Search

CN-121997940-A - National talk double-channel medicine information semantic management system

CN121997940ACN 121997940 ACN121997940 ACN 121997940ACN-121997940-A

Abstract

The invention relates to the technical field of medicine information semantic management, and discloses a national talk double-channel medicine information semantic management system. The system comprises a state observation unit, a correlation measurement unit, an attribution analysis unit, a characteristic purification unit and a semantic integration unit. The system extracts semantic features through framing and synthesizes feature profiles, and a multi-scale semantic association matrix is constructed to identify key transition points and generate a track map. The transition points are mapped to the original information elements, and the evolution paths of the transition points across the contexts are tracked to form attribution chains. And generating a causal-related semantic factor set based on the chain, and iteratively stripping noise similarity factors through feedback to obtain a purified core factor set. And carrying out time sequence reconstruction and context filling on the core factors according to the evolution sequence, and outputting a standardized semantic record. The system can deeply analyze the causal and dynamic processes of semantic evolution, and improves the accuracy and the interpretability of information management.

Inventors

  • SU DAN

Assignees

  • 常州市第二人民医院
  • 医顺通信息科技(江苏)有限公司

Dates

Publication Date
20260508
Application Date
20260130

Claims (10)

  1. 1. The national talk binary channels medicine information semantic management system is characterized by comprising: The state observation unit is used for continuously collecting data waveforms of the semantic information flow, dividing waveform data into a plurality of data frames with fixed time length, extracting and quantizing multi-order semantic features of each data frame, and synthesizing quantized features of all data frames into feature profiles, wherein the feature profiles correspond to the complete evolution process of the medicine information flow; The association measurement unit is used for carrying out multi-scale alignment on the characteristic sections, respectively constructing semantic association matrixes under different time scales, identifying key semantic transition points by calculating the difference degree among the matrixes, and generating a track map of information variation; The attribution analysis unit analyzes the track map of the information variation, maps transition points in the track map to specific information elements in the original medicine data stream, and tracks evolution paths of the specific information elements in different semantic contexts to form a structured attribution chain; The feature purification unit is used for generating a semantic factor set with causal relation with information variation based on the structured attribution chain, stripping semantic factors similar to a noise mode in the semantic factor set through feedback iterative computation, and outputting a purified core semantic factor set; the semantic integration unit inputs the core semantic factor set into a sequence generation model, and the model carries out time sequence reconstruction and context filling on the core semantic factors according to the evolution sequence recorded in the structured attribution chain and outputs standardized drug information semantic records.
  2. 2. The national talk dual channel drug information semantic management system according to claim 1, wherein the state observation unit comprises the following: Continuously receiving original data of a medicine information stream through a sensor or an interface, and converting the original data stream into an analyzable waveform signal; Uniformly cutting the waveform signal on a time axis to form a series of non-overlapping data frames with fixed duration; For each data frame, performing extraction operation of multi-order semantic features in parallel, wherein the extracted semantic features at least comprise word frequency distribution features, syntax structure features and context association features; respectively converting the extracted multi-order semantic features into numerical vectors to finish the quantization process; Splicing and integrating the quantized feature vectors of all the data frames according to the time sequence to construct a feature profile representing the complete information evolution period.
  3. 3. The national talk dual channel drug information semantic management system according to claim 2, wherein the association metric unit comprises the following: receiving the characteristic section, and presetting a plurality of different observation scales for the characteristic section, wherein each observation scale corresponds to a specific time window length; Dividing the feature profile into continuous segments under each preset observation scale, calculating the similarity between feature vectors of adjacent segments, and arranging all similarity values to form a semantic association matrix under the observation scale; Comparing a plurality of semantic association matrixes generated under different observation scales, and generating a scale difference matrix by calculating absolute differences of corresponding position elements of the matrixes; Identifying peak points with values exceeding a preset threshold in the scale difference matrix, wherein the peak points are defined as key semantic transition points; and connecting all the identified key semantic transition points according to the time sequence of occurrence of the key semantic transition points, and marking the corresponding scale difference intensity of the key semantic transition points to form a track map of information variation.
  4. 4. The national talk dual channel drug information semantic management system according to claim 3, wherein the attribution parsing unit comprises: Backtracking to the characteristic section and the original medicine information flow according to the timestamp of the key semantic transition point recorded in the track map of the information variation, and positioning a specific text unit or a data field which causes the key semantic transition point; Taking the specific text unit as a starting point, carrying out bidirectional scanning forwards and backwards along a time axis, and collecting all context information blocks containing the specific text unit or highly related to the semantics of the specific text unit; Arranging all the collected context information blocks according to a time line, and analyzing semantic role changes, attribute modification changes and association relation changes of the specific text units in different information blocks; Recording the semantic role change, the attribute modification change and the association relation change of the specific text unit as a series of event nodes with causal or time sequence connection, and connecting all the event nodes in series to form a structured attribution chain from a starting state to a transition state.
  5. 5. The national talk dual channel drug information semantic management system according to claim 4, wherein the feature purification unit comprises: receiving the structured attribution chain, and extracting all event nodes recorded as generating changes from the chain; abstracting the information change corresponding to each event node into an independent semantic factor, wherein all the semantic factors form an initial semantic factor set; establishing a noise feature library parallel to the semantic factor set, wherein the noise feature library stores common interference modes obtained by statistics of historical data; comparing each factor in the initial semantic factor set with all modes in the noise feature library one by one, and calculating the morphological similarity of the factors; If the similarity between one semantic factor and any noise pattern exceeds a set threshold, starting a stripping program, and subtracting the common part of the noise patterns from the characteristic expression of the semantic factor; After the comparison and stripping operation of multiple iterations, the rest semantic factors which are not highly similar to the noise mode are formed into a purified core semantic factor set.
  6. 6. The national talk dual channel drug information semantic management system according to claim 5, wherein the semantic integration unit comprises the following: The core semantic factor set is used as the content input of the sequence generation model, and the evolution sequence of the event nodes recorded in the structured attribution chain is used as the structure input of the sequence generation model; the sequence generation model firstly determines the approximate time sequence position of the core semantic factors in the output sequence according to the structural input; Generating a context description segment which accords with the drug information field specification for each core semantic factor according to the content input and the position information by the model; The model combines all the generated description fragments according to the determined time sequence positions, ensures the logical continuity among the fragments, and outputs a part of medicine information semantic record with complete structure and semantic specification.
  7. 7. The national talk dual channel drug information semantic management system according to claim 6, further comprising a profile updating unit comprising: continuously monitoring the newly generated medicine information flow, and generating a new characteristic section by utilizing the state observation unit; Comparing the new characteristic section with the historical information variation track map to identify a new semantic transition mode which appears and is not recorded by the map; And (3) expanding nodes and paths of the existing information variation track map by using the complementary structured attribution chain to form an updated information variation track map.
  8. 8. The national talk dual channel drug information semantic management system according to claim 7, wherein the system further comprises a verification closed loop unit comprising: acquiring a medicine information semantic record output by the semantic integration unit, and comparing the medicine information semantic record with an authoritative medicine information database in the real world; Reversely mapping the difference points found in the comparison to the structured attribution chain and the core semantic factor set, and positioning source nodes or factors generating the difference; Generating a calibration instruction according to the property of the difference, wherein the calibration instruction is used for adjusting a scale difference threshold value of the correlation measurement unit or the noise peeling strength of the characteristic purification unit; And feeding the calibration instruction back to a corresponding processing unit of the system, and driving the system parameters to perform self-adaptive optimization.
  9. 9. The national talk dual channel drug information semantic management system according to claim 8, wherein the system further comprises a cache acceleration unit comprising: A characteristic profile buffer area is arranged between the state observation unit and the associated measurement unit and is used for temporarily storing characteristic profile data generated in a last period of time; When the system needs to process new medicine information flow, firstly searching whether a history characteristic section which is highly similar to the current information flow exists in a characteristic section buffer area; If the information variation exists, directly calling a track map of the pre-calculated information variation corresponding to the historical characteristic section and a subsequent processing result, and performing incremental calculation on only a difference part by taking the track map of the pre-calculated information variation and the subsequent processing result as initial references of current processing.
  10. 10. The national talk dual channel drug information semantic management system according to claim 9, wherein the system further comprises an interface adaptation unit comprising: when the system receives the external medicine information flow, the original data with different sources and different formats are converted into uniform waveform signals in the system through protocol analysis; When the system outputs the medicine information semantic record, a corresponding semantic description template is called according to the requirement of a target application scene, and an internal unified record format is converted into output data conforming to a specific interface specification.

Description

National talk double-channel medicine information semantic management system Technical Field The invention relates to the technical field of medicine information semantic management, in particular to a national talk double-channel medicine information semantic management system. Background Existing technology for semantic management of drug information generally adopts a rule-based template matching or statistical machine learning model for processing. The method realizes classification and monitoring of information by word segmentation and feature vectorization of texts and calculation of co-occurrence frequency or statistical correlation among features. The interest is to identify key words from semantic streams, extract static features, or judge the category to which the information belongs through a classification model, which essentially discretizes and describes and generalizes the semantic information in a cross-section manner. The main drawbacks of the prior art are two. The feature extraction and analysis process has difficulty distinguishing causal factors from concomitant noise in semantic evolution. The conventional method relies on statistical correlation to construct a feature set, can not effectively strip pseudo features formed by data noise and accidental co-occurrence, causes semantic understanding to deviate from a real driving factor, and lacks the capability of describing the continuous evolution process of information entities across contexts. Most of records in the prior art are information state snapshots or independent events at discrete time points, and state association of the same information element under different semantic scenes cannot be established, so that a complete dynamic evolution path cannot be restored, and the realization of depth attribution and accurate tracing is limited. The technical scheme is needed, the core semantic factors with causal interpretation ability can be stripped from dynamic semantic streams, and the whole evolution process of specific information elements in continuous semantic contexts can be tracked, so that deep analysis of the semantic evolution essence and paths of medicine information can be realized. Disclosure of Invention The invention aims to provide a national talk double-channel medicine information semantic management system so as to solve the problems in the background technology. In order to achieve the above object, the present invention provides a national talk dual channel drug information semantic management system, the system comprising: The state observation unit is used for continuously collecting data waveforms of the semantic information flow, dividing waveform data into a plurality of data frames with fixed time length, extracting and quantizing multi-order semantic features of each data frame, and synthesizing quantized features of all data frames into feature profiles, wherein the feature profiles correspond to the complete evolution process of the medicine information flow; The association measurement unit is used for carrying out multi-scale alignment on the characteristic sections, respectively constructing semantic association matrixes under different time scales, identifying key semantic transition points by calculating the difference degree among the matrixes, and generating a track map of information variation; The attribution analysis unit analyzes the track map of the information variation, maps transition points in the track map to specific information elements in the original medicine data stream, and tracks evolution paths of the specific information elements in different semantic contexts to form a structured attribution chain; The feature purification unit is used for generating a semantic factor set with causal relation with information variation based on the structured attribution chain, stripping semantic factors similar to a noise mode in the semantic factor set through feedback iterative computation, and outputting a purified core semantic factor set; the semantic integration unit inputs the core semantic factor set into a sequence generation model, and the model carries out time sequence reconstruction and context filling on the core semantic factors according to the evolution sequence recorded in the structured attribution chain and outputs standardized drug information semantic records. Preferably, the state observing unit includes the following: Continuously receiving original data of a medicine information stream through a sensor or an interface, and converting the original data stream into an analyzable waveform signal; For each data frame, performing extraction operation of multi-order semantic features in parallel, wherein the extracted semantic features at least comprise word frequency distribution features, syntax structure features and context association features; respectively converting the extracted multi-order semantic features into numerical vectors to finish the quantization process;