Quantitative prediction method and device for qualitative distance description, and readable storage medium

Abstract

The present invention discloses a quantitative prediction method and device for qualitative distance description, and a readable storage medium. The method comprises constructing a sample set of a specified area, wherein each sample in the sample set comprises a qualitative distance descriptor and a real distance corresponding to the qualitative distance descriptor; obtaining a distance interval corresponding to each qualitative distance descriptor in the sample set according to the relationship between the qualitative distance descriptors and the real distances of all samples in the sample set; training a quantitative prediction model with all the obtained qualitative distance descriptors and their corresponding distance intervals; inputting the qualitative distance descriptor of a sample to be tested in the specified area into the trained quantitative prediction model, and obtaining a distance interval corresponding to the qualitative distance descriptor of the sample to be tested. The present invention can convert the qualitative distance description into a distance quantitative interval, thereby improving the rationality and accuracy of positioning the target address.

Description

Technical Field

[0001] The present invention relates to the field of quantitative prediction, and in particular to a quantitative prediction method and device for qualitative distance description, and a readable storage medium. Background Art

[0002] Urban public security alarm systems, such as fire alarm systems, contain a large amount of alarm description records. These alarm descriptions usually contain several common elements, such as time, address, person, event, etc. Among them, address information is one of the important key elements because it is related to the dispatch plan of urban public security departments.

[0003] The address elements in the original police description record information are mostly verbal expressions and colloquial descriptions of the parties involved, and their patterns are usually:

[0004] The first type: the alarm address element = [pre-qualitative distance description] <reference object>. Common words for pre-qualitative distance description include: "close to", "approaching", "far away", etc. For example, "close to XX Building".

[0005] The second type: the alarm address element = [preposition] <reference> [postposition qualitative distance description]. The preposition can be empty or exist. Common prepositions include: "at", "from", "distance", etc.; postposition qualitative distance descriptions include "near", "next to", "very close", "very far", "relatively far", etc. For example, "near the intersection of XX Road and XX Road", "next to XX supermarket".

[0006] These police address elements all contain colloquial pre- or post-position qualitative distance descriptors such as "near", "close to", and "next to". How to infer how far the police address is from the reference object in the original police description record information (for example, the above-mentioned "XX Road" and "XX Supermarket") based on these qualitative distance descriptors is a difficult problem to solve. Inaccurate address reasoning will lead to inaccurate allocation of the case management agency or deviation in the dispatch plan of the disposal force.

[0007] At present, the use of natural language processing technology (NLP technology) to process police descriptions is still limited to extracting reference objects, prepending / postpending qualitative distance descriptors, and vaguely expressing the vicinity of a certain reference object, but it is still unable to quickly and accurately provide the specific distance from the reference object. Generally speaking, after extracting qualitative distance descriptors, manual distance judgment is performed, which is inefficient and inaccurate. Summary of the invention

[0008] One of the purposes of the present invention is to overcome the deficiencies in the prior art and to provide a quantitative prediction method and device for qualitative distance description and a readable storage medium.

[0009] The technical solution provided by the present invention is as follows:

[0010] Quantitative prediction methods described by qualitative distances include:

[0011] Constructing a sample set of a specified area, each sample in the sample set includes a qualitative distance descriptor and a real distance corresponding to the qualitative distance descriptor;

[0012] According to the relationship between the qualitative distance descriptors and the true distances of all samples in the sample set, a distance interval corresponding to each qualitative distance descriptor in the sample set is obtained;

[0013] Use all the obtained qualitative distance descriptors and their corresponding distance intervals to train the quantitative prediction model;

[0014] The qualitative distance descriptor of a sample to be tested in the designated area is input into the trained quantitative prediction model to obtain the distance interval corresponding to the qualitative distance descriptor of the sample to be tested.

[0015] In one embodiment, constructing a sample set of a specified area includes:

[0016] Collect historical alarm cases in designated areas;

[0017] Extracting qualitative distance descriptors, reference objects and corresponding crime locations from the alarm address descriptions of the historical alarm cases;

[0018] The distance between the reference object and the crime scene is calculated, and the distance is recorded as the real distance corresponding to the qualitative distance descriptor.

[0019] In one embodiment, the step of obtaining the distance interval corresponding to each qualitative distance descriptor in the sample set according to the relationship between the qualitative distance descriptors and the true distances of all samples in the sample set includes:

[0020] Setting a number of consecutive preset distance intervals;

[0021] Traversing all samples in the sample set, and determining the association relationship between the qualitative distance descriptor of each sample and each preset distance interval according to the relationship between the real distance corresponding to the qualitative distance descriptor of each sample and each preset distance interval;

[0022] Count the cumulative number of associations of each qualitative distance descriptor in each preset distance interval;

[0023] For each qualitative distance descriptor, a preset distance interval with the highest cumulative association times is selected as the distance interval corresponding to the qualitative distance descriptor.

[0024] In one embodiment, the step of setting a plurality of continuous preset distance intervals includes: obtaining a probability distribution function graph of the designated area according to the actual distances corresponding to all qualitative distance descriptors of the sample set; and setting a plurality of continuous preset distance intervals according to the probability distribution function graph in a manner of equal cumulative probability.

[0025] In one embodiment, the step of training the quantitative prediction model with all the obtained qualitative distance descriptors and their corresponding distance intervals includes: encoding the input qualitative distance descriptors using a word embedding model, converting the qualitative distance descriptors into corresponding word vectors; and training the quantitative prediction model using a machine learning method.

[0026] In one embodiment, the step of inputting a qualitative distance descriptor of a sample to be tested in the specified area into the trained quantitative prediction model to obtain a distance interval corresponding to the qualitative distance descriptor of the sample to be tested includes: if the qualitative distance descriptor of the sample to be tested has not appeared in the sample set, the quantitative prediction model returns the distance interval corresponding to the qualitative distance descriptor that has appeared and has similar semantics.

[0027] The present invention also provides a quantitative prediction device for qualitative distance description, comprising:

[0028] A sample set construction module, used to construct a sample set of a specified area, each sample in the sample set includes a qualitative distance descriptor and a real distance corresponding to the qualitative distance descriptor;

[0029] A data processing module, used to obtain the distance interval corresponding to each qualitative distance descriptor in the sample set according to the relationship between the qualitative distance descriptors and the true distances of all samples in the sample set;

[0030] The model training module uses all the obtained qualitative distance descriptors and their corresponding distance intervals to train the quantitative prediction model;

[0031] The quantitative prediction module is used to input a sample to be tested in the designated area into the trained quantitative prediction model to obtain a distance interval corresponding to a qualitative distance description word of the sample to be tested.

[0032] In one embodiment, the sample set construction module is also used to collect historical alarm cases in a specified area; extract qualitative distance descriptors, reference objects and corresponding crime scenes from the alarm address descriptions of the historical alarm cases; calculate the distance between the reference object and the crime scene, and record the distance as the actual distance corresponding to the qualitative distance descriptor.

[0033] In one embodiment, the data processing module includes:

[0034] A preset interval setting unit, used to set a number of continuous preset distance intervals;

[0035] A statistical unit is used to traverse all samples in the sample set, determine the association relationship between the qualitative distance descriptor of each sample and each preset distance interval according to the relationship between the actual distance corresponding to the qualitative distance descriptor of each sample and each preset distance interval; and count the cumulative number of associations of each qualitative distance descriptor in each preset distance interval;

[0036] The determining unit is used to select, for each qualitative distance descriptor, a preset distance interval with the highest cumulative association times as the distance interval corresponding to the qualitative distance descriptor.

[0037] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the quantitative prediction method described by the qualitative distance are implemented.

[0038] Compared with the prior art, the quantitative prediction method and device of qualitative distance description and the readable storage medium provided by the present invention have at least one of the following beneficial effects:

[0039] 1. The present invention obtains the distance interval corresponding to the qualitative distance descriptor by studying the relationship between the qualitative distance descriptor and the actual distance in the historical cases. In the actual police situation, the reference object is combined with the distance interval corresponding to the qualitative distance descriptor to quickly determine the distance range of the target location, which is convenient for the staff to accurately locate the target address.

[0040] 2. The present invention can improve the prediction efficiency and accuracy of quantitative prediction and eliminate the differences in human judgment.

[0041] 3. The present invention can train models and predict distances for different regions based on local historical alarm information, and can better adapt to local alarm receiving and handling systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The preferred implementation modes will be described below in a clear and understandable manner with reference to the accompanying drawings to further illustrate the above-mentioned characteristics, technical features, advantages and implementation methods of the present invention.

[0043] Figure 1 is a flow chart of an embodiment of the quantitative prediction method of qualitative distance description of the present invention;

[0044] Figure 2 It is a schematic diagram of a probability distribution function graph;

[0045] Figure 3 It is a structural schematic diagram of an embodiment of a quantitative prediction device for qualitative distance description of the present invention;

[0046] Figure 4 It is a structural schematic diagram of an embodiment of a data processing module in the quantitative prediction device of qualitative distance description of the present invention;

[0047] Figure 5 It is a schematic diagram of a specific implementation scenario example of the present invention.

[0048] Description of Figure Numbers:

[0049] 100. Sample set construction module, 200. Data processing module, 300. Model training module, 400. Quantitative prediction module, 210. Preset interval setting unit, 220. Statistics unit, 230. Determination unit. DETAILED DESCRIPTION

[0050] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the specific implementation methods of the present invention will be described below with reference to the accompanying drawings. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings and other implementation methods can be obtained based on these drawings without creative work.

[0051] In order to simplify the drawings, only the parts related to the invention are schematically shown in each figure, and they do not represent the actual structure of the product. In addition, in order to simplify the drawings and facilitate understanding, in some figures, only one of the parts with the same structure or function is schematically drawn or marked. In this article, "one" not only means "only one", but also means "more than one".

[0052] As mentioned above, the address elements in the original police situation description are mostly colloquial descriptions of the parties involved, and their colloquial descriptions usually contain qualitative distance descriptors and reference objects. These qualitative distance descriptors cannot accurately provide the distance from the reference object, causing certain interference to the police officers.

[0053] In order to solve the above problem, the present invention obtains the distance interval corresponding to the qualitative distance descriptor of the specified area by studying the relationship between the qualitative distance descriptor and the actual distance in the historical alarm cases of the specified area. According to the distance interval and the reference object, the quantitative range of the target location can be obtained, thereby helping the police officers to better judge the actual location of the target address.

[0054] The present invention is specifically described below in conjunction with the qualitative distance descriptors in the alarm information. However, it should be noted that the quantitative prediction method provided by the present invention can be applied to other fields in addition to the fire protection field.

[0055] In one embodiment, Figure 1As shown, a quantitative prediction method for qualitative distance description includes:

[0056] Step S100 constructs a sample set of a specified area, each sample in the sample set includes a qualitative distance descriptor and a real distance corresponding to the qualitative distance descriptor;

[0057] Step S200 obtains the distance interval corresponding to each qualitative distance descriptor in the sample set according to the relationship between the qualitative distance descriptors and the true distances of all samples in the sample set;

[0058] Step S300 uses all the obtained qualitative distance descriptors and their corresponding distance intervals to train a quantitative prediction model;

[0059] Step S400 inputs a qualitative distance descriptor of a sample to be tested in a designated area into a trained quantitative prediction model to obtain a distance interval corresponding to the qualitative distance descriptor of the sample to be tested.

[0060] The present invention obtains the distance intervals corresponding to the qualitative distance descriptors that have appeared in the sample set by analyzing the relationship between the qualitative distance descriptors and the true distances in the sample set of the specified area; the quantitative prediction model is trained with the qualitative distance descriptors that have appeared and their corresponding distance intervals, so that the quantitative prediction model can predict the distance intervals corresponding to the qualitative distance descriptors that have not appeared, thereby realizing the quantitative prediction of the qualitative distance descriptors of any sample to be tested.

[0061] A sample to be tested is input into a trained quantitative prediction model, and the quantitative prediction model outputs the distance interval corresponding to the qualitative distance descriptor of the sample to be tested. Based on the distance interval and the reference object of the sample to be tested, the distance range of the target location can be quickly determined, assisting the staff in accurately locating the target address.

[0062] Because the distance intervals corresponding to the qualitative distance descriptors are obtained based on historical cases, they are more accurate and objective than the distance intervals predicted manually. Compared with the traditional manual prediction method, the present invention can improve the prediction efficiency and accuracy, and can also eliminate the differences in human judgment (different people may have different understandings of far and near, and the same person may have different understandings of the same qualitative distance descriptor under different cases), and the prediction results are more standardized and unified.

[0063] Because people in different regions may have a large deviation in their understanding of the same qualitative distance descriptor, it is recommended to collect historical cases in a specified area to build a sample set for the specified area, use the sample set for the specified area to train the quantitative prediction model, obtain a quantitative prediction model suitable for the specified area, and use the quantitative prediction model for the specified area to achieve quantitative prediction of the qualitative distance description of the samples to be tested in the specified area. This can further improve the prediction accuracy of the quantitative prediction model.

[0064] In one embodiment, step S100 includes:

[0065] Step S110 collects historical alarm cases in a specified area;

[0066] Step S120 extracts qualitative distance description words, reference objects and corresponding crime locations from the alarm address descriptions of historical alarm cases;

[0067] Step S130 calculates the distance between the reference object and the crime scene, and records the distance as the real distance corresponding to the qualitative distance descriptor.

[0068] Through the above steps, the correspondence between the qualitative distance descriptor and the true distance of each case can be obtained.

[0069] There are many ways to determine the distance interval corresponding to the qualitative distance descriptor that has appeared based on the sample set. For example, in one embodiment, an average value is calculated based on all the real distances corresponding to the same qualitative distance descriptor in the sample set as the average distance, and the distance interval corresponding to the qualitative distance descriptor is obtained based on the average distance.

[0070] In a preferred embodiment, step S200 includes:

[0071] Step S210 sets a number of consecutive preset distance intervals;

[0072] Step S220 traverses all samples in the sample set, and determines the association relationship between the qualitative distance descriptor of the corresponding sample and each preset distance interval according to the relationship between the actual distance corresponding to the qualitative distance descriptor of each sample and each preset distance interval;

[0073] Step S230 counts the cumulative number of associations of each qualitative distance descriptor in each preset distance interval;

[0074] Step S240 selects, for each qualitative distance descriptor, a preset distance interval with the highest cumulative association times as the distance interval corresponding to the qualitative distance descriptor.

[0075] There are many optional ways to set the preset distance interval. For example, obtain the maximum value Dmax of the real distance d in the sample set, divide Dmax into N equal parts, and obtain N preset distance intervals.

[0076] Traverse all samples in the sample set: if the true distance d corresponding to the qualitative distance descriptor of the sample is in the preset distance interval, then the qualitative distance descriptor of the sample is associated with the preset distance interval once; otherwise, the qualitative distance descriptor of the sample is not associated with the preset distance interval.

[0077] According to the above data, count the cumulative association times of each qualitative distance descriptor in each preset distance interval, and select the preset distance interval with the highest cumulative association times as the distance interval corresponding to the qualitative distance descriptor.

[0078] Suppose Dmax = 10m, divide 10m into 10 intervals evenly, such as interval 1: {d|d <= 1m}, interval 2: {d|1m < d <= 2m}, interval 3: {d|2m < d <= 3m}, and so on.

[0079] There are 10 cases in the sample set that contain the qualitative distance descriptor "nearby". Among them, the true distances of 3 cases are 1m, which fall into interval 1; the true distances of 4 cases are 3m, which fall into interval 3; the true distances of 3 cases are 5m, which fall into interval 5; these reflect the relationship between the qualitative distance descriptor "nearby" and the true distance.

[0080] Among them, the true distances of 3 cases fall into interval 1, the true distances of 4 cases fall into interval 3, and the true distances of 3 cases fall into interval 5. Therefore, interval 1, interval 3, and interval 5 are associated with "nearby", and the cumulative association times are 3, 4, and 5 respectively. Obviously, the cumulative association times of interval 3 are the highest. So, adopting the solution of this preferred embodiment, interval 3 is used as the distance interval corresponding to "nearby".

[0081] It can be seen that the distance intervals corresponding to the qualitative distance descriptors that have appeared in this preferred embodiment are affected by the preset distance intervals. In order to further improve the rationality of the preset distance intervals, in one embodiment, step S210 specifically includes:

[0082] Step S211 takes the distance as a random variable, and obtains the probability distribution function graph of this random variable in the specified area according to all the true distance values in the sample set;

[0083] Step S212 sets a number of consecutive preset distance intervals according to this probability distribution function graph in the way that the cumulative probabilities are equal.

[0084] Specifically, the horizontal axis of the probability distribution function graph represents the distance (i.e., the true distance between the reference object and the crime scene), and the vertical axis represents the cumulative probability. By counting each true distance d in the sample set, the probability distribution function graph of the specified area can be obtained.

[0085] Figure 2 It is a schematic diagram of a probability distribution function. Suppose the probability distribution function graph of the specified area is as Figure 2 shown. According to this curve, arbitrarily take L cumulative probabilities, denoted as y1%, y2%... y L %, and the corresponding distance points Q1, Q2,... Q L, respectively satisfying the following relations: P{d<=Q1}=y1%, P{d<=Q2}=y2%,…, P{d<=Q L}=y L %.

[0086] According to the distance points Q1, Q2, ...Q L There are L distance intervals: (0, Q1], (Q1, Q2], ... (Q L-1 ,Q L ]. These L distance intervals may be used as preset distance intervals.

[0087] If y2%-y1%=y3%-y2%=...=y L %-y L-1 %, then the cumulative probability of the alarm occurring in each preset distance interval is equal. This method of setting the preset distance interval is also called the method of equal cumulative probability. For example, assuming y1% = 20%, y2% = 40%, y3% = 60%, y4% = 80%, y5% = 100%, then the cumulative probabilities of the five preset distance intervals obtained are equal.

[0088] In one embodiment, step S300 includes:

[0089] Step S310 uses a word embedding model to encode the input qualitative distance description words and convert the qualitative distance description words into corresponding word vectors;

[0090] Step S320 uses a machine learning method to train the quantitative prediction model.

[0091] The prerequisite for using natural language processing (NLP) technology is to convert text data into a vectorized form that can be recognized and processed by computers. The purpose of encoding qualitative distance description words is to convert qualitative distance description words into corresponding word vectors.

[0092] The word embedding model (Word to Vector: Word2vec, word vector generation model) is used to encode the qualitative distance description words. Through the Word2vec encoding method, the distance between two words with similar semantics in the word vector space is also similar. In this way, the generalization of the quantitative prediction model can be improved.

[0093] Qualitative distance descriptors with corresponding distance intervals are called known distance words, and qualitative distance descriptors without corresponding distance intervals are called unknown distance words. Through step S200, qualitative distance descriptors appearing in the sample set become known distance words. Qualitative distance descriptors that have not appeared in the sample set are unknown distance words.

[0094] When encountering a qualitative distance descriptor that has not appeared before, the quantitative prediction model can use its corresponding word embedding vector to make predictions and infer its corresponding distance interval. Generally speaking, if the unknown distance word is semantically similar to a known distance word, then the corresponding distance interval obtained by the model is usually similar.

[0095] Other encoding methods may also be used for qualitative distance description words, as long as the encoding method satisfies: the distance between two vectors in the word vector space can reflect the semantic similarity between the corresponding two words, and the present invention does not impose any limitation on this.

[0096] In one embodiment, step S400 includes:

[0097] Step S410: If the qualitative distance descriptor of the sample to be tested has appeared in the sample set, the quantitative prediction model returns the distance interval corresponding to the qualitative distance descriptor;

[0098] In step S420, if the qualitative distance descriptor of the sample to be tested has not appeared in the sample set, the quantitative prediction model returns the distance interval corresponding to the qualitative distance descriptor that has appeared and is closest to it in semantics.

[0099] This embodiment requires the quantitative prediction model to use the Word2vec model encoding or similar encoding for the input qualitative distance descriptors, so that the semantic gap between the qualitative distance descriptors can be measured by the distance in the word vector space. The problem of finding the closest semantics to the qualitative distance descriptor of the sample to be tested among the qualitative distance descriptors that have appeared is converted into the problem of finding the word vector with the smallest word vector distance to the qualitative distance descriptor of the sample to be tested in the word vector space.

[0100] This embodiment matches qualitative distance descriptors with actual distances and performs training in different cities and different situations, thereby making faster and more accurate inferences about unknown target locations.

[0101] In one embodiment, Figure 3 As shown, a quantitative prediction device for qualitative distance description includes:

[0102] The sample set construction module 100 is used to construct a sample set of a specified area, each sample in the sample set includes a qualitative distance descriptor and a real distance corresponding to the qualitative distance descriptor;

[0103] The data processing module 200 is used to obtain the distance interval corresponding to each qualitative distance descriptor in the sample set according to the relationship between the qualitative distance descriptors and the true distances of all samples in the sample set;

[0104] A model training module 300 uses all the obtained qualitative distance descriptors and their corresponding distance intervals to train a quantitative prediction model;

[0105] The quantitative prediction module 400 is used to input a sample to be tested in a designated area into a trained quantitative prediction model to obtain a distance interval corresponding to a qualitative distance description word of the sample to be tested.

[0106] The present invention obtains the distance intervals corresponding to the qualitative distance descriptors that have appeared in the sample set by analyzing the relationship between the qualitative distance descriptors and the true distances in the sample set of the specified area; the quantitative prediction model is trained with the qualitative distance descriptors that have appeared and their corresponding distance intervals, so that the quantitative prediction model can predict the distance intervals corresponding to the qualitative distance descriptors that have not appeared, thereby realizing the quantitative prediction of the qualitative distance descriptors of any sample to be tested.

[0107] In one embodiment, the sample set construction module 100 is also used to collect historical alarm cases in a specified area; extract qualitative distance descriptors, reference objects and corresponding crime scenes from the alarm address descriptions of the historical alarm cases; calculate the distance between the reference object and the crime scene, and record the distance as the actual distance corresponding to the qualitative distance descriptor.

[0108] In one embodiment, Figure 4 As shown, the data processing module 200 includes:

[0109] The preset interval setting unit 210 is used to set a number of consecutive preset distance intervals;

[0110] The statistical unit 220 is used to traverse all samples in the sample set, determine the association relationship between the qualitative distance descriptor of each sample and each preset distance interval according to the relationship between the actual distance corresponding to the qualitative distance descriptor of each sample and each preset distance interval; and count the cumulative number of associations of each qualitative distance descriptor in each preset distance interval;

[0111] The determination unit 230 is configured to select, for each qualitative distance descriptor, a preset distance interval with the highest cumulative association times as the distance interval corresponding to the qualitative distance descriptor.

[0112] In one embodiment, the preset interval setting unit 210 is also used to take distance as a random variable, and obtain a probability distribution function graph of the random variable in a specified area based on all true distance values ​​in a sample set; based on the probability distribution function graph, set a number of continuous preset distance intervals in a manner with equal cumulative probability.

[0113] In one embodiment, the model training module 300 is further used to encode the input qualitative distance description words using the Word2vec model, convert the qualitative distance description words into corresponding word vectors, and train the quantitative prediction model using a machine learning method.

[0114] In one embodiment, the quantitative prediction module 400 is further configured to: if the qualitative distance descriptor of the sample to be measured appears in the sample set, the quantitative prediction model returns the distance interval corresponding to the qualitative distance descriptor; if the qualitative distance descriptor of the sample to be measured does not appear in the sample set, the quantitative prediction model returns the distance interval corresponding to the qualitative distance descriptor that has appeared and is semantically similar to it.

[0115] It should be noted that the embodiments of the quantitative prediction device for qualitative distance description provided by the present invention and the embodiments of the quantitative prediction method for qualitative distance description provided above are both based on the same inventive concept and can achieve the same technical effects. Therefore, other specific contents of the embodiments of the quantitative prediction device for qualitative distance description can refer to the records of the embodiments of the quantitative prediction method for qualitative distance description above.

[0116] In one embodiment, a computer-readable storage medium stores a computer program thereon. When the computer program is executed by a processor, it can implement the quantitative prediction method for qualitative distance description as recorded in the foregoing embodiments. That is, when part or all of the technical solutions that contribute to the prior art in the foregoing embodiments of the present invention are embodied in the form of a computer software product, the foregoing computer software product is stored in a computer-readable storage medium. The computer-readable storage medium can be any portable computer program code entity device or equipment, such as a USB flash drive, a removable disk, a magnetic disk, an optical disk, a computer memory, a read-only memory, a random access memory, etc.

[0117] The present invention also provides a specific implementation scenario example, applying the foregoing quantitative prediction method and device for qualitative distance description to the processing of the police situation address elements in the police call handling system.

[0118] First, collect the historical alarm information of the police call handling system in a certain city to construct a sample set, preprocess the samples in the sample set, extract the reference objects and the corresponding qualitative distance descriptors in the alarm information provided by the reporter, calculate the true distance d (by Euclidean distance) between the target location (the location of the incident) and the reference object, and count all the qualitative distance descriptors and the true distances in this area to obtain a probability distribution function graph.

[0119] According to this probability distribution function graph, in the way of equal cumulative probability, (L + 1) preset distance intervals are obtained, and each preset distance interval corresponds to a label (Label), as follows:

[0120] When 0 < d <= Q1, the corresponding Label is 1;

[0121] When Q1 < d <= Q2, the corresponding Label is 2;

[0122] …

[0123] When Q L-1 <d <= Q L , the corresponding Label is L;

[0124] When Q L < d, the corresponding Label is L + 1.

[0125] Encode all qualitative distance descriptors. Using the Word2vec model, encode N qualitative distance descriptors into vectors V1, V2, …, V N . Traverse all samples in the sample set, and count the cumulative association times of each qualitative distance descriptor in each preset distance interval; for each qualitative distance descriptor, select the preset distance interval with the highest cumulative association times as the distance interval corresponding to this qualitative distance descriptor. In this way, each qualitative distance descriptor vector can correspond to a label, that is, each qualitative distance descriptor corresponds to a real distance interval.

[0126] As Figure 5 shown, use a machine learning model to train the data of the associated qualitative distance descriptors and real distance intervals. The machine learning model mentioned here is a concept of a generalized machine learning model, including a logistic model, a deep neural network model, etc.

[0127] After the model is trained, the above machine learning training part is completed. The following is the prediction and inference part. After receiving a new police situation information, first extract the qualitative distance descriptors in it. When encountering a qualitative distance descriptor that has not appeared before, the trained model returns the distance interval corresponding to the qualitative distance descriptor that has appeared and is semantically closest to it.

[0128] In different regions and different situations, extract the reference objects and qualitative distance descriptors respectively, count to obtain the probability distribution function, obtain the corresponding preset distance interval, encode the real distance between the reference object and the target address and the above qualitative distance descriptors, and use machine learning methods to optimize and train respectively to obtain different models, and finally predict and return the auxiliary inference distance interval. In different situations, the auxiliary inference distance intervals corresponding to the same colloquial qualitative description may be different.

[0129] It should be noted that the above embodiments can be freely combined according to needs. The above is only the preferred embodiment of the present invention. It should be pointed out that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A quantitative prediction method for qualitative distance description, characterized in that: include: Constructing a sample set of a specified area, each sample in the sample set includes a qualitative distance descriptor and a real distance corresponding to the qualitative distance descriptor; According to the relationship between the qualitative distance descriptors and the true distances of all samples in the sample set, a distance interval corresponding to each qualitative distance descriptor in the sample set is obtained; Use all the obtained qualitative distance descriptors and their corresponding distance intervals to train the quantitative prediction model; The qualitative distance descriptor of a sample to be tested in the designated area is input into the trained quantitative prediction model to obtain the distance interval corresponding to the qualitative distance descriptor of the sample to be tested.

2. The quantitative prediction method of qualitative distance description according to claim 1, characterized in that: The constructing of a sample set of a specified area includes: Collect historical alarm cases in designated areas; Extracting qualitative distance descriptors, reference objects and corresponding crime locations from the alarm address descriptions of the historical alarm cases; The distance between the reference object and the crime scene is calculated, and the distance is recorded as the real distance corresponding to the qualitative distance descriptor.

3. The quantitative prediction method of qualitative distance description according to claim 1, characterized in that: The step of obtaining the distance interval corresponding to each qualitative distance descriptor in the sample set according to the relationship between the qualitative distance descriptors and the true distances of all samples in the sample set includes: Setting a number of consecutive preset distance intervals; Traversing all samples in the sample set, and determining the association relationship between the qualitative distance descriptor of each sample and each preset distance interval according to the relationship between the real distance corresponding to the qualitative distance descriptor of each sample and each preset distance interval; Count the cumulative number of associations of each qualitative distance descriptor in each preset distance interval; For each qualitative distance descriptor, a preset distance interval with the highest cumulative association times is selected as the distance interval corresponding to the qualitative distance descriptor.

4. The quantitative prediction method of qualitative distance description according to claim 3, characterized in that: The steps for setting a number of consecutive preset distance intervals include: Obtaining a probability distribution function graph of the designated area according to the real distances corresponding to all qualitative distance descriptors of the sample set; According to the probability distribution function diagram, a plurality of continuous preset distance intervals are set in a manner with equal cumulative probabilities.

5. The quantitative prediction method of qualitative distance description according to claim 1, characterized in that: Step 1: Use all the obtained qualitative distance descriptors and their corresponding distance intervals to train the quantitative prediction model, including: Using a word embedding model to encode the input qualitative distance description words, and converting the qualitative distance description words into corresponding word vectors; The quantitative prediction model is trained using a machine learning method.

6. The quantitative prediction method of qualitative distance description according to claim 5, characterized in that: The step of inputting the qualitative distance descriptor of a sample to be tested in the designated area into the trained quantitative prediction model to obtain the distance interval corresponding to the qualitative distance descriptor of the sample to be tested comprises: If the qualitative distance descriptor of the sample to be tested has not appeared in the sample set, the quantitative prediction model returns the distance interval corresponding to the qualitative distance descriptor that has appeared and has similar semantics.

7. A quantitative prediction device for qualitative distance description, characterized in that: include: A sample set construction module, used to construct a sample set of a specified area, each sample in the sample set includes a qualitative distance descriptor and a real distance corresponding to the qualitative distance descriptor; A data processing module, used to obtain the distance interval corresponding to each qualitative distance descriptor in the sample set according to the relationship between the qualitative distance descriptors and the true distances of all samples in the sample set; The model training module uses all the obtained qualitative distance descriptors and their corresponding distance intervals to train the quantitative prediction model; The quantitative prediction module is used to input a sample to be tested in the designated area into the trained quantitative prediction model to obtain a distance interval corresponding to a qualitative distance description word of the sample to be tested.

8. The quantitative prediction device of qualitative distance description according to claim 7, characterized in that: The sample set construction module is also used to collect historical alarm cases in a specified area; extract qualitative distance descriptors, reference objects and corresponding crime scenes from the alarm address descriptions of the historical alarm cases; calculate the distance between the reference object and the crime scene, and record the distance as the actual distance corresponding to the qualitative distance descriptor.

9. The quantitative prediction device of qualitative distance description according to claim 7, characterized in that: The data processing module comprises: A preset interval setting unit, used to set a number of continuous preset distance intervals; A statistical unit is used to traverse all samples in the sample set, determine the association relationship between the qualitative distance descriptor of each sample and each preset distance interval according to the relationship between the actual distance corresponding to the qualitative distance descriptor of each sample and each preset distance interval; and count the cumulative number of associations of each qualitative distance descriptor in each preset distance interval; The determining unit is used to select, for each qualitative distance descriptor, a preset distance interval with the highest cumulative association times as the distance interval corresponding to the qualitative distance descriptor.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the quantitative prediction method described by the qualitative distance as claimed in any one of claims 1 to 6 are implemented.

Images