A gastrointestinal tumor patient symptom self-evaluation data intelligent acquisition and dynamic management method

Abstract

The application relates to the technical field of natural language data processing, and discloses a kind of gastrointestinal tumor patient symptom self-evaluation data intelligent acquisition and dynamic management method, collect this round self-evaluation text and collection time, form semantic unit by standard, split and context relationship reservation;Extract symptom entity, degree and time information, form symptom vector, degree sequence value, alignment period and relationship vector;Restore previous version record, generate entity similarity, relationship consistency and time consistency, determine semantic anchor by weightless geometric convergence, correlation threshold and bidirectional unique maximum condition;According to semantic anchor, write new item and change field, reconstruct dynamic symptom table composed of symptom name, degree level and time interval. The application separates the processing of symptom identity association and state update, reduces the record misdisassembly, mismerger and improper coverage in continuous self-evaluation, and improves version consistency and traceability.

Description

Technical Field

[0001] This application relates to the field of natural language data processing technology, specifically to a method for intelligent collection and dynamic management of symptom self-assessment data of patients with gastrointestinal tumors. Background Technology

[0002] Patients with gastrointestinal tumors often submit symptom information such as abdominal pain, nausea, vomiting, abdominal distension, changes in appetite, and abnormal bowel movements through paper questionnaires, mobile devices, or web pages during diagnosis, treatment, and follow-up. Existing symptom self-assessment systems typically perform word segmentation, entity recognition, and attribute extraction on a single text sample, writing the symptom name, severity, and time into structured fields, and saving records according to the collection round. Some systems also use text similarity, field comparison, or version control methods to identify newly added and modified content, in order to generate symptom data arranged by time.

[0003] Continuous self-report texts are conversational and sequential; the same symptom may omit its name, use pronouns, abbreviations, or synonyms in different rounds, and the relative time may also change with the time of collection. At the same time, the degree and duration themselves may change. When using a method of independently extracting and comparing texts or fields by round, changes in expression and state are easily treated as the same type of difference. Degree and time modifiers in compound sentences may also be mismatched during the structuring process, resulting in duplicate records of the same symptom, merging of different symptoms, or improper overwriting of original records. It is difficult to maintain a stable correspondence and traceable version chain of symptom records in continuous self-report data. Summary of the Invention

[0004] This application provides a method for intelligent collection and dynamic management of symptom self-assessment data of patients with gastrointestinal tumors, in order to solve the technical problems of reducing the mis-splitting, mis-merging and link breakage of symptom records in the continuous self-assessment text updates of patients with gastrointestinal tumors.

[0005] The first aspect of this application provides a method for intelligent collection and dynamic management of symptom self-assessment data of patients with gastrointestinal tumors, including:

[0006] Step S1: Collect the self-evaluation text and collection time of this round, standardize and segment the self-evaluation text, and form semantic units that retain contextual relationships;

[0007] Step S2: Extract symptom entities, degree words, negation words, disappearance words, time expressions and contextual relationships from semantic units to form symptom vectors, degree order values, relative time periods and relationship vectors. Combine these with the collection time to form aligned time periods and compose symptom data items.

[0008] Step S3: Recover the symptom data items from the previous version record, form candidate pairs with the symptom data items in the current round, generate entity similarity from the symptom vector, and do not introduce degree order values;

[0009] Step S4: Generate relation consistency quantity and time consistency quantity according to relation vector and alignment time period respectively, and geometrically converge them with entity similarity quantity without weight to form semantic association quantity. Generate semantic anchor mark according to association threshold and bidirectional unique maximum condition.

[0010] Step S5: In this step, candidate pairs marked as true with semantic anchors are included in the symptom semantic chain, the degree order value and the alignment time period are compared, and only the change field is written; the current round of symptom data items without semantic anchors are added as new items to form the current round version record.

[0011] Step S6: Recover the symptom name, severity level, and time interval from the current version record to form a dynamic symptom table.

[0012] The second aspect of this application provides an information processing and storage system, including a processor and a memory. The memory is used to store computer programs and version records. When the computer program is executed by the processor, it causes the processor to perform the following operations:

[0013] Collect the self-evaluation text and collection time of this round, standardize and segment the self-evaluation text, and form semantic units that retain contextual relationships;

[0014] Extract symptom entities, degree words, negation words, disappearance words, time expressions, and contextual relationships from semantic units to form symptom vectors, degree order values, relative time periods, and relationship vectors. Combine these with the collection time to form aligned time periods and compose symptom data items.

[0015] The symptom data items from the previous version are restored and paired with the symptom data items in the current round to form candidate pairs. Entity similarity is generated from the symptom vectors without introducing degree order values.

[0016] Based on the relation vector and the alignment time period, relation consistency quantity and time consistency quantity are generated respectively. They are geometrically converged with entity similarity quantity without weight into semantic association quantity. Semantic anchor mark is generated based on association threshold and bidirectional unique maximum condition.

[0017] Candidate pairs marked as true with semantic anchors are included in the symptom semantic chain, and the degree order value and alignment time period are compared. Only the change field is written. The current round of symptom data items without semantic anchors are added as new items to form the current round version record.

[0018] The symptom names, severity levels, and time intervals are recovered from the symptom semantic chain in this version to form a dynamic symptom table.

[0019] The third aspect of this application provides a computer-readable storage medium storing computer instructions for causing a computer to execute the above-described method for intelligent collection and dynamic management of self-assessment data of gastrointestinal tumor patients' symptoms.

[0020] The beneficial effects of this application are as follows: This application standardizes, segments, and preserves the context of continuous self-assessment texts, forming symptom identity information, severity status, and time information under a unified time coordinate. When associating across versions, it eliminates the interference of severity order values ​​on symptom identity judgment, and combines relational and time information to filter bidirectionally unique corresponding items. After association is established, only newly added items and changed fields are attached to the version record; original records not mentioned in the current round remain valid. Only when an expression is explicitly absent or has disappeared is the lowest severity level written. This processing ensures that expression rewriting, relative time migration, and actual status changes are processed accordingly, reducing duplicate linking of the same symptom, erroneous merging of different symptoms, and instances where unexpressed symptoms are treated as disappeared. This improves the consistency, traceability, and adaptability of the dynamic symptom table to continuous data collection scenarios. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0022] Figure 1 A flowchart illustrating an intelligent collection and dynamic management method for symptom self-assessment data of gastrointestinal tumor patients provided in this application embodiment.

[0023] Figure 2 This is a schematic diagram illustrating the scenario of symptom self-assessment data collection and dynamic management for patients with gastrointestinal tumors, as provided in an embodiment of this application. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0025] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The terms "first," "second," etc., in this application are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0026] It is understood that before using the technical solutions disclosed in the various embodiments of this application, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this application in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0027] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0028] like Figure 1 As shown, a method for intelligent collection and dynamic management of symptom self-assessment data in patients with gastrointestinal tumors includes:

[0029] Step S1: Collect the self-evaluation text and collection time of this round, standardize and segment the self-evaluation text, and form semantic units that retain contextual relationships;

[0030] Step S2: Extract symptom entities, degree words, negation words, disappearance words, time expressions and contextual relationships from semantic units to form symptom vectors, degree order values, relative time periods and relationship vectors. Combine these with the collection time to form aligned time periods and compose symptom data items.

[0031] Step S3: Recover the symptom data items from the previous version record, form candidate pairs with the symptom data items in the current round, generate entity similarity from the symptom vector, and do not introduce degree order values;

[0032] Step S4: Generate relation consistency quantity and time consistency quantity according to relation vector and alignment time period respectively, and geometrically converge them with entity similarity quantity without weight to form semantic association quantity. Generate semantic anchor mark according to association threshold and bidirectional unique maximum condition.

[0033] Step S5: In this step, candidate pairs marked as true with semantic anchors are included in the symptom semantic chain, the degree order value and the alignment time period are compared, and only the change field is written; the current round of symptom data items without semantic anchors are added as new items to form the current round version record.

[0034] Step S6: Recover the symptom name, severity level, and time interval from the current version record to form a dynamic symptom table.

[0035] In one embodiment of this application, step S1 includes the following steps:

[0036] Step S101: Collect the self-assessment text and collection time of the gastrointestinal tumor patients in this round. Perform character normalization, punctuation restoration and colloquial redundancy removal on the self-assessment text in sequence to maintain the original order of symptom words, degree words, negation words, time words, and the original sequence of words.

[0037] This round of self-assessment text refers to the natural language description of symptoms entered by the patient in a single submission. This self-assessment text can be obtained through keyboard input or generated by a speech recognition device before submission. The system saves the complete original text of this submission, without collecting blood pressure, heart rate, or imaging values.

[0038] The acquisition time refers to the timestamp at which the server fully receives the self-evaluation text for this round. The acquisition time is automatically generated by the server clock and stored using the same time zone and time unit to avoid the impact of terminal clock differences on time expression parsing.

[0039] Specifically, character normalization sequentially performs unified character encoding, conversion between full-width and half-width characters, merging of consecutive whitespace, standardization of repeated punctuation, and case normalization. Punctuation restoration first preserves the original sentence-ending punctuation, then adds commas or periods based on pause markers, parallel conjunctions, adversative conjunctions, and new symptom trigger words. Pause markers are the pause position information output by the speech recognition device along with the transcribed text. For keyboard-input text, or speech-transcribed text that does not contain pause position information, punctuation is not added based on pause markers.

[0040] Furthermore, redundant spoken language is cleaned up by removing modal particles and repeated words that do not convey meanings of symptoms, degree, negation, disappearance, or time. During the process, a mapping table of original text positions is established, ensuring that symptom words, degree words, negation words, time words, and their original order remain unchanged, while disappearing words are retained as state expressions.

[0041] Step S102: Determine the segmentation position of the text fragment based on the sentence-end punctuation, parallel conjunctions, adversative conjunctions, and symptom trigger words, and form text fragments according to the segmentation position;

[0042] A text segment refers to a continuous sequence of characters cut from the self-assessment text in this round according to semantic boundaries. Each text segment retains its start and end positions and order in the original text. Symptom trigger words are standardized symptom names, synonyms, or common abbreviations in the symptom glossary that can trigger symptom entity recognition.

[0043] Specifically, the segmentation position is determined in the order of sentence-end punctuation, adversative conjunctions, coordinating conjunctions, and symptom trigger words. Existing sentence-end punctuation forms a mandatory boundary. The content before and after an adversative conjunction forms separate text segments. A boundary is formed when symptom trigger words appear on both sides of a coordinating conjunction. If the current content already contains a symptom trigger word and a new symptom trigger word appears, a boundary is formed before the new symptom trigger word. No segmentation position is set between degree words or time words and their adjacent symptom words to prevent the modified content from deviating from the corresponding symptom. When multiple rules fall in the same position, only one segmentation position is retained.

[0044] Step S103: Preserve the contextual relationships between adjacent text segments, which are composed of continuation, reference, parallelism, contrast, degree modification and time modification, and form semantic units from the text segments and the aforementioned contextual relationships.

[0045] Contextual relationships represent the connections between adjacent text segments, including succession, synonymy, parallelism, contrast, degree modification, and temporal modification. A semantic unit is composed of a text segment, its original position, and its corresponding contextual relationship, used to maintain a consistent correspondence between text content and relational structure.

[0046] Specifically, contextual relationships are identified using conjunctions, pronouns, word order, and adjacency distance. Adjacency distance is determined based on character positions in the original text position mapping table. When a relevant expression precedes a symptom entity, the distance is calculated as the number of character positions between the end position of the relevant expression and the beginning position of the symptom entity; when a relevant expression follows a symptom entity, the distance is calculated as the number of character positions between the end position of the symptom entity and the beginning position of the relevant expression. Candidates with smaller character position counts are considered closer candidates.

[0047] For text fragments containing continuation expressions such as "still," "or," and "like this," but lacking a clear symptom entity, first search for a unique continuation symptom entity among the earlier adjacent text fragments in the current round. If no unique object is found, the text fragment is marked as a continuation relationship to be parsed. The referents of pronouns such as "its" and "this symptom" are determined in the same way. Parallel and adversative relationships are determined by their corresponding conjunctions. Degree and time modification relationships are respectively connected to the nearest symptom entity within the same text fragment that is not separated by an adversative boundary. If two equidistant candidates exist and cannot be excluded by the connection relationship, no modification relationship is established, and the original text is preserved for verification.

[0048] Furthermore, semantic units store text fragments, the start and end positions of the original text, the relationship type, the relationship direction, and the position of the associated text fragments in a record format. When a text fragment has multiple contextual relationships, they are stored separately without overwriting each other. This record structure allows the symptom entities and modifying content in compound self-evaluation statements to be restored according to the original textual relationships after segmentation.

[0049] In one embodiment of this application, step S2 includes the following steps:

[0050] Step S201: Perform symptom entity recognition, contextual continuation analysis, degree word parsing, and time expression parsing on the semantic unit to obtain symptom entities, degree words, negation words, disappearance words, and time expressions;

[0051] Symptom entities are symptom names identified from semantic units or symptom references determined through continuation analysis. Degree words are words describing the strength of a symptom. Negation words are words that explicitly indicate the absence of a symptom. Disappearance words are words that explicitly indicate the termination of an existing symptom. Time expressions are textual content indicating dates, times, durations, or relative time ranges. The degree word parsing in this embodiment includes the identification of degree words, complete negation expressions, and explicit disappearance expressions, as well as the determination of their scope.

[0052] Specifically, symptom entity recognition employs a symptom terminology list and a longest-matching rule. The symptom terminology list stores at least the standardized symptom names, synonyms, and common abbreviations, establishing a unified mapping for the same meaning. During recognition, longer terms are matched first, followed by shorter terms. If a shorter term is completely contained within an already recognized longer term, no duplicate symptom entity is generated. Degree words, negation words, disappearing words, and time expressions are identified using their corresponding terminologies, with their scope limited by text fragment boundaries and contextual relationships.

[0053] The target of a complete negation or explicitly disappeared expression is determined by the symptom entity within the same text segment that is closest to the expression and not separated by a transition boundary. When candidate symptom entities exist on both sides of the expression, the candidate symptom entity with the smaller character position distance in the original text position mapping table is selected as the target. If the distances are the same and ambiguity cannot be resolved based on context, no definitive action relationship is established, and the original semantic unit is preserved for verification. When a new symptom entity appears after a transition conjunction, the scope of action is redefined.

[0054] Furthermore, contextual analysis searches for candidate symptom entities in the order of the current semantic unit, earlier semantic units in this round, and the most recent valid self-assessment text. The most recent valid self-assessment text refers to the most recent self-assessment text that has been parsed and formed at least one symptom data item before the current collection round. A still valid symptom entity refers to a symptom entity that can be recovered from the previous version record and has a currently valid symptom vector.

[0055] For the succession relationships marked as to be parsed in step S103, first read the adjacent relationships saved in the current semantic unit, then read the symptom entities in the earlier semantic units in this round, and finally read the last appearing and still valid symptom entity in the most recent valid self-assessment text. If there is only one candidate object, that candidate object is taken as the succession result. If there are parallel symptoms and the succession expression does not provide distinguishable clues, no definite succession result is generated. Succession expressions without clear symptom entities do not form symptom data items, and the original semantic unit is still saved in the collection record.

[0056] Step S202: Encode the symptom entity and its synonyms, abbreviations and continuation context into a non-zero symptom vector, and map degree words, negation words and disappearance words to degree order values ​​according to the same ordered dictionary. The lowest level of degree order value corresponds to the absence or disappearance of the symptom.

[0057] A symptom vector is a fixed-order numerical representation of a symptom entity and its synonyms, abbreviations, and context, used to characterize symptom identity. A symptom vector has at least one non-zero component. The same ordered dictionary is a degree mapping table used consistently during deployment. The degree ordinal value is the ordinal value in the same ordered dictionary corresponding to the current degree expression.

[0058] The ordered dictionary preferably has five levels, where 0 corresponds to non-existence or disappearance, 1 corresponds to mild, 2 to moderate, 3 to severe, and 4 to very severe. Synonymous degree expressions in the dictionary are grouped into the same level. Complete negation expressions and explicitly disappeared expressions are preferentially mapped to 0. When no degree word, negation word, or disappeared word appears, the degree order value field retains an empty value, and the degree of symptoms is not presumed.

[0059] Specifically, fixed-dimensional sparse coding is used to form symptom vectors. Each canonical symptom name in the symptom glossary occupies a fixed dimension, and synonyms and abbreviations are mapped to the same dimension of the corresponding canonical symptom name. Symptom entities obtained from contextual analysis are written into this same dimension. The matching dimension is set to 1, and the other dimensions are set to 0. The glossary and dimension order are fixed in the same deployment environment, and the glossary version is saved with the version record. Each confirmed symptom entity has a canonical symptom name, keeping the symptom vector non-zero.

[0060] Furthermore, when mapping degree words, complete negation expressions and explicitly disappeared expressions are determined first, followed by the degree expression. Complete negation expressions such as "none," "absent," and "not appeared," as well as explicitly disappeared expressions such as "disappeared" and "no longer appears after relief," are mapped to the lowest level. Weakened expressions such as "not so much" and "slightly present" are treated as degree words and not as complete negations. When a symptom entity has multiple degree words, the degree word with the closest character position determined by the original text position mapping table and which does not cross the transition boundary is used. When conflicting degree expressions with the same distance appear within the same scope, the degree order value field is kept null, and the original text is retained in the collection record. When the degree order value is null, the previous valid degree order value is not overwritten in the incremental update. Newly established symptom semantic chains can be displayed as ungraded in the dynamic symptom table.

[0061] Step S203: The time expression is parsed into a binary start and end offset relative to the acquisition time to form a relative time period. The relationships of succession, same reference, parallel, transition, degree modification and time modification are encoded into a fixed-dimensional binary relationship vector.

[0062] The relative time period is the start and end offset of the time expression relative to the acquisition time. The relation vector is a fixed-dimensional binary representation of the context relationship.

[0063] Specifically, the time expression parsing uses a fixed time vocabulary and calendar conversion rules. "Today" is parsed from the start time of the collection date to the collection time. "Yesterday" is parsed from the start time of the day before the collection date to the end time. "The last 3 days" is parsed from 3 days before the collection time to the collection time. "3 days ago" is parsed from 3 days before the collection time to the collection time. Dates are explicitly converted to start and end times according to a unified time zone. When no time expression appears, both the start and end offsets are set to 0, indicating the observation point of this round of collection, and the symptom onset time is not inferred from this. The start and end offsets use the same time unit, preferably hours.

[0064] Furthermore, the relation vector can be set to 12 dimensions. The first six dimensions represent the continuation, synonymy, parallelism, contrast, degree modification, and time modification relationships to the preceding adjacent text segment, respectively. The last six dimensions represent the correspondence relationships to the following adjacent text segment in the same order. Dimensions with corresponding relationships are set to 1, and dimensions without corresponding relationships are set to 0. When the same semantic unit has multiple relationships, multiple dimensions are allowed to be 1 simultaneously. All semantic units use the same dimensions and the same order.

[0065] Step S204: Add the acquisition time to the two start and end components of the relative time period to form an aligned time period under a unified time coordinate.

[0066] Aligned time periods are the start and end times obtained after converting relative time periods to a unified time coordinate system.

[0067] Specifically, the calculation formula for the alignment period is as follows:

[0068]

[0069] In the formula, Indicates the first Round Alignment time period for each symptom data item Indicates the first Round The relative time period obtained by parsing semantic units Indicates the first The timing of collecting the self-evaluation text;

[0070] Specifically, the two components of the aligned time period represent the start and end times under a unified time coordinate system. The acquisition time and the relative time period use the same time unit, and the start and end components are converted separately. If the start time is later than the end time after conversion, the time expression parsing result is considered invalid, the original time expression is retained, and both the start and end times of the aligned time period for that symptom data item are recorded as the acquisition time.

[0071] Step S205: Symptom data items are composed of symptom vectors, severity order values, aligned time periods, and relation vectors, while retaining relative time periods;

[0072] Symptom data items are structured records stored in a fixed field order according to symptom vector, severity ordinal value, alignment time period, and relation vector.

[0073] Specifically, the structural expression of the symptom data item is as follows:

[0074]

[0075] In the formula, Indicates the first Round Each symptom data item, Indicates the first Round Symptom vector for each symptom data item, Indicates the first Round The severity ordinal value of each symptom data item. Indicates the first Round Alignment time period for each symptom data item Indicates the first Round The relationship vector of each symptom data item.

[0076] Furthermore, the fields of the symptom data items retain their original units and data types, and cross-field numerical addition is not performed. The relative time period is saved as an independent traceability field associated with the symptom data items to distinguish between the relative time expression in the original text and the converted time interval.

[0077] In one embodiment of this application, step S3 includes the following steps:

[0078] Step S301: Recover the symptom data items of each symptom semantic chain from the previous version record, and form candidate pairs by pairing the symptom data items of the current round with the symptom data items of the previous version.

[0079] The previous version record is a collection of incremental records attached and saved according to the collection round, including the inheritance relationship of newly added items, changed fields, and unchanged fields. A candidate pair is a pair consisting of a symptom data item from the current round and a symptom data item recovered from the previous version record.

[0080] Specifically, the system reads the previous version of records according to the collection round. When a new item is encountered, the current record for the corresponding symptom semantic chain is created. When a changed field is encountered, the corresponding old field in the symptom semantic chain is replaced with the new field value. Fields that have not changed continue to use the most recent valid value, thereby restoring the symptom data items for each symptom semantic chain. The restored result includes at least the symptom vector, degree order value, aligned time period, and relationship vector.

[0081] Subsequently, each symptom data item in this round is paired with each symptom data item in the previous version to form a complete set of candidate pairs. If there is no record in the previous version, no candidate pair is generated, and all symptom data items in this round are treated as new items in step S502.

[0082] Step S302: Extract two non-zero symptom vectors from the candidate pair and calculate the cosine similarity value of the two non-zero symptom vectors;

[0083] Cosine similarity value represents the degree of consistency in the directions of two non-zero symptom vectors.

[0084] Specifically, the two symptom vectors in a candidate pair use the same terminology version and the same dimension. During terminology upgrades, the old symptom vectors are first converted to the current dimension according to the standardized symptom name mapping relationship between the old and new terms, and then the cosine similarity value is calculated. If neither side yields a non-zero symptom vector, the candidate pair is not included in the entity similarity calculation, and the original record is retained for verification.

[0085] Step S303: Add 1 to the cosine similarity value and divide by 2 to obtain the entity similarity value;

[0086] The entity similarity measure is a dimensionless result obtained by shifting the cosine similarity value. It is used to represent the degree of similarity in symptoms and identities among candidate pairs, and its value is between 0 and 1.

[0087] Specifically, the formula for calculating the entity similarity is as follows:

[0088]

[0089] In the formula, Indicates the first Round The symptom data item and the first Round Entity similarity between symptom data items Indicates the first Round A non-zero symptom vector for each symptom data item, Indicates the first Round A non-zero symptom vector for each symptom data item;

[0090] Furthermore, entity similarity only characterizes the degree of proximity of symptom identity information. The component order of the two symptom vectors is kept consistent during computation. When using fixed-dimensional sparse coding, synonyms, abbreviations, or successor expressions mapped to the same canonical symptom name have a consistent primary non-zero dimension.

[0091] Step S304: When generating entity similarity, the degree order values ​​of the current symptom data items and the previous version of the symptom data items are not introduced.

[0092] Specifically, the degree order value does not participate in symptom vector construction, cosine similarity calculation, or entity similarity generation. Symptoms becoming milder, more severe, explicitly absent, or having disappeared only change the state field, not the symptom identity input of the candidate pair.

[0093] In one embodiment of this application, step S4 includes the following steps:

[0094] Step S401: Add 1 to the intersection and union counts of the two relation vectors in the candidate pair, then divide them to obtain the relation consistency value;

[0095] The relation consistency measure is the degree of consistency between two relation vectors in a candidate pair on the same relation dimension.

[0096] Specifically, the formula for calculating the consistency quantity of the relationship is as follows:

[0097]

[0098] In the formula, Indicates the first Round The symptom data item and the first Round The consistency of the relationships between the symptom data items Indicates the first Round A fixed-dimensional binary relation vector for each symptom data item. Indicates the first Round A fixed-dimensional binary relation vector for each symptom data item. This represents the number of intersections between two relation vectors. This represents the number of elements in the union of two relation vectors.

[0099] Specifically, the intersection number of relation vectors is the number of dimensions in which both relation vectors have a value of 1. The union number is the number of dimensions in which at least one relation vector has a value of 1. The intersection and union numbers are each incremented by 1 and then divided to ensure that two relation vectors with no contextual relationship still yield a definite result. When both relation vectors are all zero, the relation consistency value is 1. When other dimensions remain constant, and the number of relation dimensions where only one relation vector has a value of 1 increases, the relation consistency value decreases.

[0100] Step S402: Select the upper quartile as the time scale from the aligned time difference samples that are manually confirmed to belong to the same symptom continuous records. The aligned time difference samples contain non-zero time differences and the time scale is fixed in the same deployment environment. Divide the aligned time difference of the candidate pair by the time scale, add 1 and take the reciprocal to obtain the time consistency quantity.

[0101] The time scale is a positive calibration parameter that normalizes the alignment time difference, and its dimensions are the same as the alignment time difference. The time consistency quantity represents the degree of closeness of candidate pairs under a unified time coordinate.

[0102] Specifically, the formula for calculating the time consistency quantity is as follows:

[0103]

[0104] In the formula, Indicates the first Round The symptom data item and the first Round Time consistency between symptom data items Indicates the first Round Alignment time period for each symptom data item Indicates the first Round Alignment time period for each symptom data item This represents the sum of the absolute values ​​of the differences between the start and end components of two aligned time periods. Represents a positive time scale determined by the upper quartile of the aligned time difference samples;

[0105] Specifically, the time scale is calibrated before the system is put into use. Calibration samples are selected from text pairs in adjacent data collection rounds that have been manually confirmed to belong to the same symptom semantic chain, covering common symptom names, synonyms, sequential expressions, durations, and relative date expressions. The sample set used for calibration must contain at least one non-zero aligned time lag. Upper quartile calibration is not performed if no non-zero aligned time lag sample is available.

[0106] For each sample, using the same time unit, the sum of the absolute values ​​of the starting and ending component differences between two aligned time periods is calculated to obtain the aligned time difference. Samples with an aligned time difference of 0 are not included in the upper quartile calculation. Non-zero aligned time differences are arranged in ascending order, and the number of samples with non-zero aligned time differences is multiplied by 0.75 and rounded up to obtain the target index. The aligned time difference corresponding to this target index is taken as the time scale.

[0107] The preferred timescale is in hours. The calibrated timescale is saved along with the glossary version and the time resolution rule version, and remains unchanged within the same deployment environment. If the time resolution rule version changes and alters the calculation method for the alignment period, or if the data collection cycle configuration changes and alters the time interval between adjacent data collection rounds, the timescale is recalibrated using newly collected and validated samples, and the updated timescale is used from the new version onwards.

[0108] Furthermore, the time consistency measure is used to convert the alignment time difference of candidate pairs into a degree of closeness within the range of 0 to 1. The starting and ending components of two alignment time periods use the same time unit; the absolute values ​​of their differences are first calculated separately, and then summed to obtain the alignment time difference. When the alignment time difference is 0, the time consistency measure is 1. With a fixed time scale, the time consistency measure monotonically decreases as the alignment time difference increases. The time scale only serves as a dimension normalization function and is not used as a weight among entity information, relational information, and time information.

[0109] Step S403: Perform three unweighted geometric convergences on entity similarity, relation consistency, and time consistency to obtain semantic association.

[0110] The semantic association quantity is a dimensionless result obtained by performing three unweighted geometric convergences of entity similarity quantity, relational consistency quantity, and temporal consistency quantity.

[0111] Specifically, the formula for calculating the semantic association quantity is as follows:

[0112]

[0113] In the formula, Indicates the first Round The symptom data item and the first Round The semantic association between symptom data items This represents the entity similarity measure for the same candidate pair. This represents the relational consistency quantity of the same candidate pair. This indicates the time consistency of the same candidate pair;

[0114] Specifically, the three unweighted geometric convergence parameters treat entity similarity, relational consistency, and temporal consistency equally, without assigning adjustable weights. The semantic association value ranges from 0 to 1. When any input value decreases while the other two remain unchanged, the semantic association value decreases accordingly. When any input is 0, the semantic association value is 0. This process avoids amplifying any type of information through manual weighting.

[0115] Step S404: Based on manually annotated text pairs of the same symptom semantic chain and different symptom semantic chains, a verification sample is constructed. The discrimination boundary is traversed within the range of greater than 0 and less than 1. The discrimination boundary with the smallest sum of the number of erroneous merging and the number of erroneous splitting is selected as the association threshold. Among the parallel minimum values, the discrimination boundary with the larger value is selected.

[0116] The validation samples consist of manually annotated text pairs with the same symptom semantic chain and text pairs with different symptom semantic chains. The association threshold is a dimensionless discrimination boundary obtained from the validation samples, with a value greater than 0 and less than 1.

[0117] Specifically, the validation samples contain both text pairs with the same symptom semantic chain and text pairs with different symptom semantic chains. Labeling is based on standardized symptom names, original context, collection time, and manually confirmed longitudinal records; labels are not generated by reverse engineering the association threshold to be calibrated. Validation samples are stored separately from the time-scale calibration samples.

[0118] For each validation sample, a semantic association value is generated and sorted by numerical value in ascending order to remove duplicates. The median value between two adjacent distinct values ​​is used as the candidate discrimination boundary. When the minimum value is greater than 0, half of the minimum value is used as the low-end candidate. When the maximum value is less than 1, the median value between the maximum value and 1 is used as the high-end candidate. Finally, only candidate discrimination boundaries greater than 0 and less than 1 are retained.

[0119] For each candidate discrimination boundary, the number of text pairs with different symptom semantic chains that are judged as related is recorded as the number of incorrect merging, and the number of text pairs with the same symptom semantic chain that are judged as not related is recorded as the number of incorrect splitting. The candidate discrimination boundary with the smallest sum of incorrect merging and incorrect splitting is selected as the association threshold. When multiple minimum values ​​exist, the candidate discrimination boundary with the larger value is selected. The association threshold is associated with and saved for the validation sample version, terminology version, and timescale version, and is used consistently in the same deployment environment.

[0120] Step S405: Filter candidate pairs using the association threshold, and compare the semantic association of the current symptom data item with all previous version symptom data items and the semantic association of the paired previous version symptom data item with all current symptom data items according to the bidirectional unique maximum condition. Set the semantic anchor flag to true only when both directions are unique maximum values, and set the semantic anchor flag corresponding to the parallel maximum values ​​to false.

[0121] The semantic anchor flag is a Boolean value that indicates whether a candidate pair simultaneously satisfies the association threshold and the bidirectional unique maximum condition, and its value is either true or false.

[0122] Specifically, candidate pairs with semantic association values ​​below the association threshold are first removed. The remaining semantic association values ​​are then arranged into a pairing matrix between the current symptom data item and the previous version's symptom data item. A candidate pair satisfies the bidirectional unique maximum condition only if it has a unique maximum value in both its row and column. If there are tied maximum values ​​in a row or column, no candidate pair exists in either direction, or the semantic association value does not reach the association threshold, the semantic anchor flag is set to false. When both directional unique maximum conditions are met, the semantic anchor flag is set to true, and the same symptom data item corresponds to at most one candidate pair with a true semantic anchor flag.

[0123] It should be noted that symptom severity changes with the course of the disease and the patient's subjective experience. The severity ordinal value is variable state information, while the symptom name, synonym mapping, and inheriting object are used to represent symptom identity. The severity ordinal value is not introduced when generating entity similarity metrics, which avoids the actual decrease in severity or disappearance of the symptom from indicating a lower degree of similarity. Relational consistency metrics and temporal consistency metrics supplement local sentence structure and time coordinate information, and then one-to-many pairings are restricted by a bidirectional unique maximum condition. When there are parallel maximum values, the unassociated state is maintained, and pairing objects are not selected in an arbitrary order, keeping symptom identity judgment and state updates separate, thereby reducing duplicate link building caused by state changes and erroneous merging caused by local semantic similarity.

[0124] In one embodiment of this application, step S5 includes the following steps:

[0125] Step S501: In this step, candidate pairs marked as true by semantic anchors are grouped into the same symptom semantic chain. The degree order value and alignment period of the current symptom data item are compared with those of the previous version symptom data item. Only the changed fields are written into the current version record.

[0126] The symptom semantic chain is a sequence of records connecting the valid states of the same symptom according to the collection round. A changed field is a field in the current round of symptom data that has changed relative to the previous valid symptom data item. The current round version record is an incremental record appended to the previous version record, used to store newly added items and changed fields in the current round, and to obtain unchanged fields through inheritance.

[0127] Specifically, each symptom semantic chain is assigned a unique record number. Candidate pairs with a true semantic anchor flag retain the record number of the symptom semantic chain to which the previous version of the symptom data item belongs. When the degree ordinal value of the current symptom data item is not null, it is compared with the previous valid degree ordinal value. If the values ​​are different, the current degree ordinal value is written to the change field; if the values ​​are the same, the previous valid degree ordinal value is inherited. When the degree ordinal value is null, the previous valid degree ordinal value is not overwritten.

[0128] The alignment period is compared separately according to the start and end times. If any component differs, a new alignment period is written. To enable the next round to recover complete symptom data items from the previous version, the symptom vector and relation vector, as technical fields, are stored using the same incremental method. If the symptom vector or relation vector differs from the previous valid value, a change field is written; if they are the same, inheritance continues. When the technical fields change, no new symptom semantic chain is created separately, nor are business fields added to the dynamic symptom table.

[0129] Furthermore, each changed field record includes the collection round, the record number of the symptom semantic chain, the field name, the new field value, and the collection time.

[0130] Step S502: If there are no candidate pairs with semantic anchor flags set to true in the current round of symptom data items, a new symptom semantic chain is established and written as a new item into the current round version record.

[0131] The new item is a candidate symptom data item in this round that has not formed a semantic anchor flag that is true with any symptom data item in the previous version.

[0132] Specifically, when none of the candidate pairs for a symptom data item in this round have a true semantic anchor, a new record number is assigned to that symptom data item, a new symptom semantic chain is established, and the symptom vector, degree ordinal value, alignment period, and relation vector are added as new items to the current version record. Even when the degree ordinal value is null, a symptom semantic chain can still be established, and the degree level in the dynamic symptom table will be displayed as ungraded.

[0133] Step S503: If there are no candidate pairs with true semantic anchors for the symptom data items in the previous version, retain the original record; if the self-assessment text in this round clearly expresses that the symptoms do not exist or have disappeared, form the lowest level of the degree order value according to the same ordered dictionary, and after forming a candidate pair with true semantic anchors between the symptom data items in this round and the symptom data items in the previous version, write the lowest level of the degree order value into the record of this round.

[0134] Specifically, if no candidate pair with a true semantic anchor exists for a symptom data item from the previous version, the original symptom semantic chain and its currently valid fields remain unchanged and are not deleted simply because the symptom is not mentioned in this round. Only when a complete negative expression or a clearly disappeared expression appears in the self-assessment text of this round, and a candidate pair with a true semantic anchor is formed between the symptom data item of this round and the symptom data item from the previous version, will the lowest degree ordinal value in the same ordered dictionary be written into the symptom semantic chain. If a negative expression or a disappeared expression does not form a candidate pair with a true semantic anchor, the existing symptom semantic chain is not changed, and the expression is retained for manual verification.

[0135] Step S504: Attach the current version record to the previous version record according to the collection round. The current version record only saves the newly added items and changed fields, while the unchanged fields are inherited from the previous version record.

[0136] Specifically, the current version record is appended to the previous version record without modifying existing historical records. When reading the status of any collection round, new items and changed fields are applied sequentially according to the collection round, starting from the first version. If a field has no changed records in the current round, the most recent valid value of that field is used. Before writing to the current version record, the record number, collection round, and field name of the symptom semantic chain are validated to avoid conflicting incremental records for the same field in the same round.

[0137] It should be noted that omissions in continuous self-reports do not equate to symptom disappearance; patients may only report changes. Previous records are retained when symptom data items lack candidate pairs with true semantic anchors, preventing omissions from being interpreted as state termination. A disappearance status is only written when a clear negation or disappearance word forms a minimum degree ordinal value and establishes a candidate pair with a true semantic anchor in the existing symptom semantic chain. New items and changed fields are appended and saved according to the collection round, while unchanged fields inherit from the most recent valid value. This ensures that each state change can be located to the corresponding collection round and that the valid state of any round can be restored without rewriting historical records. It also reduces duplicate storage of unchanged fields and maintains a traceability relationship between the dynamic symptom table and historical versions.

[0138] In one embodiment of this application, step S6 includes the following steps:

[0139] Step S601: Apply the new items and changed fields in the current version record in the order of the collection rounds, and inherit the unchanged fields to restore the symptom vector, degree order value and alignment time period of each symptom semantic chain;

[0140] Specifically, the system reads version records from earliest to latest according to the collection round. New entries are used to establish the initial valid values ​​for the symptom semantic chain. Changed fields are used to replace the valid values ​​of corresponding fields in the same symptom semantic chain. Fields without changed records continue to use the most recent valid value. After completing the application for all rounds, the system obtains the currently valid symptom vector, degree order value, and alignment time for each symptom semantic chain. When reading historical rounds, only version records no later than the target round are applied, thereby restoring the state corresponding to the target round.

[0141] Step S602: Map the symptom vector to the symptom name, map the degree sequence value to the degree level, and convert the aligned time period into a time interval;

[0142] The symptom name is the canonical symptom name corresponding to the symptom vector. The severity level is the textual level of the severity ordinal value in the same ordered dictionary. The time interval is the start and end time obtained by aligning time periods and formatting them according to a unified time zone.

[0143] Specifically, the symptom name is retrieved by querying the symptom terminology table based on the non-zero canonical dimension of the symptom vector. Synonyms, abbreviations, and successive expressions all return the canonical symptom names corresponding to that dimension. The degree level is obtained by querying the same ordered dictionary based on the degree ordinal value. A null degree ordinal value indicates no level, and the lowest level indicates it does not exist or has disappeared.

[0144] Furthermore, the start and end times of the aligned time period are converted into year, month, day, hour, and minute format according to a unified time zone. When the start and end times are the same, a time interval is formed based on a single observation point; when the start and end times are different, a time interval is formed from the start time to the end time.

[0145] Step S603: A dynamic symptom table is formed by symptom name, severity level, and time interval.

[0146] The dynamic symptom table is a current status result consisting of three fields: symptom name, severity level, and time interval.

[0147] Specifically, each currently valid symptom semantic chain generates one row in the dynamic symptom table. The dynamic symptom table only sets three business fields: symptom name, severity level, and time interval; it does not include entity similarity, relationship consistency, time consistency, semantic association, or semantic anchor flags. Symptom semantic chains that are non-existent or have disappeared are still retained in the dynamic symptom table, with their current status represented by the corresponding severity level. Historical changes are queried through version records and are not repeatedly expanded in the dynamic symptom table.

[0148] like Figure 2As shown, this application can be deployed in a data acquisition system consisting of a symptom self-assessment terminal and a server. The self-assessment terminal receives natural language text or speech-to-text input from the keyboard. The server generates the acquisition time based on the complete submission arrival time and performs semantic unit formation, symptom data item construction, cross-version association, incremental writing, and dynamic symptom table reconstruction under a unified time unit. For example, the self-assessment text collected at 09:00 on day 1 is "Upper abdominal pain, moderate, occasional nausea for the past 3 days." The system forms symptom data items corresponding to upper abdominal pain and nausea, and saves the corresponding degree order value and alignment time period. The self-assessment text collected at 09:00 on day 2 is "Feeling better today, no more nausea." The system associates "feeling better today" with upper abdominal pain based on the continuation relationship, and sets the degree order value corresponding to nausea to the lowest level based on the explicit disappearance words. After the two symptom data items in this round form a semantic anchor with the previous version record, only the degree order value and alignment time period change of upper abdominal pain, as well as the lowest degree level of nausea, are written; unchanged fields continue to be inherited. The reconstructed dynamic symptom table displays the current severity level and time interval of upper abdominal pain, and indicates that nausea is absent or has disappeared; the version history still retains the content from day 1. This result can be used for viewing symptom records, organizing follow-up data, and statistical processing, but does not automatically generate diagnostic or treatment conclusions.

[0149] This application provides an information processing and storage system, including a processor and a memory. The memory stores computer programs and version records. When the computer program is executed by the processor, it causes the processor to perform the following operations:

[0150] Collect the self-evaluation text and collection time of this round, standardize and segment the self-evaluation text, and form semantic units that retain contextual relationships;

[0151] Extract symptom entities, degree words, negation words, disappearance words, time expressions, and contextual relationships from semantic units to form symptom vectors, degree order values, relative time periods, and relationship vectors. Combine these with the collection time to form aligned time periods and compose symptom data items.

[0152] The symptom data items from the previous version are restored and paired with the symptom data items in the current round to form candidate pairs. Entity similarity is generated from the symptom vectors without introducing degree order values.

[0153] Based on the relation vector and the alignment time period, relation consistency quantity and time consistency quantity are generated respectively. They are geometrically converged with entity similarity quantity without weight into semantic association quantity. Semantic anchor mark is generated based on association threshold and bidirectional unique maximum condition.

[0154] Candidate pairs marked as true with semantic anchors are included in the symptom semantic chain, and the degree order value and alignment time period are compared. Only the change field is written. The current round of symptom data items without semantic anchors are added as new items to form the current round version record.

[0155] The symptom names, severity levels, and time intervals are recovered from the symptom semantic chain in this version to form a dynamic symptom table.

[0156] This application also provides a computer-readable storage medium storing a computer program thereon; when the computer program is executed by a processor, it implements the intelligent collection and dynamic management method for symptom self-assessment data of gastrointestinal tumor patients as described in any embodiment of this application.

[0157] A portion of this application can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to this application through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0158] The technical solutions provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for intelligent collection and dynamic management of symptom self-assessment data from patients with gastrointestinal tumors, characterized in that: The method includes the following steps: Step S1: Collect the self-evaluation text and collection time of this round, standardize and segment the self-evaluation text, and form semantic units that retain contextual relationships; Step S2: Extract symptom entities, degree words, negation words, disappearance words, time expressions and contextual relationships from semantic units to form symptom vectors, degree order values, relative time periods and relationship vectors. Combine these with the collection time to form aligned time periods and compose symptom data items. Step S3: Recover the symptom data items from the previous version record, form candidate pairs with the symptom data items in the current round, generate entity similarity from the symptom vector, and do not introduce degree order values; Step S4: Generate relation consistency quantity and time consistency quantity according to relation vector and alignment time period respectively, and geometrically converge them with entity similarity quantity without weight to form semantic association quantity. Generate semantic anchor mark according to association threshold and bidirectional unique maximum condition. Step S5: In this step, candidate pairs marked as true with semantic anchors are included in the symptom semantic chain, the degree order value and the alignment time period are compared, and only the change field is written; the current round of symptom data items without semantic anchors are added as new items to form the current round version record. Step S6: Recover the symptom name, severity level, and time interval from the current version record to form a dynamic symptom table.

2. The method for intelligent collection and dynamic management of symptom self-assessment data of gastrointestinal tumor patients according to claim 1, characterized in that, Step S1 includes: Collect the self-assessment texts of patients with gastrointestinal tumors in this round and the time of collection. Perform character standardization, punctuation restoration and colloquial redundancy removal on the self-assessment texts to maintain symptom words, degree words, negation words, time words and the original order. The text is segmented based on sentence-ending punctuation, coordinating conjunctions, adversative conjunctions, and symptom trigger words; the contextual relationships between adjacent text segments, consisting of continuation, reference, coordination, adversation, degree modification, and time modification, are preserved, and the text segments and the aforementioned contextual relationships form semantic units.

3. The method for intelligent collection and dynamic management of symptom self-assessment data of gastrointestinal tumor patients according to claim 1, characterized in that, Step S2 includes: Perform symptom entity recognition, contextual analysis, degree word parsing, and time expression parsing on semantic units; encode symptom entities and their synonyms, abbreviations, and contextual relationships into non-zero symptom vectors; map degree words, negation words, and disappearance words to degree order values ​​according to the same ordered dictionary, with the lowest level corresponding to the absence or disappearance of the symptom; The time expression is parsed into a binary start and end offset relative to the acquisition time to form a relative time period. The relationships of succession, same reference, parallel, transition, degree modification and time modification are encoded into a fixed-dimensional binary relationship vector. The data collection time is added to the two start and end components of the relative time period to form the aligned time period. The symptom data item consists of the symptom vector, the degree order value, the aligned time period, and the relationship vector, and the relative time period is retained.

4. The method for intelligent collection and dynamic management of symptom self-assessment data of gastrointestinal tumor patients according to claim 1, characterized in that, Step S3 includes: Recover the symptom data items of each symptom semantic chain from the previous version record, and form candidate pairs by pairing the symptom data items of the current round with the symptom data items of the previous version; Calculate the cosine similarity value of the two non-zero symptom vectors in the candidate pair, add 1 to the cosine similarity value and divide by 2 to obtain the entity similarity value. No degree order value is introduced when generating the entity similarity value.

5. The method for intelligent collection and dynamic management of symptom self-assessment data of gastrointestinal tumor patients according to claim 1, characterized in that, Step S4 includes: Add 1 to the intersection and union counts of the two relation vectors in the candidate pair, then divide them to obtain the relation consistency value; The timescale is taken from the upper quartile of the aligned time difference sample of consecutive records of the same symptom confirmed by manual verification, the sample contains non-zero time differences, and the timescale is fixed in the same deployment environment; Divide the alignment time difference of the candidate pair by the time scale, add 1, and take the reciprocal to obtain the time consistency quantity; Perform three unweighted geometric convergences on entity similarity, relational consistency, and temporal consistency to obtain semantic association.

6. The method for intelligent collection and dynamic management of symptom self-assessment data of gastrointestinal tumor patients according to claim 1, characterized in that, Step S4 also includes: The validation samples consist of text pairs of the same symptom semantic chain and different symptom semantic chains that are manually annotated. The discrimination boundary is traversed within the range of greater than 0 and less than 1. The discrimination boundary with the smallest sum of the number of erroneous merging and the number of erroneous splitting is used as the association threshold. The larger of the two minimum values ​​is taken as the discrimination boundary. Candidate pairs are filtered by association threshold. The semantic association of the current symptom data item with all previous version symptom data items is compared with the semantic association of the previous version symptom data item with all current symptom data items according to the bidirectional unique maximum condition. The semantic anchor flag is set to true only when both directions are unique maximum values, and the semantic anchor flag of the symmetrical maximum values ​​is set to false.

7. The method for intelligent collection and dynamic management of symptom self-assessment data of gastrointestinal tumor patients according to claim 1, characterized in that, Step S5 includes: Candidate pairs with semantic anchors marked as true are grouped into the same symptom semantic chain. The degree ordinal value and alignment period of the current symptom data item are compared with those of the previous version symptom data item. Only the changed fields are written to the current version record. When there are no candidate pairs with semantic anchors that are true for symptom data items in the current round, a new symptom semantic chain is established and written as a new item in the current version record; when there are no candidate pairs with semantic anchors that are true for symptom data items in the previous version, the original record is retained. The self-assessment text in this round clearly expresses the lowest level of degree order value when the symptoms do not exist or have disappeared, and writes it into the version record of this round after forming candidate pairs with semantic anchor flags as true. The current version record is attached to the previous version record according to the collection round, and only the newly added items and changed fields are saved. Unchanged fields are inherited from the previous version record.

8. The method for intelligent collection and dynamic management of symptom self-assessment data of gastrointestinal tumor patients according to claim 1, characterized in that, Step S6 includes: The newly added items and changed fields in the current version record are applied in the order of the collection rounds, and the unchanged fields are inherited to restore the symptom vector, degree order value and alignment time period of each symptom semantic chain; The symptom vector is mapped to the symptom name, the severity ordinal value is mapped to the severity level, and the aligned time period is converted into a time interval, forming a dynamic symptom table composed of symptom name, severity level, and time interval.

9. An information processing and storage system, characterized in that, It includes a processor and a memory, the memory being used to store computer programs and version records, the computer programs being executed by the processor to cause the processor to perform the following operations: Collect the self-evaluation text and collection time of this round, standardize and segment the self-evaluation text, and form semantic units that retain contextual relationships; Extract symptom entities, degree words, negation words, disappearance words, time expressions, and contextual relationships from semantic units to form symptom vectors, degree order values, relative time periods, and relationship vectors. Combine these with the collection time to form aligned time periods and compose symptom data items. The symptom data items from the previous version are restored and paired with the symptom data items in the current round to form candidate pairs. Entity similarity is generated from the symptom vectors without introducing degree order values. Based on the relation vector and the alignment time period, relation consistency quantity and time consistency quantity are generated respectively. They are geometrically converged with entity similarity quantity without weight into semantic association quantity. Semantic anchor mark is generated based on association threshold and bidirectional unique maximum condition. Candidate pairs marked as true with semantic anchors are included in the symptom semantic chain, and the degree order value and alignment time period are compared. Only the change field is written. The current round of symptom data items without semantic anchors are added as new items to form the current round version record. The symptom names, severity levels, and time intervals are recovered from the symptom semantic chain in this version to form a dynamic symptom table.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute a method for intelligent collection and dynamic management of symptom self-assessment data of gastrointestinal tumor patients as described in any one of claims 1 to 8.

Images