Dictionary construction and classification method and device for spacecraft test identification and evaluation

Abstract

The present invention discloses a method and device for dictionary construction and classification for spacecraft test identification and evaluation, which solves the problems that spacecraft telemetry name components are unique, common word segmentation algorithms are not applicable, and telemetry feature extraction is difficult. A spacecraft telemetry concept space is created, and the spacecraft telemetry concept space includes multiple telemetry concepts; a spacecraft experiment identification dictionary structure is defined, and the dictionary stores multiple nodes, each node represents a Chinese character, English character or symbol, and the dictionary is a multi-layer dictionary, each layer stores multiple nodes, and the number of layers stored in the dictionary is determined based on the position of the character or character in a certain word; each node has a pair of keys and values, wherein the key is used to represent Chinese characters, English characters or symbols, and the value is used to represent the attribute corresponding to the key, as well as the connection relationship of the node in the dictionary; the phrases corresponding to each telemetry concept in the spacecraft telemetry concept space are inserted into the dictionary to construct a spacecraft experiment identification dictionary.

Description

Technical Field

[0001] The present invention relates to the field of information processing, and in particular to a method and device for dictionary construction and classification for spacecraft test identification and evaluation. Background Art

[0002] In the process of spacecraft production and development, test and appraisal is an important task for quality evaluation and mission effectiveness assessment. The evaluation input information and data sources are wide and diverse. The design documents of spacecraft models in different fields vary greatly, and there are a lot of parameter entry and configuration work when matching the interface with the test and appraisal system. With the application, launch and delivery of new models, the back-end evaluation, interpretation, analysis and report generation modules of the test and appraisal system need to make a lot of adjustments according to the changes in mission requirements, resulting in the automation means being limited to the implementation of test and appraisal, and it is difficult to further expand to the preparation and result generation of test and appraisal assessment. Standardizing and constraining from the input end and unifying the data access format of multiple types of spacecraft face great management difficulties and execution resistance. If the file content can be intelligently classified and the content format that meets the standards of the test and appraisal system itself can be automatically formed, it will effectively solve the problem of the impact of input changes on the configuration of the test and appraisal system, which can greatly save the workload of related tasks, save labor costs, and further shorten the development cycle. In order to extract key content from different spacecraft design documents and test and appraisal work documents and form automated configuration files, it is necessary to effectively segment the telemetry names in the field of spacecraft. However, there are several problems in the implementation process.

[0003] (1) Unique components of spacecraft telemetry names

[0004] Spacecraft belong to a specialized technical field. Their telemetry names are composed of professional technical vocabulary or obscure abbreviations, with Chinese, English and symbols mixed together. The vocabulary is unique, the combinations of words are varied, and they do not follow the syntax in daily habits.

[0005] (2) Common dictionaries and word insertion algorithms are not applicable

[0006] The word segmentation of spacecraft telemetry involves the design of dictionary structure and word insertion algorithm. The commonly used tree-shaped dictionary takes up a lot of space, is difficult to maintain incrementally, and has low retrieval efficiency. Its word insertion algorithm is not optimized for the characteristics of spacecraft telemetry, and its correctness is difficult to guarantee, which cannot meet the application requirements of telemetry word segmentation.

[0007] (3) Difficulty in extracting feature information

[0008] To achieve classification, the feature extraction process of spacecraft telemetry names requires vectorization of the word segmentation results. Since telemetry names have different lengths, vectorized telemetry names cannot simply digitize Chinese characters, English letters, and symbols and then combine them in sequence. Since most telemetry names are short, the concepts involved have important meanings, and the frequency of occurrence of a single word is low, it is not suitable to use common design ideas such as keywords, word frequency weights, and stop words. Summary of the invention

[0009] In view of this, the present invention provides a method and device for dictionary construction and classification for spacecraft test identification and evaluation, which is aimed at spacecraft test identification and evaluation and solves the problems of unique components of spacecraft telemetry names, inapplicability of commonly used word segmentation algorithms, and difficulty in extracting telemetry features.

[0010] In order to solve the above technical problems, the present invention is implemented as follows.

[0011] A dictionary construction and classification method for spacecraft test identification and evaluation, including:

[0012] Step S1: Create a spacecraft telemetry concept space, the spacecraft telemetry concept space includes multiple telemetry concepts, each telemetry concept has five attributes, namely, number, phrase, alias, weight, and category; the number represents the sequence number of the telemetry concept, the phrase stores the name of the telemetry concept itself, the alias represents the alias of the telemetry concept, the category attribute is used to represent the category to which the telemetry concept belongs, and the weight is used to represent the importance of the category to which the telemetry concept belongs when classified by semantics;

[0013] Step S2: define the structure of the spacecraft experiment identification dictionary, the dictionary stores a plurality of nodes, each node represents a Chinese character, English character or symbol, the dictionary is a multi-layer dictionary, each layer stores a plurality of nodes, based on the position of the character or the character in a certain word, the number of layers in which it is stored in the dictionary is determined; each node has a pair of key and value, wherein the key is used to represent the Chinese character, English character or symbol, the value is used to represent the attribute corresponding to the key, and the connection relationship of the node in the dictionary, the connection relationship is used to indicate the pre-order node and post-order node of the node, the pre-order node is the node located before the node in a certain word, and the post-order node is the node located after the node in a certain word;

[0014] Step S3: inserting the phrases corresponding to each telemetry concept in the spacecraft telemetry concept space into the dictionary to construct a spacecraft experiment identification dictionary.

[0015] Preferably, in step S2, the spacecraft experiment identification dictionary structure is a multi-layer mesh dictionary structure, wherein the dictionary has multiple nodes, each node is described by a key and a value, wherein the key corresponds to a word, a character or a symbol, and the value can represent the relationship attribute between the node and the node word; the node value is composed of a node index, a plurality of node word indexes, a node word state corresponding to each node word index, a node word parent node index corresponding to each node word index, and a node word child node index corresponding to each node word index;

[0016] The node index represents the number of the node in the current layer of the dictionary. In the same layer, the number increases in the order of insertion. In different layers, the same key can correspond to different node indexes.

[0017] The node words corresponding to the node refer to all the phrases in the spacecraft telemetry concept space, including the key corresponding to the node, and all the phrases whose positions in the phrases are the same as the level of the current level;

[0018] The node word index indicates the number of the node word corresponding to the node. Each node word has a unique number, and the number is a global quantity, which has nothing to do with the layer number of the node corresponding to the node word. A node can correspond to several node words, each node word has a number, and each node can correspond to several node word indexes.

[0019] The node word state indicates the position information of the node in the node word, which is specifically divided into a continuing state marked as 0 and a terminal state marked as 1. The continuing state indicates that the node is not at the last position of the node word, and the terminal state indicates that the node is at the last position of the node word.

[0020] The parent node index of a node word indicates the node index of the previous node in the previous layer of the current layer corresponding to the node in the node word;

[0021] The node word subnode index represents the node index of the subsequent node corresponding to the node in the node word in the next layer of the current layer;

[0022] If the node word corresponding to the node does not have a parent node, the parent node is indexed as 0. If the node word corresponding to the node does not have a child node, the node word child node is indexed as 0.

[0023] Preferably, the step S3 comprises the following steps:

[0024] Step S31: If all telemetry concepts in the spacecraft telemetry concept space have been inserted into the dictionary, the method ends; otherwise, proceed to step S32;

[0025] Step S32: obtaining a telemetry concept in the spacecraft telemetry concept space that has not been inserted into the dictionary, obtaining a phrase corresponding to the telemetry concept, using the phrase as a node word, and setting a node word index ii for the node word;

[0026] Step S33: read an unprocessed character code from the node word in order from left to right, and obtain the position num of the character in the node word; if the character code is not the last character of the node word, proceed to step S34; if the character code is the last character of the node word, proceed to step S35;

[0027] Step S34: If the unprocessed character code is not included in the node of the numth layer of the dictionary, a node is inserted in the numth layer of the dictionary, the key of the node is set to the unprocessed character code, and the node word index ii is added to the node word index of the node; if the unprocessed character code is included in the node corresponding to the numth layer of the dictionary, the node word index ii is added to the node word index of the node corresponding to the character code; the node word state corresponding to the node word index ii is marked as the continue state; the node word parent node index corresponding to the node word index ii is marked as the preceding character of the character code in the node word in the dictionary The node index corresponding to the num-1 layer; the node word child node index corresponding to the node word index ii is marked as the node index corresponding to the subsequent character of the character code in the node word in the num+1th layer of the dictionary; wherein, if the preceding character of the character code is empty, the node word parent node index corresponding to the node word index ii is marked as 0; if the num+1th layer does not contain the subsequent character of the character code of the node word, the node word child node index corresponding to the node word index ii is marked as the current total number of nodes in the num+1th layer of the dictionary + 1; the unprocessed character code is marked as processed, and return to step S33;

[0028] Step S35: If the node at the numth layer of the dictionary does not contain the unprocessed character code, then insert a node at the numth layer of the dictionary, set the key of the node to the unprocessed character code, and add the node word index ii to the node word index of the node; if the node at the numth layer of the dictionary contains the unprocessed character code, then add the node word index ii to the node word index of the node corresponding to the character code; the node word state corresponding to the node word index ii is marked as the terminal state, the node word parent node index corresponding to the node word index ii is marked as the node index corresponding to the preceding character of the character code in the node word in the num-1th layer of the dictionary, and the node word child node index corresponding to the node word index ii is marked as 0; return to step S31.

[0029] A classification method for spacecraft test evaluation is provided, the method is based on the spacecraft test evaluation dictionary as described above, and the method comprises the following steps:

[0030] Step S41: Acquire telemetry vocabulary for spacecraft test, identification and evaluation, and perform word segmentation preprocessing on the telemetry vocabulary. The word segmentation preprocessing method is to segment the hyphens of the telemetry vocabulary for spacecraft test, identification and evaluation, and remove invalid characters and interference characters.

[0031] Step S42: based on the spacecraft experiment identification dictionary, segmenting the spacecraft test identification evaluation telemetry vocabulary obtained after word segmentation preprocessing into words, and using the segmented words as node words;

[0032] Step S43: define word weight, word category weight, keyword adjustment weight and word order weight, and vectorize each node word;

[0033] The word weight calculates the root mean square of the sum of the character codes in the word; the word category weight refers to the weight of each category in the concept space, which indicates the relative importance of the category to which the telemetry concept corresponding to the word belongs; the keyword adjustment weight is the adjustment weight added on the basis of the word category weight, which indicates further adjustment of the word category weight, and is divided into keywords, secondary words and invalid words; the word sequence weight refers to the weight assigned according to the order in which the words appear in the telemetry, and decays according to an exponential; the normalized word weight is defined according to the word weight, word category weight, keyword adjustment weight and word sequence weight, and the normalized similar word weight and the total normalized word weight under each concept category are calculated respectively; the word confidence is calculated, which indicates the percentage of known concept participles in the word to the total number of word parts; the normalized word weight, various normalized similar word weights and confidence are combined to obtain a 12-dimensional whole word weight vector.

[0034] Step S44: using the 12-dimensional whole word weight vector corresponding to each node word as input, hierarchical clustering is used to classify all node words.

[0035] Preferably, the step S42 includes:

[0036] Step S421: The result obtained after word segmentation preprocessing of a certain spacecraft test evaluation telemetry vocabulary is recorded as a set S, S = {S1, ..., S i ,…,S n}, where S i is the i-th subword formed after the telemetry vocabulary is preprocessed by word segmentation, 1≤i≤n, n is the total number of subwords formed after the telemetry vocabulary is segmented; define the word segmentation state C, C={C1,…,C j ,…,C m}, where C jis the state corresponding to the jth word or character of the telemetry vocabulary, and its value is 0 for the continuation state, 1 for the word break point, or 2 for the unknown concept, 1≤j≤m, and m is the total number of words or characters in the telemetry vocabulary;

[0037] Step S422: If all subwords have been segmented, the method ends; otherwise, a new subword S to be processed is selected. i , the pointer points to the subword S to be processed i The first character of ;

[0038] Step S423: If the subword to be processed S i If the word segmentation is completed, go to step S422; otherwise, obtain the segment to be processed S i The character k pointed to by the pointer in the middle, go to step S424;

[0039] Step S424: Search the first-level nodes of the dictionary to see if they contain the character k pointed to by the pointer. If there is a node that is consistent with the character k, then record the node corresponding to the character k as the current node, and record the node word corresponding to the node word index of the current node as the valid query node word, and then go to S425; if there is no node that is consistent with the character k in the first-level nodes of the dictionary, then set the segmentation state C corresponding to the character k to the valid query node word. j Marked as unknown concept, the pointer points to the current subword to be processed S i The next character of character k in the figure, if there is a next character, the step returns to S424, otherwise the step turns to S429;

[0040] Step S425: Read the current subword to be processed S i The next character of the character k pointed to by the pointer is recorded as character k+1, and the process goes to step S426; if the current subword to be processed S i If there is no next character in the sequence, go to step S428;

[0041] Step S426: Search all valid query node words of the node corresponding to character k, and determine whether the next-layer dictionary node corresponding to the node word sub-node index of all valid query node words is consistent with character k+1. If not, go to step S427; otherwise, set the word segmentation state C corresponding to character k to the same as the word segmentation state C of character k. j Mark as the continue state, update the valid query node word to the valid query node word whose node word subnode index is consistent with the character k+1, point the pointer to the character k+1, and record the node corresponding to the character k+1 as the current node, and return to step S425;

[0042] Step S427: Determine whether the node word state corresponding to the valid query node word of the current node has a terminal state. If so, change the word segmentation state C corresponding to the character k to jMark as the word segmentation point, point the pointer to character k+1, clear and reset the recorded current node and valid query node words, and return to S423; otherwise, set the word segmentation state C corresponding to character k j Mark as unknown concept; point the pointer to character k+1, clear and reset the recorded current node and valid query node words, and return to S423;

[0043] Step S428: Search all valid query node words of the current node corresponding to the pointer pointing to character k, and determine whether the node word states corresponding to all valid query node words have a terminal state. If so, change the word segmentation state C corresponding to character k to j Mark as a word segmentation point, clear and reset the pointer, the recorded current node and the valid query node word, and go to S429; otherwise, the word segmentation state C corresponding to the character k is j Mark as unknown concept; clear and reset the pointer, the recorded current node and the valid query node word, and go to S429;

[0044] Step S429: Segment S according to the segmentation status i Perform word segmentation. If the word segmentation state is a plurality of consecutive continuation states and a word-breaking point state, combine the characters corresponding to the plurality of consecutive continuation states and the word-breaking point state to form a subdivided sub-word, which is marked as a known concept deterministic word segmentation. If the word segmentation state is an unknown concept state, mark the characters corresponding to the unknown concept state as an unknown concept deterministic word segmentation. If the word segmentation state is a plurality of consecutive continuation states and an unknown concept state, combine the characters corresponding to the plurality of consecutive continuation states and the unknown concept state to form a subdivided sub-word, which is marked as an uncertain word segmentation. Go to step S422.

[0045] Preferably, the word order weight refers to the weight assigned according to the order in which the words appear in the telemetry, and the weight decays exponentially. The calculation formula for exponential decay is:

[0046] W=e -aX

[0047] Where a is the word order importance attenuation factor, and X is the order position of the word in the telemetry;

[0048] For a telemetry vocabulary of a spacecraft test identification and evaluation, the segmented words formed after word segmentation preprocessing and word segmentation are used to calculate the normalized word weight according to the word weight, word class weight, keyword adjustment weight and word order weight of each segmented word. The normalized word weight calculation method is as follows

[0049]

[0050] The word weight, word class weight, keyword adjustment weight and word order weight are calculated for each subdivided word index, and then the normalized word weight is obtained; the total number of subdivided words is the total number of subdivided words formed after word segmentation preprocessing and word segmentation in the telemetry vocabulary of the spacecraft test identification and evaluation;

[0051] Calculate the normalized similar word weights for each word category separately:

[0052]

[0053] The word weight, word class weight, keyword adjustment weight and word order weight are calculated for each word subdivision index of the same word class, and then the normalized word weight of the same word class is obtained, where the total number of words of the same word class is the number of words of the same word class in the telemetry vocabulary of the spacecraft test identification and evaluation;

[0054] Define confidence, calculated as follows

[0055]

[0056] The normalized word weight, normalized similar word weight and confidence are combined to obtain a 12-dimensional whole word weight vector.

[0057] A device for constructing a dictionary for spacecraft test identification and evaluation, characterized in that the device comprises:

[0058] A concept space construction module: configured to create a spacecraft telemetry concept space, wherein the spacecraft telemetry concept space includes a plurality of telemetry concepts, each of which has five attributes, namely, a number, a phrase, an alias, a weight, and a category; the number indicates a sequence number of the telemetry concept, the phrase stores the name of the telemetry concept itself, the alias indicates an alias of the telemetry concept, the category attribute is used to indicate a category to which the telemetry concept belongs, and the weight is used to indicate the importance of the category to which the telemetry concept belongs when semantically classified;

[0059] Dictionary structure building module: configured to define a spacecraft experiment identification dictionary structure, the dictionary stores multiple nodes, each node represents a Chinese character, English character or symbol, the dictionary is a multi-layer dictionary, each layer stores multiple nodes, based on the position of the word or character in a certain word, the number of layers in which it is stored in the dictionary is determined; each node has a pair of keys and values, wherein the key is used to represent the Chinese character, English character or symbol, the value is used to represent the attribute corresponding to the key, and the connection relationship of the node in the dictionary, the connection relationship is used to indicate the pre-order node and post-order node of the node, the pre-order node is the node located before the node in a certain word, and the post-order node is the node located after the node in a certain word;

[0060] Dictionary generation module: configured to insert the phrases corresponding to each telemetry concept in the spacecraft telemetry concept space into the dictionary to generate a spacecraft experiment identification dictionary.

[0061] A device for classification of spacecraft test identification and evaluation, characterized in that the device is based on the spacecraft test identification dictionary as described above, and comprises:

[0062] Preprocessing module: configured to obtain telemetry vocabulary of spacecraft test identification and evaluation, and perform word segmentation preprocessing on the telemetry vocabulary, wherein the word segmentation preprocessing method is to segment the hyphens of the telemetry vocabulary of the spacecraft test identification and evaluation, and remove invalid characters and interference characters;

[0063] A segmentation module: configured to segment the spacecraft test identification evaluation telemetry vocabulary obtained after word segmentation preprocessing into words based on the spacecraft test identification dictionary, and use the segmented words as node words;

[0064] Node word vectorization module: configured to define word weight, word category weight, keyword adjustment weight and word order weight, and vectorize each node word;

[0065] The word weight calculates the root mean square of the sum of the character codes in the word; the word category weight refers to the weight of each category in the concept space, which indicates the relative importance of the category to which the telemetry concept corresponding to the word belongs; the keyword adjustment weight is the adjustment weight added on the basis of the word category weight, which indicates further adjustment of the word category weight, and is divided into keywords, secondary words and invalid words; the word sequence weight refers to the weight assigned according to the order in which the words appear in the telemetry, and decays according to an exponential; the normalized word weight is defined according to the word weight, word category weight, keyword adjustment weight and word sequence weight, and the normalized similar word weight and the total normalized word weight under each concept category are calculated respectively; the word confidence is calculated, which indicates the percentage of known concept participles in the word to the total number of word parts; the normalized word weight, various normalized similar word weights and confidence are combined to obtain a 12-dimensional whole word weight vector.

[0066] Classification module: It is configured to take the 12-dimensional whole word weight vector corresponding to each node word as input and use hierarchical clustering to classify all node words.

[0067] The present invention proposes a technology for word segmentation and classification of spacecraft telemetry names. By designing a spacecraft test and identification application concept domain, word order weight and normalized word class weight are used to improve the effectiveness of feature extraction; word segmentation accuracy and efficiency are improved by designing a network concept dictionary containing matrix nodes, using word segmentation preprocessing, and improving the word segmentation algorithm. Finally, the spacecraft telemetry names are automatically clustered according to their meanings. The content of the present invention lays a technical foundation for realizing the automated processing of spacecraft test identification and evaluation.

[0068] Beneficial effects:

[0069] (1) Fill the gap in methods for automatically extracting relevant information from various types of spacecraft and test identification design documents during the test identification and evaluation process.

[0070] (2) Design and implement a dictionary and word segmentation algorithm based on the unique structure of spacecraft telemetry names to improve the accuracy and efficiency of word segmentation;

[0071] (3) Transition from manual identification and entry of spacecraft test identification information to automated identification and classification of content to improve the efficiency of test identification and evaluation work. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 Schematic flow diagram of the methodology for building a lexicon for spacecraft test qualification assessment;

[0073] Figure 2 This is a diagram of the dictionary structure;

[0074] Figure 3 Schematic flow diagram of the methodology for the classification of spacecraft test qualification assessments;

[0075] Figure 4 Schematic diagram of the method for dictionary construction and classification for spacecraft test identification and evaluation. DETAILED DESCRIPTION

[0076] The present invention is described in detail below in conjunction with the accompanying drawings and embodiments.

[0077] The present invention proposes a method for constructing a dictionary for spacecraft test identification and evaluation, such as Figure 1 As shown, the method comprises the following steps:

[0078] Step S1: Create a spacecraft telemetry concept space, the spacecraft telemetry concept space includes multiple telemetry concepts, each telemetry concept has five attributes, namely, number, phrase, alias, weight, and category; the number represents the sequence number of the telemetry concept, the phrase stores the name of the telemetry concept itself, the alias represents the alias of the telemetry concept, the category attribute is used to represent the category to which the telemetry concept belongs, and the weight is used to represent the importance of the category to which the telemetry concept belongs when classified by semantics;

[0079] Step S2: define the structure of the spacecraft experiment identification dictionary, the dictionary stores a plurality of nodes, each node represents a Chinese character, English character or symbol, the dictionary is a multi-layer dictionary, each layer stores a plurality of nodes, based on the position of the character or the character in a certain word, the number of layers in which it is stored in the dictionary is determined; each node has a pair of key and value, wherein the key is used to represent the Chinese character, English character or symbol, the value is used to represent the attribute corresponding to the key, and the connection relationship of the node in the dictionary, the connection relationship is used to indicate the pre-order node and post-order node of the node, the pre-order node is the node located before the node in a certain word, and the post-order node is the node located after the node in a certain word;

[0080] Step S3: inserting the phrases corresponding to each telemetry concept in the spacecraft telemetry concept space into the dictionary to construct a spacecraft experiment identification dictionary.

[0081] The step S1 describes the specific spacecraft telemetry concept by defining five attributes, namely, number, phrase, alias, weight, and category. The attribute values ​​corresponding to the category attribute include system class, entity, dynamics, polarity, unit, abstraction, state, adjunct, preposition, and unknown concept, as shown in Table 1.

[0082] serial number word Other Names Weight category 1 sensor Sensor XXX entity 2 Magnetic torquer MT MTP XXX entity 3 Base XXX entity 4 Earth XXX entity 5 Earth Sensor ES Dimin XXX entity

[0083] Table 1

[0084] The system class is used to indicate the concept category of the composition of the spacecraft subsystem (such as "control", "payload", "integrated electronics"); the entity is used to indicate the concept category of specific equipment or single machine (such as "gyroscope", "reaction wheel", "transponder"); dynamics is used to indicate the concept category related to dynamics or physical quantity (such as "yaw angle", "voltage", "orbital coordinate system"); polarity is used to indicate the concept category related to direction, position, positive and negative, and order (such as "1", "+", "south", "C"); unit is used to indicate the concept category related to the unit of physical quantity (such as "V", "A"); state is used to indicate the concept category related to Concept categories related to the state of the object (such as "open", "locked", "overflow"); abstract is used to indicate concept categories that have certain logical meanings but need to be combined with category words such as entity, dynamics, and state to form specific meanings (such as "flag", "constant", and "current"); dependent words are used to indicate concept categories that will not affect the understanding of the telemetry meaning after removal (such as "reference", "operation", and "work"); prepositions are used to indicate words, words, or characters used to maintain grammatical correctness (such as "in" and "from"); unknown concepts are used to indicate words, words, or characters that do not belong to the above categories and have never appeared in the vocabulary list or dictionary.

[0085] In this embodiment, a spacecraft telemetry concept space is formed by constructing spacecraft telemetry concepts one by one.

[0086] In step S2, in this embodiment, the spacecraft experiment identification dictionary structure defined is a multi-layer mesh dictionary structure, with multiple nodes in the dictionary. The nodes are described by keys and values. The keys correspond to words, characters or symbols, and the values ​​can represent the relationship attributes between the node and the node word, where the node word is a vocabulary in the telemetry concept space, that is, a vocabulary formed by connecting the nodes in the dictionary in sequence. The connection order of the nodes is consistent with the connection order of the words or characters in the vocabulary, so nodes with the same key can be in different layers in the dictionary. Figure 2 As shown in the figure, each word, character or symbol constitutes a node, and the number of layers of a node in the dictionary is determined by its position in the vocabulary. The structure of the value is defined as shown in Table 2, which specifically describes the sequence number of the node corresponding to the word, character or symbol, the connected previous node (parent node) and the next node (child node).

[0087] As shown in Table 2, the value of the node is composed of a node index, several node word indexes, a node word state corresponding to each node word index, a node word parent node index corresponding to each node word index, and a node word child node index corresponding to each node word index.

[0088] Among them, the node index represents the number of the node in the current layer of the dictionary. In the same layer, the number increases in the order of insertion. In different layers, the same key can correspond to different node indexes. The node word corresponding to the node refers to all phrases in the spacecraft telemetry concept space, including the key corresponding to the node, and all phrases whose positions in the phrases are the same as the number of the current layer. The node word index represents the number of the node word corresponding to the node. Each node word has a unique number, and the number is a global quantity, which has nothing to do with the number of layers of the node corresponding to the node word. A node can correspond to several node words, and each node word has a number. Therefore, each node can correspond to several node word indexes. The node word state indicates the position information of the node in the node word, which is specifically divided into the continuing state (represented as 0) and the terminal state (represented as 1). The continuing state indicates that the node is not at the last position of the node word, and the terminal state indicates that the node is at the last position of the node word. The node word parent node index indicates the node index of the previous node corresponding to the node in the node word in the previous layer of the current layer, and the node word child node index indicates the node index of the subsequent node corresponding to the node in the node word in the next layer of the current layer. If the node word corresponding to the node does not have a parent node, the parent node index is 0. If the node word corresponding to the node does not have a child node, the node word child node index is 0.

[0089]

[0090]

[0091] Table 2

[0092] For example Figure 2 As shown in [example], the first characters of both "spacecraft" and "heading angle" are "hang" (航 in Chinese). Therefore, "hang" is located at a certain node in the first layer, which contains two node words: "spacecraft" and "heading angle". In the node word attributes of the "hang" node, the sub-node index corresponding to the node word "spaceflight personnel" under the "hang" node points to the "tian" (天) node in the second layer of the dictionary, while the sub-node index corresponding to the node word "heading angle" under the "hang" node points to the "xiang" (向) character node in the second layer. On the other hand, the second character of the node word "direction angle" is also the "xiang" node. Therefore, in addition to "heading angle", the "xiang" character node also includes the node word "direction angle", and its corresponding parent node index points to the "fang" (方) character node.

[0093] The multi-layer network dictionary structure makes the vocabulary architecture clearer, saves space occupancy, and greatly simplifies the design and implementation of the word insertion algorithm.

[0094] The said step S3 includes the following steps

[0095] Step S31: If all the telemetry concepts in the spacecraft telemetry concept space have been inserted into the dictionary, the method ends; otherwise, go to step S32;

[0096] Step S32: Obtain a telemetry concept that has not been inserted into the dictionary in the spacecraft telemetry concept space, obtain the corresponding phrase of this telemetry concept, use this phrase as a node word, and set a node word index ii for this node word;

[0097] Step S33: Read an unprocessed character code from the said node word in the order from left to right, and obtain the position num of this character in the node word; if this character code is not the last character of the node word, go to step S34; if this character code is the last character of the node word, go to step S35;

[0098] Step S34: If the unprocessed character code is not included in the node of the numth layer of the dictionary, a node is inserted in the numth layer of the dictionary, the key of the node is set to the unprocessed character code, and the node word index ii is added to the node word index of the node; if the unprocessed character code is included in the node corresponding to the numth layer of the dictionary, the node word index ii is added to the node word index of the node corresponding to the character code; the node word state corresponding to the node word index ii is marked as the continue state; the node word parent node index corresponding to the node word index ii is marked as the preceding character of the character code in the node word in the dictionary The node index corresponding to the num-1 layer; the node word child node index corresponding to the node word index ii is marked as the node index corresponding to the subsequent character of the character code in the node word in the num+1th layer of the dictionary; wherein, if the preceding character of the character code is empty, the node word parent node index corresponding to the node word index ii is marked as 0; if the num+1th layer does not contain the subsequent character of the character code of the node word, the node word child node index corresponding to the node word index ii is marked as the current total number of nodes in the num+1th layer of the dictionary + 1; the unprocessed character code is marked as processed, and return to step S33;

[0099] Step S35: If the node at the numth layer of the dictionary does not contain the unprocessed character code, then insert a node at the numth layer of the dictionary, set the key of the node to the unprocessed character code, and add the node word index ii to the node word index of the node; if the node at the numth layer of the dictionary contains the unprocessed character code, then add the node word index ii to the node word index of the node corresponding to the character code; the node word state corresponding to the node word index ii is marked as the terminal state, the node word parent node index corresponding to the node word index ii is marked as the node index corresponding to the preceding character of the character code in the node word in the num-1th layer of the dictionary, and the node word child node index corresponding to the node word index ii is marked as 0; return to step S31.

[0100] In this embodiment, the continuing state indicates that the current node is not located at the last position of the node word, and the ending state indicates that the current node is located at the last position of the node word.

[0101] The present invention proposes a classification method for spacecraft test identification and evaluation, which is based on the spacecraft test identification dictionary as described above. Figure 3-4 As shown, the method comprises the following steps:

[0102] Step S41: Acquire telemetry vocabulary for spacecraft test, identification and evaluation, and perform word segmentation preprocessing on the telemetry vocabulary. The word segmentation preprocessing method is to segment the hyphens of the telemetry vocabulary for spacecraft test, identification and evaluation, and remove invalid characters and interference characters.

[0103] Step S42: based on the spacecraft experiment identification dictionary, segmenting the spacecraft test identification evaluation telemetry vocabulary obtained after word segmentation preprocessing into words, and using the segmented words as node words;

[0104] Step S43: define word weight, word category weight, keyword adjustment weight and word order weight, and vectorize each node word;

[0105] The word weight calculates the root mean square of the sum of the character codes in the word; the word category weight refers to the weight of each category in the concept space, which indicates the relative importance of the category to which the telemetry concept corresponding to the word belongs; the keyword adjustment weight is the adjustment weight added on the basis of the word category weight, which indicates further adjustment of the word category weight, and is divided into keywords, secondary words and invalid words; the word sequence weight refers to the weight assigned according to the order in which the words appear in the telemetry, and decays according to an exponential; the normalized word weight is defined according to the word weight, word category weight, keyword adjustment weight and word sequence weight, and the normalized similar word weight and the total normalized word weight under each concept category are calculated respectively; the word confidence is calculated, which indicates the percentage of known concept participles in the word to the total number of word parts; the normalized word weight, various normalized similar word weights and confidence are combined to obtain a 12-dimensional whole word weight vector.

[0106] Step S44: using the 12-dimensional whole word weight vector corresponding to each node word as input, hierarchical clustering is used to classify all node words.

[0107] In step S41, some spacecraft telemetry uses hyphens to separate different concept parts, which need to be pre-segmented to improve the accuracy and efficiency of word segmentation. The words to be segmented are pre-segmented to remove invalid characters and interference characters. For spacecraft telemetry and remote control vocabulary, invalid characters refer to the following:

[0108] [`~~! @#$%^&*()()—_={}[]【】| / \\、::;;''""“”,,。 <>《》? ? ]

[0109] Furthermore, in this embodiment, a replacement method is used to remove redundant characters. The strategy for removing interfering characters "+" and "-" from telemetry vocabulary for spacecraft test identification and evaluation is:

[0110] 1. Based on the actual meaning of telemetry vocabulary in spacecraft test evaluation, it is determined that "+" or "-" must be followed by a character or number, and will not appear at the end of a word after invalid symbols are replaced;

[0111] 2. If a number is immediately after "+" or "-", it is split before "+" and "-". At this time, "+" or "-" indicates the positive or negative number respectively;

[0112] 3. If a non-numeric character is immediately following "+" or "-" and it is not the first letter, when the preceding and succeeding characters of "+" or "-" are both English letters, it is determined that the "+" or "-" symbol represents a hyphen and the "+" or "-" symbol is deleted; when the preceding and succeeding characters of "+" or "-" exist, it is determined that the "+" or "-" symbol represents a polarity concept and the word is split before the "+" or "-" symbol.

[0113] The step S42, based on the spacecraft experiment identification dictionary, segmenting the spacecraft test identification evaluation telemetry vocabulary obtained after word segmentation preprocessing into words, and using the segmented words as node words, further includes:

[0114] Step S421: The result obtained after word segmentation preprocessing of a certain spacecraft test evaluation telemetry vocabulary is recorded as a set S, S = {S1, ..., S i ,…,S n}, where S i is the i-th subword formed after the telemetry vocabulary is preprocessed by word segmentation, 1≤i≤n, n is the total number of subwords formed after the telemetry vocabulary is segmented; define the word segmentation state C, C={C1,…,C j ,…,C m}, where C j is the state corresponding to the jth word or character of the telemetry vocabulary, and its value is 0 for the continuation state, 1 for the word break point, or 2 for the unknown concept, 1≤j≤m, and m is the total number of words or characters in the telemetry vocabulary;

[0115] Step S422: If all subwords have been segmented, the method ends; otherwise, a new subword S to be processed is selected. i , the pointer points to the subword S to be processed i The first character of ;

[0116] Step S423: If the subword to be processed S i If the word segmentation is completed, go to step S422; otherwise, obtain the segment to be processed S i The character k pointed to by the pointer in the middle, go to step S424;

[0117] Step S424: Search the first-level nodes of the dictionary to see if they contain the character k pointed to by the pointer. If there is a node that is consistent with the character k, then record the node corresponding to the character k as the current node, and record the node word corresponding to the node word index of the current node as the valid query node word, and then go to S425; if there is no node that is consistent with the character k in the first-level nodes of the dictionary, then set the segmentation state C corresponding to the character k to the valid query node word. jMarked as unknown concept, the pointer points to the current subword to be processed S i The next character of character k in the figure, if there is a next character, the step returns to S424, otherwise the step turns to S429;

[0118] Step S425: Read the current subword to be processed S i The next character of the character k pointed to by the pointer is recorded as character k+1, and the process goes to step S426; if the current subword to be processed S i If there is no next character in the sequence, go to step S428;

[0119] Step S426: Search all valid query node words of the node corresponding to character k, and determine whether the next-layer dictionary node corresponding to the node word sub-node index of all valid query node words is consistent with character k+1. If not, go to step S427; otherwise, set the word segmentation state C corresponding to character k to the same as the word segmentation state C of character k. j Mark as the continue state, update the valid query node word to the valid query node word whose node word subnode index is consistent with the character k+1, point the pointer to the character k+1, and record the node corresponding to the character k+1 as the current node, and return to step S425;

[0120] Step S427: Determine whether the node word state corresponding to the valid query node word of the current node has a terminal state. If so, change the word segmentation state C corresponding to the character k to j Mark as the word segmentation point, point the pointer to character k+1, clear and reset the recorded current node and valid query node words, and return to S423; otherwise, set the word segmentation state C corresponding to character k j Mark as unknown concept; point the pointer to character k+1, clear and reset the recorded current node and valid query node words, and return to S423;

[0121] Step S428: Search all valid query node words of the current node corresponding to the pointer pointing to character k, and determine whether the node word states corresponding to all valid query node words have a terminal state. If so, change the word segmentation state C corresponding to character k to j Mark as a word segmentation point, clear and reset the pointer, the recorded current node and the valid query node word, and go to S429; otherwise, the word segmentation state C corresponding to the character k is j Mark as unknown concept; clear and reset the pointer, the recorded current node and the valid query node word, and go to S429;

[0122] Step S429: Segment S according to the segmentation status iPerform word segmentation. If the word segmentation state is a plurality of consecutive continuation states and a word-breaking point state, combine the characters corresponding to the plurality of consecutive continuation states and the word-breaking point state to form a subdivided sub-word, which is marked as a known concept deterministic word segmentation. If the word segmentation state is an unknown concept state, mark the characters corresponding to the unknown concept state as an unknown concept deterministic word segmentation. If the word segmentation state is a plurality of consecutive continuation states and an unknown concept state, combine the characters corresponding to the plurality of consecutive continuation states and the unknown concept state to form a subdivided sub-word, which is marked as an uncertain word segmentation. Go to step S422.

[0123] In this embodiment, three separate marks are made for each character in the vocabulary to be segmented, namely, the continuation state, the segmentation point and the unknown concept. The continuation state indicates that the character is at the first character or middle character position in a node word recorded in the dictionary, the segmentation point indicates that the character is at the last character position in a node word recorded in the dictionary, and the unknown concept indicates that the character has not found a matching node elsewhere in the corresponding layer of the dictionary. The subwords composed of the continuation state and the segmentation point belong to segmentations with known concepts (known concept deterministic segmentations) and belong to node words that are clearly recorded in the dictionary; the subwords composed of the continuation state and the unknown concept belong to segmentations with unclear concepts (uncertain segmentations) and do not belong to node words recorded in the dictionary, but the first part of the characters in the word appears in the dictionary and may belong to unrecorded vocabulary; a single unknown concept belongs to a completely unknown concept segmentation (unknown concept deterministic segmentation) and does not belong to a node recorded in the first layer of the dictionary.

[0124] Taking the telemetry "ground sensitive A power status" as an example, the word is converted into "ground sensitive", "A", and "power status" after word segmentation preprocessing, a total of three sub-words, each of which corresponds to an S in the above step S42. i The word segmentation algorithm processes the three subwords separately. Before word segmentation, "ground sensitivity", "A", "power supply", and "status" have been defined by the pre-concept space and inserted into the dictionary to form node words, among which "ground sensitivity" is the 43rd node word, "A" is the 132nd node word, "power supply" is the 129th node word, and "status" is the 285th node word. In the dictionary, these concept relationships are shown in Table 3:

[0125]

[0126]

[0127] Table 3

[0128] The dictionary defines the connection relationship between nodes at each level, corresponding node words, and nodes within each node word, and the forward and backward indexes can be realized through the sequence number. For example, the first-level dictionary node No. 33 "Electricity" has 7 node words 44, 45, 92, 93, 94, 129, and 303 under it. The child node index of the node word No. 129 points to the second-level dictionary node No. 79 "Source", which also has a node word No. 129, and its parent node index points to the first-level dictionary node No. 33 "Electricity".

[0129] Under the input of the subword "power status", the search algorithm starts from the first word "electricity". Since the "source" node can be found in the second-level dictionary, the word segmentation state corresponding to the "electricity" character is recorded as the continuing state. Since the node word subnode index under the "source" node is 0, and the node word state of the "source" node No. 129 is the terminal state 1, the word segmentation state corresponding to the "source" character is recorded as the word segmentation point, and then turns to the next character "status", and starts searching again from the first level of the dictionary to obtain the word segmentation state vector as shown in Table 4.

[0130]

[0131] Table 4

[0132] Therefore, when the subword "power status" is input, two subwords "power" and "status" are formed after word segmentation.

[0133] In addition, if the dictionary records the node word "power status", the node word sub-node index of the "source" node is not 0. At this time, when the "source" node is searched, according to step S426, since the node word of the "source" node has a sub-node, the word segmentation state corresponding to the word "source" will be marked as the continued state, so that "power status" will not be subdivided. Therefore, when the dictionary records the three node words "power", "status", and "power status" at the same time, the word segmentation algorithm performs word segmentation according to the longest node word.

[0134] In step S43, the word category weights are quantified as shown in Table 5, with a total of 10 categories.

[0135] Chinese name English name Part of Speech Rights Keyword adjustment rights System Class Class 9 Dead words entity Entity 8 Keywords dynamics Dynamics 7 Secondary words polarity Distinction 6 Secondary words unit Unit 5 Dead words abstract Abstract 4 Dead words state State 3 Dead words Adjuncts Subsidiary 2 Dead words preposition Preposition 1 Dead words Unknown concept Null 0 Dead words

[0136] Table 5

[0137] The word order weight refers to the weight assigned according to the order in which the words appear in the telemetry. The weight decays exponentially. The calculation formula for exponential decay is:

[0138] W=e -aX

[0139] Where a is the word order importance attenuation factor, and X is the order of the word in the telemetry. For a telemetry vocabulary for a spacecraft test identification and evaluation, the normalized word weight is calculated based on the word weight, word class weight, keyword adjustment weight and word order weight of each word segmentation using its word segmentation preprocessing and word segmentation. The calculation method is as follows

[0140]

[0141] The total number of subdivision words is the total number of subdivision words formed after word segmentation preprocessing and word segmentation in the telemetry vocabulary of the spacecraft test identification and evaluation. The word weight, word class weight, keyword adjustment weight and word order weight are calculated for each subdivision word index to obtain the normalized word weight.

[0142] Calculate the normalized similar word weights under each word category separately, as follows

[0143]

[0144] The total number of similar subdivision words is the number of subdivision words of the same word class in the telemetry vocabulary of the spacecraft test identification and evaluation. The word weight, word class weight, keyword adjustment weight and word order weight are calculated for each subdivision word index of the same word class to obtain the normalized similar word weight.

[0145] Define confidence, calculated as follows

[0146]

[0147] The normalized word weight, 10 categories of normalized similar word weights and confidence are combined to obtain a 12-dimensional whole word weight vector.

[0148] In the step S44, in this embodiment, the whole word weight vector corresponding to each word is taken as input, and the whole word weight vector is classified using a hierarchical clustering analysis method based on distance measurement, and the classification result of each telemetry word is obtained according to the corresponding relationship between the word and the whole word weight vector.

[0149] The following is a test example of the name of the analog telemetry of a navigation satellite control subsystem.

[0150]

[0151] The word segmentation test results are as follows. After running the algorithm, each complete telemetry name is segmented into different sub-words, and the word segmentation effect is good, without any mis-segmentation.

[0152]

[0153] After vectorizing the word segmentation results, the classification results obtained using hierarchical clustering analysis are as follows. The classification number represents the telemetry classified into one category. Each telemetry is effectively clustered, and there is no mixed classification or misclassification. For example, although both contain the "GX" subword, "GX power status" and "GX motor current" belong to different categories. At the same time, "GX power status" and "ground sensitive A power status" both contain the "power status" subword, but belong to different categories.

[0154]

[0155]

[0156] The present invention also provides a device for constructing a dictionary for spacecraft test identification and evaluation, the device comprising:

[0157] A concept space construction module: configured to create a spacecraft telemetry concept space, wherein the spacecraft telemetry concept space includes a plurality of telemetry concepts, each of which has five attributes, namely, a number, a phrase, an alias, a weight, and a category; the number indicates a sequence number of the telemetry concept, the phrase stores the name of the telemetry concept itself, the alias indicates an alias of the telemetry concept, the category attribute is used to indicate a category to which the telemetry concept belongs, and the weight is used to indicate the importance of the category to which the telemetry concept belongs when semantically classified;

[0158] Dictionary structure building module: configured to define a spacecraft experiment identification dictionary structure, the dictionary stores multiple nodes, each node represents a Chinese character, English character or symbol, the dictionary is a multi-layer dictionary, each layer stores multiple nodes, based on the position of the word or character in a certain word, the number of layers in which it is stored in the dictionary is determined; each node has a pair of keys and values, wherein the key is used to represent the Chinese character, English character or symbol, the value is used to represent the attribute corresponding to the key, and the connection relationship of the node in the dictionary, the connection relationship is used to indicate the pre-order node and post-order node of the node, the pre-order node is the node located before the node in a certain word, and the post-order node is the node located after the node in a certain word;

[0159] Dictionary generation module: configured to insert the phrases corresponding to each telemetry concept in the spacecraft telemetry concept space into the dictionary to generate a spacecraft experiment identification dictionary.

[0160] The present invention also provides a device for classification of spacecraft test identification and evaluation, the device is based on the spacecraft test identification dictionary as described above, and the device comprises:

[0161] Preprocessing module: configured to obtain telemetry vocabulary of spacecraft test identification and evaluation, and perform word segmentation preprocessing on the telemetry vocabulary, wherein the word segmentation preprocessing method is to segment the hyphens of the telemetry vocabulary of the spacecraft test identification and evaluation, and remove invalid characters and interference characters;

[0162] A segmentation module: configured to segment the spacecraft test identification evaluation telemetry vocabulary obtained after word segmentation preprocessing into words based on the spacecraft test identification dictionary, and use the segmented words as node words;

[0163] Node word vectorization module: configured to define word weight, word category weight, keyword adjustment weight and word order weight, and vectorize each node word;

[0164] The word weight calculates the root mean square of the sum of the character codes in the word; the word category weight refers to the weight of each category in the concept space, which indicates the relative importance of the category to which the telemetry concept corresponding to the word belongs; the keyword adjustment weight is the adjustment weight added on the basis of the word category weight, which indicates further adjustment of the word category weight, and is divided into keywords, secondary words and invalid words; the word sequence weight refers to the weight assigned according to the order in which the words appear in the telemetry, and decays according to an exponential; the normalized word weight is defined according to the word weight, word category weight, keyword adjustment weight and word sequence weight, and the normalized similar word weight and the total normalized word weight under each concept category are calculated respectively; the word confidence is calculated, which indicates the percentage of known concept participles in the word to the total number of word parts; the normalized word weight, various normalized similar word weights and confidence are combined to obtain a 12-dimensional whole word weight vector.

[0165] Classification module: It is configured to take the 12-dimensional whole word weight vector corresponding to each node word as input and use hierarchical clustering to classify all node words.

[0166] The above specific embodiments only describe the design principle of the present invention. The shapes and names of the components in the description may be different and are not limited. Therefore, those skilled in the art in the field of the present invention may modify or replace the technical solutions recorded in the above embodiments; and these modifications and replacements do not deviate from the creative purpose and technical solutions of the present invention and should all fall within the protection scope of the present invention.

Claims

1. A method for constructing a dictionary for spacecraft test evaluation, characterized in that: The method comprises the following steps: Step S1: Create a spacecraft telemetry concept space, the spacecraft telemetry concept space includes multiple telemetry concepts, each telemetry concept has five attributes, namely, number, phrase, alias, weight, and category; the number represents the sequence number of the telemetry concept, the phrase stores the name of the telemetry concept itself, the alias represents the alias of the telemetry concept, the category attribute is used to represent the category to which the telemetry concept belongs, and the weight is used to represent the importance of the category to which the telemetry concept belongs when classified by semantics; Step S2: define the structure of a spacecraft test and identification dictionary, wherein the dictionary stores a plurality of nodes, each node represents a Chinese character, English character or symbol, and the dictionary is a multi-layer dictionary, wherein each layer stores a plurality of nodes, and the number of layers in which the character or the symbol is stored in the dictionary is determined based on the position of the character or the symbol in a certain word; each node has a pair of keys and values, wherein the key is used to represent the Chinese character, English character or symbol, and the value is used to represent the attribute corresponding to the key, and the connection relationship of the node in the dictionary, and the connection relationship is used to indicate the pre-order node and post-order node of the node, wherein the pre-order node is the node located before the node in a certain word, and the post-order node is the node located after the node in a certain word; Step S3: inserting the phrases corresponding to each telemetry concept in the spacecraft telemetry concept space into the dictionary to construct a spacecraft test identification dictionary; In step S2, the spacecraft test and identification dictionary structure is a multi-layer mesh dictionary structure, wherein the dictionary has a plurality of nodes, each node is described by a key and a value, wherein the key corresponds to a word, a character or a symbol, and the value can represent the relationship attribute between the node and the node word; the node value is composed of a node index, a plurality of node word indexes, a node word state corresponding to each node word index, a node word parent node index corresponding to each node word index, and a node word child node index corresponding to each node word index; The node index represents the number of the node in the current layer of the dictionary. In the same layer, the number increases in the order of insertion. In different layers, the same key can correspond to different node indexes. The node words corresponding to the node refer to all the phrases in the spacecraft telemetry concept space, including the key corresponding to the node, and all the phrases whose positions in the phrases are the same as the level of the current level; The node word index indicates the number of the node word corresponding to the node. Each node word has a unique number, and the number is a global quantity, which has nothing to do with the layer number of the node corresponding to the node word. A node can correspond to several node words, each node word has a number, and each node can correspond to several node word indexes. The node word state indicates the position information of the node in the node word, which is specifically divided into a continuing state marked as 0 and a terminal state marked as 1. The continuing state indicates that the node is not at the last position of the node word, and the terminal state indicates that the node is at the last position of the node word. The parent node index of a node word indicates the node index of the previous node in the previous layer of the current layer corresponding to the node in the node word; The node word subnode index represents the node index of the subsequent node corresponding to the node in the node word in the next layer of the current layer; If the node word corresponding to the node does not have a parent node, the parent node is indexed as 0. If the node word corresponding to the node does not have a child node, the node word child node is indexed as 0. The step S3 comprises the following steps: Step S31: If all telemetry concepts in the spacecraft telemetry concept space have been inserted into the dictionary, the method ends; otherwise, proceed to step S32; Step S32: obtaining a telemetry concept in the spacecraft telemetry concept space that has not been inserted into the dictionary, obtaining a phrase corresponding to the telemetry concept, using the phrase as a node word, and setting a node word index ii for the node word; Step S33: read an unprocessed character code from the node word in order from left to right, and obtain the position num of the character in the node word; if the character code is not the last character of the node word, proceed to step S34; if the character code is the last character of the node word, proceed to step S35; Step S34: If the unprocessed character code is not included in the nodes of the numth layer of the dictionary, a node is inserted in the numth layer of the dictionary, the key of the node is set to the unprocessed character code, and a node word index ii is added to the node word index of the node; if the unprocessed character code is included in the nodes of the numth layer of the dictionary, a node word index ii is added to the node word index of the node corresponding to the character code; the node word state corresponding to the node word index ii is marked as the continue state; the node word parent node index corresponding to the node word index ii is marked as the preceding character of the character code in the node word in the dictionary The node index corresponding to the num-1 layer; the node word child node index corresponding to the node word index ii is marked as the node index corresponding to the subsequent character of the character code in the node word in the num+1th layer of the dictionary; wherein, if the preceding character of the character code is empty, the node word parent node index corresponding to the node word index ii is marked as 0; if the num+1th layer does not contain the subsequent character of the character code of the node word, the node word child node index corresponding to the node word index ii is marked as the current total number of nodes in the num+1th layer of the dictionary + 1; the unprocessed character code is marked as processed, and return to step S33; Step S35: If the unprocessed character code is not included in the node of the numth layer of the dictionary, a node is inserted into the numth layer of the dictionary, the key of the node is set to the unprocessed character code, and the node word index ii is added to the node word index of the node; if the unprocessed character code is included in the node of the numth layer of the dictionary, the node word index ii is added to the node word index of the node corresponding to the character code; the node word state corresponding to the node word index ii is marked as the terminal state, the node word parent node index corresponding to the node word index ii is marked as the node index corresponding to the preceding character of the character code in the node word in the num-1th layer of the dictionary, and the node word child node index corresponding to the node word index ii is marked as 0; return to step S31.

2. A classification method for spacecraft test evaluation, characterized in that: The method is based on the spacecraft test and evaluation dictionary constructed by the method for constructing a dictionary for spacecraft test and evaluation as claimed in claim 1, and the method comprises the following steps: Step S41: Acquire telemetry vocabulary for spacecraft test, identification and evaluation, and perform word segmentation preprocessing on the telemetry vocabulary. The word segmentation preprocessing method is to segment the hyphens of the telemetry vocabulary for spacecraft test, identification and evaluation, and remove invalid characters and interference characters. Step S42: based on the spacecraft test and identification dictionary, segment the spacecraft test and identification evaluation telemetry vocabulary obtained after word segmentation preprocessing into words, and use the segmented words as node words; Step S43: define word weight, word category weight, keyword adjustment weight and word order weight, and vectorize each node word; The word weight is the root mean square of the sum of the character codes in the word; the word category weight refers to the weight of each category in the concept space, indicating the relative importance of the category to which the telemetry concept corresponding to the word belongs; the keyword adjustment weight is the adjustment weight added on the basis of the word category weight, indicating further adjustment of the word category weight, which is divided into keywords, secondary words and invalid words; the word order weight refers to the weight assigned according to the order in which the word appears in the telemetry, and decays according to the exponential; the normalized word weight is defined according to the word weight, word category weight, keyword adjustment weight and word order weight, and the normalized similar word weight and the total normalized word weight under each concept category are calculated respectively; Calculate the word confidence, which represents the percentage of known concept segmentations in the word to the total number of segmentations of the word; combine the normalized word weight, various normalized similar word weights and confidence to obtain a 12-dimensional whole word weight vector; Step S44: using the 12-dimensional whole word weight vector corresponding to each node word as input, hierarchical clustering is used to classify all node words.

3. The method according to claim 2, characterized in that The step S42 comprises: Step S421: The result obtained after word segmentation preprocessing of a certain spacecraft test evaluation telemetry vocabulary is recorded as a set S, S = {S1, ..., S i ,...,S n }, where S i is the i-th subword formed after the telemetry vocabulary is preprocessed by word segmentation, 1≤i≤n, n is the total number of subwords formed after the telemetry vocabulary is segmented; define the word segmentation state C, C={C1,…,C j ,...,C m }, where C j is the state corresponding to the jth word or character of the telemetry vocabulary, and its value is 0 for the continuation state, 1 for the word break point, or 2 for the unknown concept, 1≤j≤m, and m is the total number of words or characters in the telemetry vocabulary; Step S422: If all subwords have been segmented, the method ends; otherwise, a new subword S to be processed is selected. i , the pointer points to the subword S to be processed i The first character of ; Step S423: If the subword to be processed S i If the word segmentation is completed, go to step S422; otherwise, obtain the segment to be processed S i The character k pointed to by the pointer in the middle, go to step S424; Step S424: Search the first-level nodes of the dictionary to see if they contain the character k pointed to by the pointer. If there is a node that is consistent with the character k, then record the node corresponding to the character k as the current node, and record the node word corresponding to the node word index of the current node as the valid query node word, and then go to S425; if there is no node that is consistent with the character k in the first-level nodes of the dictionary, then set the segmentation state C corresponding to the character k to the valid query node word. j Marked as unknown concept, the pointer points to the current subword to be processed S i The next character of character k in the figure, if there is a next character, the step returns to S424, otherwise the step turns to S429; Step S425: Read the current subword to be processed S i The next character of the character k pointed to by the pointer is recorded as character k+1, and the process goes to step S426; if the current subword to be processed S i If there is no next character in the sequence, go to step S428; Step S426: Search all valid query node words of the node corresponding to character k, and determine whether the next-layer dictionary node corresponding to the node word sub-node index of all valid query node words is consistent with character k+1. If not, go to step S427; otherwise, set the word segmentation state C corresponding to character k to the same as the word segmentation state C of character k. j Mark as the continue state, update the valid query node word to the valid query node word whose node word subnode index is consistent with the character k+1, point the pointer to the character k+1, and record the node corresponding to the character k+1 as the current node, and return to step S425; Step S427: Determine whether the node word state corresponding to the valid query node word of the current node has a terminal state. If so, change the word segmentation state C corresponding to the character k to j Mark as a word segmentation point, point the pointer to character k+1, clear and reset the recorded current node and valid query node words, and return to S423; otherwise, set the word segmentation state C corresponding to character k j Mark as unknown concept; point the pointer to character k+1, clear and reset the recorded current node and valid query node words, and return to S423; Step S428: Search all valid query node words of the current node corresponding to the pointer pointing to character k, and determine whether the node word states corresponding to all valid query node words have a terminal state. If so, change the word segmentation state C corresponding to character k to j Mark as a word segmentation point, clear and reset the pointer, the recorded current node and the valid query node word, and go to S429; otherwise, the word segmentation state C corresponding to the character k is j Mark as unknown concept; clear and reset the pointer, the recorded current node and the valid query node word, and go to S429; Step S429: Segment S according to the segmentation status i Perform word segmentation. If the word segmentation state is a plurality of consecutive continuation states and a word segmentation point state, combine the characters corresponding to the plurality of consecutive continuation states and the word segmentation point state to form a subdivided sub-word, which is marked as a known concept deterministic word segmentation. If the word segmentation state is an unknown concept state, mark the characters corresponding to the unknown concept state as an unknown concept deterministic word segmentation. If the word segmentation state is a plurality of consecutive continuation states and an unknown concept state, combine the characters corresponding to the plurality of consecutive continuation states and the unknown concept state to form a subdivided sub-word, which is marked as an uncertain word segmentation. Go to step S422.

4. The method according to claim 3, characterized in that Step S43, the word order weight refers to the weight assigned according to the order in which the words appear in the telemetry, and the weight decays exponentially. The calculation formula for exponential decay is: W=e -aX Where a is the word order importance attenuation factor, and X is the order position of the word in the telemetry; For a telemetry vocabulary of a spacecraft test identification and evaluation, the segmented words formed after word segmentation preprocessing and word segmentation are used to calculate the normalized word weight according to the word weight, word class weight, keyword adjustment weight and word order weight of each segmented word. The normalized word weight calculation method is as follows The word weight, word class weight, keyword adjustment weight and word order weight are calculated for each subdivided word index, and then the normalized word weight is obtained; the total number of subdivided words is the total number of subdivided words formed after word segmentation preprocessing and word segmentation in the telemetry vocabulary of the spacecraft test identification and evaluation; Calculate the normalized similar word weights for each word category separately: The word weight, word class weight, keyword adjustment weight and word order weight are calculated for each word subdivision index of the same word class, and then the normalized word weight of the same word class is obtained, where the total number of words of the same word class is the number of words of the same word class in the telemetry vocabulary of the spacecraft test identification and evaluation; Define confidence, calculated as follows The normalized word weight, normalized similar word weight and confidence are combined to obtain a 12-dimensional whole word weight vector.

5. A classification device for spacecraft test evaluation, characterized in that: The device constructs a spacecraft test identification dictionary based on a device for constructing a dictionary for spacecraft test identification and evaluation; wherein the device for constructing a dictionary for spacecraft test identification and evaluation comprises: A concept space construction module: configured to create a spacecraft telemetry concept space, wherein the spacecraft telemetry concept space includes a plurality of telemetry concepts, each of which has five attributes, namely, a number, a phrase, an alias, a weight, and a category; the number indicates a sequence number of the telemetry concept, the phrase stores the name of the telemetry concept itself, the alias indicates an alias of the telemetry concept, the category attribute is used to indicate a category to which the telemetry concept belongs, and the weight is used to indicate the importance of the category to which the telemetry concept belongs when semantically classified; Dictionary structure building module: configured to define a spacecraft test and identification dictionary structure, the dictionary stores multiple nodes, each node represents a Chinese character, English character or symbol, the dictionary is a multi-layer dictionary, each layer stores multiple nodes, based on the position of the word or character in a certain word, the number of layers in which it is stored in the dictionary is determined; each node has a pair of keys and values, wherein the key is used to represent the Chinese character, English character or symbol, the value is used to represent the attribute corresponding to the key, and the connection relationship of the node in the dictionary, the connection relationship is used to indicate the pre-order node and post-order node of the node, the pre-order node is the node located before the node in a certain word, and the post-order node is the node located after the node in a certain word; A dictionary generation module: configured to insert the phrases corresponding to each telemetry concept in the spacecraft telemetry concept space into the dictionary to construct a spacecraft test identification dictionary; The device for classification of spacecraft test identification and evaluation includes: Preprocessing module: configured to obtain telemetry vocabulary of spacecraft test identification and evaluation, and perform word segmentation preprocessing on the telemetry vocabulary, wherein the word segmentation preprocessing method is to segment the hyphens of the telemetry vocabulary of the spacecraft test identification and evaluation, and remove invalid characters and interference characters; A segmentation module: configured to segment the spacecraft test identification evaluation telemetry vocabulary obtained after word segmentation preprocessing into words based on the spacecraft test identification dictionary, and use the segmented words as node words; Node word vectorization module: configured to define word weight, word category weight, keyword adjustment weight and word order weight, and vectorize each node word; The word weight is calculated by the root mean square of the sum of the character codes in the word; the word category weight refers to the weight of each category in the concept space, indicating the relative importance of the category to which the telemetry concept corresponding to the word belongs; the keyword adjustment weight is the adjustment weight added on the basis of the word category weight, indicating further adjustment of the word category weight, which is divided into keywords, secondary words and invalid words; the word order weight refers to the weight assigned according to the order in which the words appear in the telemetry, and decays according to the exponential; the normalized word weight is defined according to the word weight, word category weight, keyword adjustment weight and word order weight, and the normalized similar word weight and the total normalized word weight under each concept category are calculated respectively; the word confidence is calculated, indicating the percentage of known concept participles in the word to the total number of word parts; the normalized word weight, various normalized similar word weights and confidence are combined to obtain a 12-dimensional whole word weight vector; Classification module: It is configured to take the 12-dimensional whole word weight vector corresponding to each node word as input and use hierarchical clustering to classify all node words.

Images