Information evaluation method and device, electronic equipment, storage medium and program product

Abstract

The invention discloses an information evaluation method and device, electronic equipment, a storage medium and a program product. The specific implementation scheme comprises the steps of determining to-be-evaluated information; determining a multi-kernel convolution model comprising multiple groups of dynamic convolution kernels; extracting features of the to-be-evaluated information based on the multi-kernel convolution model, and generating a length sensing text vector corresponding to the to-be-evaluated information; and based on the first length perception text vector and the second length perception text vector, determining a recruitment score corresponding to the to-be-evaluated information. The to-be-evaluated information is analyzed through the multi-kernel convolution model comprising multiple groups of dynamic convolution kernels, so that truncation of the to-be-evaluated information is avoided, and information loss in the evaluation process is reduced. According to the method, the length sensing text vector of the to-be-evaluated information is obtained, and the recruitment score is obtained based on the length sensing text vector, so that the rationality of the recruitment information in the to-be-evaluated information is evaluated, and the rationality of the recruitment information is ensured.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to an information evaluation method, apparatus, electronic device, storage medium, and program product. Background Technology

[0002] In today's society, job postings are crucial for recruiting talent. However, with the development of the internet and information technology, various fake job postings have emerged, causing numerous problems and risks for job seekers and wasting their time and energy.

[0003] Therefore, evaluating long-text job postings and determining their reasonableness is a pressing issue that needs to be addressed. Summary of the Invention

[0004] This invention provides an information evaluation method, apparatus, electronic device, storage medium, and program product to evaluate the rationality of recruitment information and improve the rationality of recruitment information.

[0005] According to one aspect of the present invention, an information evaluation method is provided, comprising:

[0006] The information to be evaluated is determined, including long text information that needs to be evaluated, including recruitment information and business registration information of the companies that publish the recruitment information;

[0007] A multi-kernel convolution model is determined, comprising multiple sets of dynamic convolution kernels, wherein the size of the convolution kernels contained in the multiple sets of dynamic convolution kernels is associated with the information to be evaluated;

[0008] Based on the multi-kernel convolution model, features of the information to be evaluated are extracted, and length-aware text vectors corresponding to the information to be evaluated are generated. The length-aware text vectors include a first length-aware text vector corresponding to the recruitment information and a second length-aware text vector corresponding to the business information.

[0009] Based on the first length-aware text vector and the second length-aware text vector, a recruitment score corresponding to the information to be evaluated is determined, and the recruitment score indicates whether the recruitment information is reasonable.

[0010] According to another aspect of the present invention, an information evaluation apparatus is provided, comprising:

[0011] The first determining module is used to determine the information to be evaluated, which includes long text information that needs to be evaluated, including recruitment information and business registration information of the company that published the recruitment information;

[0012] The second determining module is used to determine a multi-kernel convolution model including multiple sets of dynamic convolution kernels, wherein the size of the convolution kernels included in the multiple sets of dynamic convolution kernels is associated with the information to be evaluated;

[0013] The generation module is used to extract features of the information to be evaluated based on the multi-kernel convolution model and generate a length-aware text vector corresponding to the information to be evaluated. The length-aware text vector includes a first length-aware text vector corresponding to the recruitment information and a second length-aware text vector corresponding to the business information.

[0014] The third determining module is used to determine the recruitment score corresponding to the information to be evaluated based on the first length-aware text vector and the second length-aware text vector, wherein the recruitment score indicates whether the recruitment information is reasonable.

[0015] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0016] At least one processor; and

[0017] A memory communicatively connected to the at least one processor; wherein,

[0018] The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the information evaluation method according to any embodiment of the present invention.

[0019] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the information evaluation method described in any embodiment of the present invention.

[0020] According to another aspect of the present invention, a computer program product is provided, the computer program product comprising a computer program that, when executed by a processor, implements the information evaluation method described in any embodiment of the present invention.

[0021] The technical solution of this invention involves: determining the information to be evaluated; determining a multi-kernel convolutional model including multiple sets of dynamic convolutional kernels; extracting features of the information to be evaluated based on the multi-kernel convolutional model to generate a length-aware text vector corresponding to the information to be evaluated; and determining a recruitment score corresponding to the recruitment information based on the first length-aware text vector and the second length-aware text vector, wherein the recruitment score indicates whether the recruitment information is reasonable. By using a multi-kernel convolutional model including multiple sets of dynamic convolutional kernels to analyze the information to be evaluated, truncation of the information to be evaluated is avoided, reducing information loss during the evaluation process. By obtaining the length-aware text vector of the information to be evaluated and obtaining the recruitment score based on the length-aware text vector, the reasonableness evaluation of the recruitment information in the information to be evaluated is achieved, ensuring the reasonableness of the recruitment information.

[0022] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart of an information evaluation method provided according to Embodiment 1 of the present invention;

[0025] Figure 2 This is a flowchart of a vector generation method provided according to Embodiment 2 of the present invention;

[0026] Figure 3 This is a schematic diagram of the structure of an information evaluation device according to Embodiment 3 of the present invention;

[0027] Figure 4 This is a block diagram of an electronic device provided according to Embodiment 4 of the present invention. Detailed Implementation

[0028] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0029] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0030] The acquisition, storage, use, and processing of data in the technical solution of this invention all comply with the relevant provisions of laws and regulations.

[0031] Example 1

[0032] Figure 1 This is a flowchart of an information evaluation method according to Embodiment 1 of the present invention. This embodiment is applicable to the evaluation of the rationality of recruitment information. The method can be executed by an information evaluation device, which can be implemented in hardware and / or software and can be configured in an electronic device. Figure 1 As shown, the method includes:

[0033] S110. Determine the information to be evaluated.

[0034] The information to be evaluated includes long text information that needs to be evaluated, including recruitment information and the business registration information of the company that posted the recruitment information.

[0035] In this embodiment, the information to be evaluated can be understood as text information that needs to be evaluated. The information to be evaluated can be a text pair formed by recruitment information and business registration information. Recruitment information can be understood as information posted by a company regarding job vacancies. Business registration information can be understood as information about the company that posted the recruitment information. Long text information can be understood as text information exceeding the model's usual processing length. Long text information contains a large amount of information and can be composed of various types of text information.

[0036] Specifically, recruitment information and business registration information of enterprises are collected. These are one-to-one text pairs, with the enterprise's business registration information corresponding to its recruitment information. The text pairs formed by these recruitment and business registration information are then used as the long text information to be evaluated.

[0037] For example, the information to be evaluated is as follows: 1. Business registration information: xxxxxx 2. Job information: xxxxxx.

[0038] S120. Determine a multi-kernel convolution model that includes multiple sets of dynamic convolution kernels.

[0039] The size of the convolution kernels contained in the multiple sets of dynamic convolution kernels is associated with the information to be evaluated.

[0040] In this embodiment, multiple sets of dynamic convolutional kernels can be understood as multiple sets of convolutional kernels obtained based on the information to be evaluated, and different sizes of convolutional kernels can be set according to the information to be evaluated. A multi-kernel convolutional model can be understood as a model with multiple different convolutional kernel sizes, which can be used to perform convolution operations on the information to be evaluated.

[0041] Specifically, information contained in the information to be evaluated can be extracted to generate multiple sets of dynamic convolutional kernels with different kernel sizes. By extracting segmentation information, sentence information, and fine-grained information from the information to be evaluated separately, different kernel sizes can be calculated, resulting in a multi-kernel convolutional model with multiple sets of dynamic convolutional kernels.

[0042] S130. Based on the multi-kernel convolution model, extract the features of the information to be evaluated and generate the length-aware text vector corresponding to the information to be evaluated.

[0043] The length-aware text vector includes a first length-aware text vector corresponding to the recruitment information and a second length-aware text vector corresponding to the business registration information.

[0044] In this embodiment, a length-aware text vector can be understood as a vector containing features of the information to be evaluated, which can be obtained by vectorizing the long text information contained in the information to be evaluated. Specifically, the first length-aware text vector contains features of recruitment information, and the second length-aware text vector contains features of business registration information.

[0045] Specifically, for the recruitment information and business registration information contained in the information to be evaluated, a multi-kernel convolutional model is used to extract features from the recruitment information and business registration information respectively. The feature extraction method can be to perform convolution using multiple sets of dynamic convolutional kernels contained in the multi-kernel convolutional model to obtain the first length-aware text vector corresponding to the recruitment information and the second length-aware text vector corresponding to the business registration information.

[0046] S140. Based on the first length-aware text vector and the second length-aware text vector, determine the recruitment score corresponding to the information to be evaluated.

[0047] The recruitment score indicates whether the recruitment information is reasonable.

[0048] In this embodiment, the recruitment score can indicate whether the recruitment information is reasonable. The recruitment score includes the similarity or relevance between the recruitment information and the business registration information.

[0049] Specifically, the first length-aware text vector and the second length-aware text vector can be integrated, and a recruitment score can be calculated based on the integrated vector. This recruitment score is the score corresponding to the information to be evaluated, and can indicate whether the recruitment information posted by the company is reasonable.

[0050] The technical solution of this invention involves: determining the information to be evaluated; determining a multi-kernel convolutional model including multiple sets of dynamic convolutional kernels; extracting features of the information to be evaluated based on the multi-kernel convolutional model to generate a length-aware text vector corresponding to the information to be evaluated; and determining a recruitment score corresponding to the recruitment information based on the first length-aware text vector and the second length-aware text vector, wherein the recruitment score indicates whether the recruitment information is reasonable. By using a multi-kernel convolutional model including multiple sets of dynamic convolutional kernels to analyze the information to be evaluated, truncation of the information to be evaluated is avoided, reducing information loss during the evaluation process. By obtaining the length-aware text vector of the information to be evaluated and obtaining the recruitment score based on the length-aware text vector, the reasonableness evaluation of the recruitment information in the information to be evaluated is achieved, ensuring the reasonableness of the recruitment information.

[0051] Based on the above embodiments, modified embodiments of the above embodiments are proposed. It should be noted that, in order to keep the description brief, only the differences from the above embodiments are described in the modified embodiments.

[0052] In one embodiment, determining a multi-kernel convolutional model comprising multiple sets of dynamic convolutional kernels includes:

[0053] Extract at least one word vector from the information to be evaluated, and determine the title word vector from the at least one word vector, wherein the title word vector is the vector corresponding to the title of the information to be evaluated;

[0054] For each word vector, calculate the inner product between the word vector and the title word vector to obtain the inner product value corresponding to the word vector;

[0055] The inner product value that satisfies the inner product condition among all the inner product values ​​is taken as the target inner product value, and the position index corresponding to the target inner product value is determined as the size of the first convolution kernel contained in the multiple sets of dynamic convolution kernels;

[0056] Extract at least one sentence information corresponding to the information to be evaluated, wherein the sentence information indicates the length of the sentence corresponding to the information to be evaluated;

[0057] The length of the sentence information that meets the length condition in each of the sentence information is determined as the size of the second convolution kernel contained in the multiple sets of dynamic convolution kernels;

[0058] Determine the squared value of the second convolution kernel and calculate the ratio of the squared value to the first convolution kernel;

[0059] The ratio is rounded down to obtain the size of the third convolutional kernel contained in the plurality of dynamic convolutional kernels;

[0060] The first convolutional kernel, the second convolutional kernel, and the third convolutional kernel are identified as multiple sets of dynamic convolutional kernels contained in the multi-kernel convolutional model.

[0061] In this embodiment, word vectors can be understood as vectors formed by the words that make up the information to be evaluated. Title word vectors can be understood as vectors formed by the titles in the information to be evaluated. Inner product conditions can be understood as conditions used to select target inner product values. Target inner product values ​​can be understood as inner product values ​​that satisfy the inner product conditions, and the target inner product value can be the largest inner product value among all inner product values. Position index can be understood as the index of the target inner product value's position in the information to be evaluated. The first convolution kernel can be understood as the size of the convolution kernel related to segmentation information among multiple sets of dynamic convolution kernels. Sentence information can be understood as information about the sentences that make up the information to be evaluated, and sentence information can include the sentence length. Length conditions can be understood as conditions used to select the size of the second convolution kernel. The second convolution kernel can be understood as the size of the convolution kernel related to sentence information among multiple sets of dynamic convolution kernels. The second convolution kernel can be understood as the size of the convolution kernel related to fine-grained information among multiple sets of dynamic convolution kernels.

[0062] For example, extract at least one word vector that makes up the information to be evaluated. And determine the title word vector in at least one word vector. This can be a vector extracted from the title (e.g., job requirements, business information). For each word vector, the inner product of the word vector and the title word vector is calculated to obtain the inner product value corresponding to the word vector. The inner product value that satisfies the inner product condition among all inner product values ​​is taken as the target inner product value. The inner product condition can be selecting the largest inner product value among all inner product values. The position index corresponding to the target inner product value is then used to determine the size of the first convolutional kernel contained within the multiple sets of dynamic convolutional kernels, i.e., the size of the first convolutional kernel. Extract at least one sentence information corresponding to the sentence that constitutes the information to be evaluated. The length of the sentence information that meets the length condition among all sentence information is determined as the size of the second convolutional kernel contained in the multiple sets of dynamic convolutional kernels. The length condition can be selected as the sentence information with the longest indicator sentence length among all sentence information, i.e., the size of the second convolutional kernel is... Calculate the squared value of the second convolution kernel and its ratio to that of the first convolution kernel. Then, round the ratio to obtain the size of the third convolution kernel contained within the multiple sets of dynamic convolution kernels. ,in This indicates rounding, meaning the size of the third convolution kernel is 1. .

[0063] In one embodiment, determining the recruitment score corresponding to the information to be evaluated based on the first length-aware text vector and the second length-aware text vector includes:

[0064] Concatenate the first length-aware text vector and the second length-aware text vector to obtain a joint vector;

[0065] The joint vector is input into the architecture model to obtain the output vector;

[0066] A mapping operation is performed on the output vector to obtain the recruitment score corresponding to the information to be evaluated.

[0067] In this embodiment, the joint vector can be understood as a vector obtained by concatenating a first length-aware text vector and a second length-aware text vector. The architecture model can be understood as a model used to analyze the joint vector; the architecture model is a deep learning model based on an attention mechanism.

[0068] For example, based on the first length-aware text vector corresponding to the recruitment information. The second length-perceptual text vector corresponding to business registration information Integrate these two vectors into a joint vector. Alternatively, two vectors can be directly concatenated. The joint vector is then input into the architecture model to obtain the output vector. The architecture model can be a pre-trained Transformer model. Then, an activation function is used to map the output vector to a range of 0 to 1, thus obtaining the recruitment score. .

[0069] In one embodiment, determining the information to be evaluated includes:

[0070] Collect initial recruitment information and initial business registration information, wherein the initial recruitment information and the initial business registration information are related information;

[0071] The text corresponding to the initial recruitment information is annotated and segmented to obtain the recruitment information;

[0072] The text corresponding to the initial business registration information is annotated and segmented to obtain the business registration information;

[0073] The recruitment information and the business registration information were identified as information to be evaluated.

[0074] In this embodiment, initial recruitment information can be understood as information collected from companies that relates to their job requirements. Initial business registration information can be understood as information collected from companies that have published initial recruitment information.

[0075] Specifically, initial recruitment information and initial business registration information of enterprises are collected. Initial business registration information refers to the business registration information of the enterprises that publish the initial recruitment information; therefore, initial recruitment information and initial business registration information are related. The texts of the initial recruitment information and initial business registration information are respectively annotated and segmented to obtain the recruitment information and business registration information contained in the information to be evaluated. Subsequently, the reasonableness of the recruitment information can be determined based on the business registration information.

[0076] Example 2

[0077] Figure 2 This is a flowchart of a vector generation method according to Embodiment 2 of the present invention. This embodiment focuses on the length-aware text vector generation method described in the above embodiments. Figure 2 As shown, the method includes:

[0078] S210. Determine the information to be evaluated.

[0079] S220. Determine a multi-kernel convolution model that includes multiple sets of dynamic convolution kernels.

[0080] S230. Based on the multi-kernel convolutional model, extract the features of the information to be evaluated, generate the length-aware text vector corresponding to the information to be evaluated, and perform operations S231-S233 on the recruitment information and the business registration information included in the information to be evaluated respectively:

[0081] S231. Determine the first convolution kernel, the second convolution kernel, and the third convolution kernel of the multiple sets of dynamic convolution kernels contained in the multi-kernel convolution model.

[0082] The first convolution kernel, the second convolution kernel, and the third convolution kernel have different kernel sizes.

[0083] For example, the multiple sets of dynamic convolutional kernels contained in the multi-kernel convolutional model are determined based on the information to be evaluated. The first convolution kernel Second convolution kernel and the third convolution kernel The size of the first convolution kernel is... The size of the second convolution kernel is The size of the third convolution kernel is .

[0084] S232. Based on the multiple sets of dynamic convolution kernels, generate the combination vector corresponding to the information to be evaluated.

[0085] The binding vector includes a first binding vector corresponding to the first convolution kernel, a second binding vector corresponding to the second convolution kernel, and a third binding vector corresponding to the third convolution kernel.

[0086] In this embodiment, the combined vector can be understood as the vector obtained by convolving the information to be evaluated through multiple sets of dynamic convolution kernels.

[0087] Specifically, the information to be evaluated is convolved using multiple sets of dynamic convolutional kernels, including the first, second, and third convolutional kernels, to obtain the first associative vector corresponding to the first convolutional kernel, the second associative vector corresponding to the second convolutional kernel, and the third associative vector corresponding to the third convolutional kernel. Since the information to be evaluated includes recruitment information and business registration information, the first, second, and third associative vectors corresponding to the recruitment information, as well as the first, second, and third associative vectors corresponding to the business registration information, can be obtained.

[0088] Optionally, generating the combination vector corresponding to the information to be evaluated based on the multiple sets of dynamic convolution kernels includes:

[0089] Based on the multiple sets of dynamic convolution kernels, a convolution operation is performed on the information to be evaluated to obtain the text segment to be evaluated.

[0090] The text segment to be evaluated is segment encoded to obtain a segment encoding vector, which indicates the text features and positional information of the text segment to be evaluated;

[0091] The text segment to be evaluated is length encoded to obtain a length encoding vector, which indicates the length feature of the text segment to be evaluated.

[0092] A selective weighting operation is performed on the paragraph encoding vector and the length encoding vector to obtain a combined vector.

[0093] In this embodiment, the text segment to be evaluated can be understood as the text segment obtained after performing a convolution operation on the information to be evaluated. The paragraph encoding vector can be understood as a vector obtained by encoding the text features and positional information of the text segment to be evaluated, and the paragraph encoding vector can represent the global features and positional information of the text segment to be evaluated. The length encoding vector can be understood as a vector obtained by encoding the length feature of the text segment to be evaluated.

[0094] For example, based on the first, second, and third convolutional kernels contained in multiple sets of dynamic convolutional kernels, convolution operations are performed on the information to be evaluated, resulting in three text segments to be evaluated. For each text segment to be evaluated, an enhanced pre-trained language model (pre-trained RoBERTa model) is deployed for that text segment, and the embedding representation fragment vector of the "[CLS]" tag is extracted from the text segment. Based on the fragment vector, the text segment vector corresponding to the information to be evaluated is obtained, which is used to represent the summary information of the text segment to be evaluated. This text segment vector captures the text features of the text segment to be evaluated. The text segment position encoding of the text segment to be evaluated is performed using the following formula, which can be implemented by an encoder: Text Segment Position Encoding ,in, This represents the embedding dimension of the model in the encoder, which can be set to 512. `i` is the column index of the vector scalar, and `pos` is the index of the text segment to be evaluated. The text segment position code is added to the text segment vector and input into the document encoder to obtain the segment code vector corresponding to the text segment to be evaluated.

[0095] To encode the length of the text segment to be evaluated, average pooling can be applied to the segment to obtain a length vector. This length vector is then input into a length encoder to obtain a learnable length-encoded vector. Finally, a selective weighting operation is performed on the segment encoding vector and the length encoding vector to obtain a combined vector. This selective weighting operation can dynamically adjust the influence of the segment encoding vector and the length encoding vector on the combined vector based on different contexts and features.

[0096] By using the first, second, and third convolution kernels, three text segments to be evaluated are obtained. For each text segment to be evaluated, the corresponding combination vector can be obtained.

[0097] Optionally, performing a selective weighting operation on the paragraph encoding vector and the length encoding vector to obtain a combined vector includes:

[0098] The paragraph encoding vector and the length encoding vector are concatenated to obtain the concatenated vector;

[0099] A linear transformation is performed on the concatenated vector, and the result of the linear transformation is mapped to a set interval to obtain a gated vector;

[0100] Using the gated vector, element-wise multiplication is performed on the paragraph encoding vector and the length encoding vector respectively to generate the gated paragraph encoding vector corresponding to the paragraph encoding vector and the gated length encoding vector corresponding to the length encoding vector;

[0101] Element-wise addition is performed on the gated segment encoding vector and the gated length encoding vector to generate a combined vector.

[0102] In this embodiment, the concatenation vector can be understood as the vector obtained by directly concatenating the paragraph encoding vector and the length encoding vector. The gating vector can be used to control the degree of influence of the paragraph encoding vector and the length encoding vector on the combined vector. The gated paragraph encoding vector can be understood as the portion of the paragraph encoding vector in the combined vector determined by the gating vector. The gated length encoding vector can be understood as the portion of the length encoding vector in the combined vector determined by the gating vector.

[0103] For example, the paragraph encoding vector is concatenated with the length encoding vector to obtain the concatenated vector. Apply a linear transformation to the concatenated vectors. ,in It is a learnable weight matrix. As a bias term, the linear transformation projects the concatenated vector into a space with the same dimension as the gate vector; that is, it applies the sigmoid activation function to the result of the linear transformation to obtain the gate vector. ,in It is the sigmoid activation function. The linearly transformed result is mapped to the interval between 0 and 1 using the above method, outputting a gated vector. Gating vectors are used to combine the paragraph encoding vector and the length encoding vector, and gating vectors are applied to the paragraph encoding vector and the length encoding vector respectively. Element-wise multiplication yields the gated segment encoding vector. Gated length encoded vector ,in This represents element-wise multiplication. The gated segment encoding vector and the gated length encoding vector are added element-wise to obtain the final combined vector. .

[0104] S233. Perform pooling operation on the first combined vector, the second combined vector and the third combined vector to generate the length-aware text vector corresponding to the information to be evaluated.

[0105] For example, the three combined vectors corresponding to the information to be evaluated (i.e., the first associative vector) Second associative vector and the third associative vector Perform max pooling to obtain a new vector. The value of this vector in each dimension is the maximum value of the corresponding dimension of the three combined vectors. ,in, Representing vectors The One element. Then for Weighted average pooling is performed to obtain the length-aware text vector corresponding to the information to be evaluated. ,in These are learnable weight parameters that satisfy... Since the information to be evaluated includes recruitment information and business registration information, a first length-aware text vector corresponding to the recruitment information can be generated using the first, second, and third combination vectors corresponding to the recruitment information. ; and the first, second, and third combined vectors corresponding to the business registration information, generating the second length-aware text vector corresponding to the business registration information. .

[0106] S240. Based on the first length-aware text vector and the second length-aware text vector, determine the recruitment score corresponding to the information to be evaluated.

[0107] The technical solution of this invention performs the following operations on the recruitment information and business registration information included in the information to be evaluated: determining the first convolutional kernel, the second convolutional kernel, and the third convolutional kernel, which are multiple sets of dynamic convolutional kernels included in the multi-kernel convolutional model; generating a combined vector corresponding to the information to be evaluated based on the multiple sets of dynamic convolutional kernels; and performing pooling operations on the first combined vector, the second combined vector, and the third combined vector to generate a length-aware text vector corresponding to the information to be evaluated. This refines the method for obtaining the length-aware text vector of the information to be evaluated through a multi-kernel convolutional model. By analyzing the information to be evaluated through multiple sets of dynamic convolutional kernels, truncation of the information to be evaluated is eliminated, reducing information loss during the generation of the length-aware text vector and improving the accuracy of the evaluation.

[0108] Example 3

[0109] Figure 3 This is a schematic diagram of the structure of an information evaluation device according to Embodiment 3 of the present invention. Figure 3 As shown, the device includes:

[0110] The first determining module 310 is used to determine the information to be evaluated, which includes long text information that needs to be evaluated, including recruitment information and business registration information of the company that published the recruitment information;

[0111] The second determining module 320 is used to determine a multi-kernel convolution model including multiple sets of dynamic convolution kernels, wherein the size of the convolution kernels included in the multiple sets of dynamic convolution kernels is associated with the information to be evaluated;

[0112] The generation module 330 is used to extract features of the information to be evaluated based on the multi-kernel convolution model and generate a length-aware text vector corresponding to the information to be evaluated. The length-aware text vector includes a first length-aware text vector corresponding to the recruitment information and a second length-aware text vector corresponding to the business information.

[0113] The third determining module 340 is used to determine the recruitment score corresponding to the information to be evaluated based on the first length-aware text vector and the second length-aware text vector, wherein the recruitment score indicates whether the recruitment information is reasonable.

[0114] The information evaluation device provided in this embodiment of the invention determines the information to be evaluated through a first determining module; determines a multi-kernel convolutional model including multiple sets of dynamic convolutional kernels through a second determining module; extracts features of the information to be evaluated based on the multi-kernel convolutional model through a generation module, generating a length-aware text vector corresponding to the information to be evaluated; and determines a recruitment score corresponding to the information to be evaluated based on the first length-aware text vector and the second length-aware text vector through a third determining module. The recruitment score indicates whether the recruitment information is reasonable. Through the cooperation between the modules, and by using a multi-kernel convolutional model including multiple sets of dynamic convolutional kernels to analyze the information to be evaluated, truncation of the information to be evaluated is avoided, and information loss during the evaluation process is reduced. By obtaining the length-aware text vector of the information to be evaluated and obtaining the recruitment score based on the length-aware text vector, the reasonableness evaluation of the recruitment information in the information to be evaluated is realized, ensuring the reasonableness of the recruitment information.

[0115] In one embodiment, the generation module 330 includes:

[0116] Perform the following operations on the recruitment information and business registration information included in the information to be evaluated:

[0117] The determining unit is used to determine the first convolution kernel, the second convolution kernel, and the third convolution kernel of the multiple sets of dynamic convolution kernels contained in the multi-kernel convolution model, wherein the first convolution kernel, the second convolution kernel, and the third convolution kernel have different convolution kernel sizes;

[0118] The generation unit is used to generate a combination vector corresponding to the information to be evaluated based on the multiple sets of dynamic convolution kernels. The combination vector includes a first combination vector corresponding to the first convolution kernel, a second combination vector corresponding to the second convolution kernel, and a third combination vector corresponding to the third convolution kernel.

[0119] A pooling unit is used to perform pooling operations on the first binding vector, the second binding vector, and the third binding vector to generate a length-aware text vector corresponding to the information to be evaluated.

[0120] In one embodiment, the generating unit includes:

[0121] A convolutional subunit is used to perform a convolution operation on the information to be evaluated based on the multiple sets of dynamic convolutional kernels to obtain the text segment to be evaluated.

[0122] A paragraph encoding subunit is used to perform paragraph encoding on the text segment to be evaluated to obtain a paragraph encoding vector, wherein the paragraph encoding vector indicates the text features and positional information of the text segment to be evaluated;

[0123] A length encoding subunit is used to encode the length of the text segment to be evaluated to obtain a length encoding vector, wherein the length encoding vector indicates the length feature of the text segment to be evaluated.

[0124] The weighting subunit is used to perform a selective weighting operation on the segment encoding vector and the length encoding vector to obtain a combined vector.

[0125] In one embodiment, the weighted subunit is specifically used for:

[0126] The paragraph encoding vector and the length encoding vector are concatenated to obtain the concatenated vector;

[0127] A linear transformation is performed on the concatenated vector, and the result of the linear transformation is mapped to a set interval to obtain a gated vector;

[0128] Using the gated vector, element-wise multiplication is performed on the paragraph encoding vector and the length encoding vector respectively to generate the gated paragraph encoding vector corresponding to the paragraph encoding vector and the gated length encoding vector corresponding to the length encoding vector;

[0129] Element-wise addition is performed on the gated segment encoding vector and the gated length encoding vector to generate a combined vector.

[0130] In one embodiment, the second determining module 320 is specifically used for:

[0131] Extract at least one word vector from the information to be evaluated, and determine the title word vector from the at least one word vector, wherein the title word vector is the vector corresponding to the title of the information to be evaluated;

[0132] For each word vector, calculate the inner product between the word vector and the title word vector to obtain the inner product value corresponding to the word vector;

[0133] The inner product value that satisfies the inner product condition among all the inner product values ​​is taken as the target inner product value, and the position index corresponding to the target inner product value is determined as the size of the first convolution kernel contained in the multiple sets of dynamic convolution kernels;

[0134] Extract at least one sentence information corresponding to the information to be evaluated, wherein the sentence information indicates the length of the sentence corresponding to the information to be evaluated;

[0135] The length of the sentence information that meets the length condition in each of the sentence information is determined as the size of the second convolution kernel contained in the multiple sets of dynamic convolution kernels;

[0136] Determine the squared value of the second convolution kernel and calculate the ratio of the squared value to the first convolution kernel;

[0137] The ratio is rounded down to obtain the size of the third convolutional kernel contained in the plurality of dynamic convolutional kernels;

[0138] The first convolutional kernel, the second convolutional kernel, and the third convolutional kernel are identified as multiple sets of dynamic convolutional kernels contained in the multi-kernel convolutional model.

[0139] In one embodiment, the third determining module 340 is specifically used for:

[0140] Concatenate the first length-aware text vector and the second length-aware text vector to obtain a joint vector;

[0141] The joint vector is input into the architecture model to obtain the output vector;

[0142] A mapping operation is performed on the output vector to obtain the recruitment score corresponding to the information to be evaluated.

[0143] In one embodiment, the first determining module 310 is specifically used for:

[0144] Collect initial recruitment information and initial business registration information, wherein the initial recruitment information and the initial business registration information are related information;

[0145] The text corresponding to the initial recruitment information is annotated and segmented to obtain the recruitment information;

[0146] The text corresponding to the initial business registration information is annotated and segmented to obtain the business registration information;

[0147] The recruitment information and the business registration information were identified as information to be evaluated.

[0148] The information evaluation device provided in this embodiment of the invention can execute the information evaluation method provided in any embodiment of the invention. Through the cooperation and coordination between the modules, the evaluation of information is completed, and it has the corresponding functional modules and beneficial effects of the execution method.

[0149] Example 4

[0150] According to embodiments of the present invention, the present invention also provides an electronic device, a computer-readable storage medium, and a computer program product.

[0151] Figure 4 This is a block diagram of an electronic device according to Embodiment 4 of the present invention, which implements the information evaluation method described in the embodiments of the present invention. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0152] like Figure 4 As shown, the electronic device 410 includes at least one processor 411 and a memory, such as a read-only memory (ROM) 412 or a random access memory (RAM) 413, communicatively connected to the at least one processor 411. The memory stores computer programs executable by the at least one processor. The processor 411 can perform various appropriate actions and processes based on the computer program stored in the ROM 412 or loaded from storage unit 418 into the RAM 413. The RAM 413 may also store various programs and data required for the operation of the electronic device 410. The processor 411, ROM 412, and RAM 413 are interconnected via a bus 414. An input / output (I / O) interface 415 is also connected to the bus 414.

[0153] Multiple components in the electronic device are connected to the I / O interface 415, including: an input unit 416, such as a keyboard, mouse, etc.; an output unit 417, such as various types of displays, speakers, etc.; a storage unit 418, such as a disk, optical disk, etc.; and a communication unit 419, such as a network card, modem, wireless transceiver, etc. The communication unit 419 allows the electronic device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0154] Processor 411 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 411 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 411 performs the various methods and processes described above, such as information evaluation methods.

[0155] In some embodiments, the information evaluation method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 418. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 410 via ROM 412 and / or communication unit 419. When the computer program is loaded into RAM 413 and executed by processor 411, one or more steps of the information evaluation method described above may be performed. Alternatively, in other embodiments, processor 411 may be configured to perform the information evaluation method by any other suitable means (e.g., by means of firmware).

[0156] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0157] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0158] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0159] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0160] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0161] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0162] In some embodiments, the computer program product includes a computer program that, when executed by a processor, implements the information evaluation method provided in the embodiments of the present invention.

[0163] The technical solution of this invention provides an information evaluation method, apparatus, electronic device, storage medium, and program product. The process involves: determining information to be evaluated; determining a multi-kernel convolutional model including multiple sets of dynamic convolutional kernels; extracting features from the information to be evaluated based on the multi-kernel convolutional model to generate a length-aware text vector corresponding to the information to be evaluated; and determining a recruitment score corresponding to the recruitment information based on the first length-aware text vector and the second length-aware text vector, whereby the recruitment score indicates whether the recruitment information is reasonable. By using a multi-kernel convolutional model including multiple sets of dynamic convolutional kernels to analyze the information to be evaluated, truncation of the information to be evaluated is avoided, reducing information loss during the evaluation process. By obtaining the length-aware text vector of the information to be evaluated and obtaining the recruitment score based on the length-aware text vector, the reasonableness evaluation of the recruitment information in the information to be evaluated is achieved, ensuring the reasonableness of the recruitment information.

[0164] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0165] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. An information evaluation method characterized by, The method comprises the following steps: determining to-be-evaluated information, wherein the to-be-evaluated information comprises long text information that needs to be evaluated, and the to-be-evaluated information comprises recruitment information and business information of an enterprise publishing the recruitment information; determining a multi-kernel convolution model comprising multiple groups of dynamic convolution kernels, wherein the multiple groups of dynamic convolution kernels comprise convolution kernels associated with the to-be-evaluated information in terms of kernel size; extracting features of the to-be-evaluated information based on the multi-kernel convolution model, and generating a length-aware text vector corresponding to the to-be-evaluated information, wherein the length-aware text vector comprises a first length-aware text vector corresponding to the recruitment information and a second length-aware text vector corresponding to the business information; determining a recruitment score corresponding to the to-be-evaluated information based on the first length-aware text vector and the second length-aware text vector, wherein the recruitment score indicates whether the recruitment information is reasonable.

2. The method of claim 1, wherein, The step of extracting features of the to-be-evaluated information based on the multi-kernel convolution model and generating a length-aware text vector corresponding to the to-be-evaluated information comprises the following steps of: respectively performing the following operations on the recruitment information and the business information included in the to-be-evaluated information: determining a first convolution kernel, a second convolution kernel and a third convolution kernel of the multiple groups of dynamic convolution kernels comprised in the multi-kernel convolution model, wherein the first convolution kernel, the second convolution kernel and the third convolution kernel are different in terms of kernel size; generating a combination vector corresponding to the to-be-evaluated information based on the multiple groups of dynamic convolution kernels, wherein the combination vector comprises a first combination vector corresponding to the first convolution kernel, a second combination vector corresponding to the second convolution kernel, and a third combination vector corresponding to the third convolution kernel; performing a pooling operation on the first combination vector, the second combination vector and the third combination vector to generate a length-aware text vector corresponding to the to-be-evaluated information.

3. The method of claim 2, wherein, The step of generating a combination vector corresponding to the to-be-evaluated information based on the multiple groups of dynamic convolution kernels comprises the following steps of: performing a convolution operation on the to-be-evaluated information based on the multiple groups of dynamic convolution kernels to obtain a to-be-evaluated text segment; performing paragraph encoding on the to-be-evaluated text segment to obtain a paragraph encoding vector, wherein the paragraph encoding vector indicates text features and position information of the to-be-evaluated text segment; performing length encoding on the to-be-evaluated text segment to obtain a length encoding vector, wherein the length encoding vector indicates length features of the to-be-evaluated text segment; performing a selective weighting operation on the paragraph encoding vector and the length encoding vector to obtain a combination vector.

4. The method of claim 3, wherein, The step of performing a selective weighting operation on the paragraph encoding vector and the length encoding vector to obtain a combination vector comprises the following steps of: splicing the paragraph encoding vector and the length encoding vector to obtain a spliced vector; performing linear transformation on the spliced vector, and mapping a result of the linear transformation to a set interval to obtain a gating vector; performing element-wise multiplication on the paragraph encoding vector and the length encoding vector respectively by using the gating vector to generate a gated paragraph encoding vector corresponding to the paragraph encoding vector and a gated length encoding vector corresponding to the length encoding vector. Performing element-wise addition on the gated passage encoding vector and the gated length encoding vector generates a combined vector.

5. The method of claim 1, wherein, The determination includes a multi-kernel convolution model including multiple groups of dynamic convolution kernels, which includes: Extract at least one word vector of the to-be-evaluated information, and determine a title word vector in the at least one word vector, the title word vector being a vector corresponding to a title of the to-be-evaluated information; For each word vector, calculate the inner product of the word vector and the title word vector to obtain an inner product value corresponding to the word vector; Determine, as a target inner product value, an inner product value that meets an inner product condition among the inner product values, and determine, as a size of a first convolution kernel included in the multiple groups of dynamic convolution kernels, a position index corresponding to the target inner product value; Extract at least one sentence information corresponding to the to-be-evaluated information, the sentence information indicating a length of a sentence corresponding to the to-be-evaluated information; Determine, as a size of a second convolution kernel included in the multiple groups of dynamic convolution kernels, a length of sentence information that meets a length condition among the sentence information; Determine a square value of the second convolution kernel, and calculate a ratio of the square value to the first convolution kernel; Take the ratio to an integer to obtain a size of a third convolution kernel included in the multiple groups of dynamic convolution kernels; Determine the first convolution kernel, the second convolution kernel, and the third convolution kernel as the multiple groups of dynamic convolution kernels included in the multi-kernel convolution model.

6. The method of claim 1, wherein, The determination of the recruitment score corresponding to the to-be-evaluated information based on the first length-aware text vector and the second length-aware text vector includes: Concatenate the first length-aware text vector and the second length-aware text vector to obtain a joint vector; Input the joint vector into an architecture model to obtain an output vector; Perform a mapping operation on the output vector to obtain the recruitment score corresponding to the to-be-evaluated information.

7. The method of claim 1, wherein, The determination of the to-be-evaluated information includes: Collect initial recruitment information and initial business information, the initial recruitment information and the initial business information being information having relevance; Perform labeling and word segmentation processing on a text corresponding to the initial recruitment information to obtain recruitment information; Perform labeling and word segmentation processing on a text corresponding to the initial business information to obtain business information; Determine the recruitment information and the business information as to-be-evaluated information.

8. An information evaluation device, characterized by comprising: It includes: A first determination module is configured to determine to-be-evaluated information, the to-be-evaluated information including long text information that needs to be evaluated, and the to-be-evaluated information including recruitment information and business information of an enterprise publishing the recruitment information; A second determination module is configured to determine a multi-kernel convolution model including multiple groups of dynamic convolution kernels, sizes of convolution kernels included in the multiple groups of dynamic convolution kernels being associated with the to-be-evaluated information; A generation module is configured to extract features of the to-be-evaluated information based on the multi-kernel convolution model, and generate length-aware text vectors corresponding to the to-be-evaluated information, the length-aware text vectors including a first length-aware text vector corresponding to the recruitment information and a second length-aware text vector corresponding to the business information; A third determining module is configured to determine a recruitment score corresponding to the to-be-evaluated information based on the first length-aware text vector and the second length-aware text vector, the recruitment score indicating whether the recruitment information is reasonable.

9. An electronic device, comprising: The electronic device comprises: at least one processor; and a memory connected with the at least one processor in communication; wherein The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to perform the information evaluation method in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions for enabling the processor to perform the information evaluation method in any one of claims 1-7 when executed.

11. A computer program product, characterised in that, The computer program product comprises a computer program, and the computer program, when executed by the processor, implements the information evaluation method according to any one of claims 1-7.

Images