finally the media work is debugged and optimized combined with the scoring results of the experts and the subjects. If the scoring result exceeds the preset scoring threshold, the final content system is obtained.

[0070] The adverse events are extracted from the questionnaire text data according to the learned multiple semantic features by the pre-trained model, and the evidence items are extracted from the associated literature, so that the extracted results reduce the deviation of human experience, and the accuracy of text extraction and the consistency of similar texts are improved; the data structure of the associated care theme and the evidence item is constructed according to the associated structure of the chapter theme and the placeholder label in the chapter template, so that the large number of extracted evidence items can be integrated into a hierarchical structure, which is beneficial to efficiently and quickly fill in the chapter template, improve the construction efficiency of the content system, and quickly form the content system with high credibility. The content system is made into a media work, and the media work can reflect the real care content. The optimization and verification of the content system are completed by the feedback of the media work, the efficiency of feedback optimization of the content system is improved, and the content system is quickly constructed.

[0071] In another embodiment, since the same adverse event can exist in the scene of caring for old people of different disability levels, but the emergency treatment measures and preventive measures of old people of different disability levels are different, the retrieved evidence items have the problem of being difficult to adapt to old people of different disability levels, and the same evidence item can also be applied to old people of multiple disability levels. At the same time, since the care process is affected by the disability level of the care object and the number and skill proficiency of the care skills possessed by the caregiver, the content system generated based on theoretical retrieval has the problem of insufficient retrieval and compatibility in actual application, such as the mismatch of the disability level corresponding to the evidence item, the difficulty or unskilled situation of the care action in the evidence item for the caregiver of the care object. Therefore, when the content system is applied, the care action, the different levels of the care object, and the skill level of the caregiver also need to be matched with each other.

[0072] After obtaining the content system, the application further comprises: grouping each evidence item according to the care theme, decomposing each evidence item into a plurality of first units representing minimum basic care actions, determining the disability level range and the skill level range of the first units, and matching appropriate first units according to the disability level of the care object and the care skill level of the care worker to form new evidence items, updating the data structure, constructing chapters of the content system, and dynamically adjusting the content system to help provide individualized care system and match the best care worker for the care object.

[0073] Further, for example Figure 3 As shown, the content of each evidence item is decomposed into a plurality of first units representing basic care actions through syntactic analysis and semantic role labeling;

[0074] Feature phrases representing functional disorders and / or diseases are extracted from the associated literature, matched with the preset disability level semantic library, and the first level range is determined; the first level range is used to represent the recommended range of the disability level of the care object applying the first unit;

[0075] The second level range is determined based on the matching result of the first unit and the care skill semantic library; the second level range is used to represent the recommended range of the skill level of the care worker applying the first unit to the care object;

[0076] The first level range and the second level range are associated with the first unit.

[0077] Specifically, the content of the evidence item is processed by an open-source natural language processing Python library and / or a pre-trained model, such as the spaCy library or a pre-trained Bert semantic model. Using a syntactic analysis function, such as the Dependency Parsing module in the spaCy library, the syntactic structure of the evidence item is analyzed, the part-of-speech relationship and sentence boundary in the evidence item sentence are identified, the words and / or phrases representing care actions, action purposes or action objects, and adverse event affected parts are extracted, and the first units are obtained in the form of "care action + action purpose or action object" by combining them in the order of the words. For example, an evidence item "the care worker should first check the food temperature, then slowly tilt the old man's head during feeding, and observe the old man's throat swallowing state to prevent choking", can be decomposed into four first units "check temperature, feed, tilt head, and observe swallowing". The plurality of first units are sequentially added to the array according to the order of extracting the care action and saved.

[0078] Further, semantic role labeling is performed by using spaCy-SRL, an extension library of spaCy, to extract words and / or phrases representing the role relationship between the action and the object in the evidence item content, to obtain the implementer and the implementee of the care action, and to form the associated role relationship, such as "caregiver, care recipient". The implementee is the care recipient, and the implementer varies according to the different disability levels of the care recipient, for example, a mild care recipient can be the care recipient himself, and the implementer of a severe care recipient is the caregiver. Taking the first unit as the key and the role relationship as the value, the formed key-value pairs are sequentially added to the array according to the order of extracting the care action, and the implementer and the implementee are associated with the first unit to represent the participable objects of the first unit.

[0079] When extracting the disability level, first, the phrase representing the care recipient ("old people" or "disabled old people") is obtained from the associated literature based on the adverb part of speech, and the phrase representing the degree of disability, such as "severe" or "level one", is obtained. The adverb phrase is directly added to the array corresponding to the first level range. If the adverb phrase fails to be obtained, the phrase representing the functional impairment and / or disease in the associated literature is extracted by the semantic role labeling function of the spaCy-SRL library. After converting the phrase and the disability feature description words in the disability level semantic library to vectors, the cosine similarity is calculated. If there is a disability feature description word with a cosine similarity exceeding a preset second threshold, the disability level of the disability feature description word is added to the array associated with the first unit and corresponding to the first level range. The disability level semantic library obtains the disability feature description words of the care recipient, including the functional impairment description, self-care ability description, and / or disease description, from the long-term care disability level evaluation standard, saves the disability feature description words and the corresponding disability level as a database record, and takes the vector of the self-care ability description and / or disease description after encoding as the primary key of the record table.

[0080] The set of care skills required by different caregivers is obtained from the national professional standards of health care workers (long-term care workers) and other professional certification standards, and the care skills and the corresponding skill levels are saved as a database record, and the vector of the care skill after encoding is taken as the primary key of the record table, to obtain the care skill semantic library. The first unit is encoded into a vector, and the cosine similarity between the vector of the first unit and the vectors of the care skills in the care skill semantic library is calculated. If there is a care skill with a cosine similarity exceeding a preset first threshold, the skill level of the care skill is added to the array, and after deduplication, the second level range associated with the first unit is obtained.

[0081] The first unit is taken as a key, and a set formed by the first level range and the second level range is taken as a value, to construct a key-value pair data structure, to realize the association of the first unit with the first level range, and the association of the first unit with the second level range. In another embodiment, an association data table can also be created, the vector after coding the first unit is taken as a primary key, and the first unit, the first level range and the second level range are taken as a database record for storage.

[0082] Further, the first disability level of the current care object is evaluated, if the first level range contains the first disability level, the first unit associated with the first level range is obtained, a plurality of first units are combined to form a new evidence item, the data structure is updated based on the new evidence item, and a new content system is constructed based on the updated data structure.

[0083] Specifically, the first disability level of the current care object is evaluated by the old-age disability assessor who obtains the professional skill level certificate, through a questionnaire designed according to the long-term care insurance disability level evaluation standard, based on the self-care ability of the care object in daily life. The first disability level and the disability level in the first level range are matched in turn, if the matching is successful, it indicates that the current care object can use the first unit associated with the first level range for care. In this way, the basic care action obtained based on the disability level of the care object constructs the individualized content system, and through the dynamic combination and adjustment of the basic care action with different disability levels, the best matching of the care action and the disability level is realized, which provides accurate care action reference standard for the old people with different disability levels, and is beneficial to improve the care satisfaction of the care object.

[0084] Further, the first skill level of the care giver is evaluated based on the skill record and the care time, if the second level range contains the first skill level, the first unit associated with the first skill level in the second level range is obtained; a plurality of first units are combined to obtain the content system of the care action that the care giver can perform.

[0085] Specifically, text data such as "complete primary feeding training" is collected from the training records, certificates and practice journals of the caregivers, and a plurality of entities related to the care skills are extracted by using a named entity recognition function, such as the Named Entity Recognition module in the spaCy library. The related entities are converted into entity vectors, the cosine similarity between the entity vectors and the care skill vectors in the care skill semantic library is calculated, the skill level corresponding to the care skill vector with a cosine similarity exceeding a first threshold is obtained, and the skill level with the largest number is determined as the first skill level of the caregiver. The first skill level and the skill levels in the second level range are matched in turn, and if the matching is successful, it indicates that the caregiver has the ability to implement the basic care actions associated with the second level range. In this way, the individualized content system constructed by the basic care actions obtained from the skill level of the caregiver is dynamically combined and adjusted with different skill levels, which realizes the best matching of care actions and caregiver skill levels, helps to quickly match the skill set already possessed by the caregiver, and provides the content system that can be used for evaluation when selecting a caregiver for the disabled elderly.

[0086] Further, for example Figure 4 As shown, the second disability level of the care object is evaluated; the second skill level of the caregiver is evaluated; the first unit in which the first level range contains the second disability level and the second level range contains the second skill level is obtained, a plurality of first units are combined to form a new evidence item, the data structure is updated based on the new evidence item, and the content system is constructed based on the data structure.

[0087] Specifically, the evaluation method of the second disability level or the second skill level is consistent with the evaluation method of the first disability level or the first skill level, which has been described in the foregoing text, and will not be repeated in this embodiment. The content system formed based on the basic care actions in this embodiment can be used for hospitals or nursing homes to assign suitable caregivers for the disabled elderly, which can not only ensure the adaptability of the care actions and the skill of the caregivers, reduce the care risks caused by the low skill proficiency of the caregivers, but also ensure the adaptability of the care actions and the disability level of the disabled elderly, improve the effectiveness of the care actions, reduce the risks brought by insufficient care or excessive care to the disabled elderly, and improve the precision of implementing the content system.

[0088] In another embodiment, the second disability level can also be set by the administrator or the caregiver, for indicating the disability level that the caregiver is expected to be able to reach. The first level range containing the second skill level and the second level range containing the second disability level are obtained, and a plurality of first units are obtained, to obtain an expected skill set; a difference set between the expected skill set and the current skill set of the caregiver is calculated, to obtain a care action that the caregiver is expected to learn, to form a care action learning path, to facilitate the construction of an individualized content system for learning according to the needs of the skill level of the caregiver, and to help the caregiver determine which advanced care skills he or she needs to master if he or she needs to take care of an elderly person with a higher disability level, so as to formulate a clear and phased career growth goal.

[0089] By decomposing the evidence items into basic care actions and associating them with the disability level of the care recipient and / or the skill level of the caregiver, the content system can be reused to generate an individualized new content system, the construction cost of the individualized content system can be reduced, and the reuse efficiency of the content system can be improved. Moreover, different basic care actions are extracted according to the disability level and / or the skill level, the content system is formed for different use purposes for the care recipient and / or the caregiver, the self-adaptive care standard system is constructed, and the expansibility of the content system is improved.

[0090] The above embodiments are only used to illustrate the present application and not to limit the technical solutions described in the present application. Although the present application has been described in detail with reference to the above embodiments, the present application is not limited to the above specific embodiments. Therefore, any modification or equivalent replacement of the present application; all technical solutions and improvements that do not deviate from the spirit and scope of the present application are covered in the scope of the claims of the present application.

Claims

1. A method for constructing a safety care education content system for disabled elderly people living at home, characterized in that, include: Obtain the first text of the survey questionnaire, and extract adverse events from the first text based on semantic features representing negative risks using a pre-trained model; Search for related documents using the aforementioned adverse events as keywords; Evidence entries are extracted from the associated documents, and care topics are extracted based on the semantic features of the associated documents that indicate care goals or intentions. The data structure is constructed using the chapter theme of the preset chapter template as the key and a set of placeholder tags as the value. Calculate the cosine similarity between the care topic and the chapter topic. If the cosine similarity is greater than the topic threshold, replace the chapter topic in the data structure with the care topic. Calculate the cosine similarity between the evidence item and the label. If the cosine similarity exceeds the label threshold, replace the label in the data structure with the evidence item. After replacing the tags in the chapter template with the evidence entries in the data structure, the evidence entries are decomposed into multiple first units representing basic care actions through syntactic parsing and semantic role labeling; Feature phrases representing functional impairment and / or disease are extracted from the relevant documents and matched with a preset disability level semantic database to determine the first level range; the first level range is the recommended range of disability levels for the care recipients of the first unit. The second level range is determined based on the matching results between the first unit and the preset care skills semantic database; the second level range represents the recommended range of skill levels for caregivers to apply the first unit to the care recipient. Associate the first level range and the second level range with the first unit; Assess the first disability level of the care recipient; if the first level range includes the first disability level, obtain the first unit associated with the first level range; combine multiple first units to form a new evidence entry; update the data structure based on the new evidence entry; and construct the content system based on the data structure.

2. The method for constructing a home-based safety care education content system for disabled elderly as described in claim 1, characterized in that, The construction of the content system also includes: Using the chapter theme as the script theme of the script template and the tag as the shot text of the script template, the chapter template is converted into the script template; if the cosine similarity calculated between the shot text and the evidence item is greater than the tag threshold, the shot text is replaced with the evidence item, a media work is produced based on the replaced script template, the media work is scored, and if the score exceeds the preset score threshold, the construction of the content system is completed.

3. The method for constructing a home-based safety care education content system for disabled elderly as described in claim 1, characterized in that, The extraction of evidence entries from the related documents includes: Evidence entries are extracted from the same related documents based on the semantic features of the affected objects, care recipients, and results of the adverse events, as well as the contextual features of the cause analysis, treatment measures, and spatial environment of the care recipients.

4. The method for constructing a home-based safety care education content system for disabled elderly as described in claim 1, characterized in that, Also includes: Assess the caregiver's first skill level; if the second level range includes the first skill level, obtain the first unit associated with the first skill level in the second level range; combine multiple first units to form a new evidence entry; update the data structure based on the new evidence entry; and construct the content system based on the data structure.

5. The method for constructing a home-based safety care education content system for disabled elderly as described in claim 1, characterized in that, Also includes: Assess the second disability level of the person under care; assess the second skill level of the caregiver; Obtain the first unit whose first level range includes the second disability level and whose second level range includes the second skill level; combine multiple first units to form a new evidence entry; update the data structure based on the new evidence entry; and construct the content system based on the data structure.

6. The method for constructing a home-based safety care education content system for disabled elderly as described in claim 5, characterized in that, The second disability level is set as the disability level that the caregiver is expected to reach.

7. A method for constructing a home-based safety care education content system for disabled elderly as described in any one of claims 1-4, characterized in that, The process of retrieving related documents using the adverse event as a keyword includes: Using the aforementioned adverse events as keywords, and based on the 6S evidence-based resource pyramid model, a search was conducted from top to bottom in databases with progressively decreasing levels of literature summarization to obtain relevant literature.

8. A system for constructing a home-based safety care education content system for disabled elderly people, characterized in that, include: The acquisition module is used to acquire the first text of the questionnaire and extract adverse events from the first text based on semantic features representing negative risks using a pre-trained model; The retrieval module is used to retrieve related documents using the adverse events as keywords; Evidence entries are extracted from the associated documents, and care topics are extracted based on the semantic features of the associated documents that indicate care goals or intentions. The building module is used to construct a data structure with the chapter theme of the preset chapter template as the key and a set of multiple placeholder tags as the value; Calculate the cosine similarity between the care topic and the chapter topic. If the cosine similarity is greater than the topic threshold, replace the chapter topic in the data structure with the care topic. Calculate the cosine similarity between the evidence item and the label. If the cosine similarity exceeds the label threshold, replace the label in the data structure with the evidence item. After replacing the label in the chapter template with the evidence item in the data structure, decompose the evidence item into multiple first units representing basic care actions through syntactic parsing and semantic role labeling. Feature phrases representing functional impairment and / or disease are extracted from the relevant documents and matched with a preset disability level semantic database to determine the first level range; the first level range is the recommended range of disability levels for the care recipients of the first unit. The second level range is determined based on the matching results between the first unit and the preset care skills semantic database; the second level range represents the recommended range of skill levels for caregivers to apply the first unit to the care recipient. Associate the first level range and the second level range with the first unit; Assess the first disability level of the care recipient; if the first level range includes the first disability level, obtain the first unit associated with the first level range; combine multiple first units to form a new evidence entry; update the data structure based on the new evidence entry; and construct the content system based on the data structure.

Images