A target text scoring system

Abstract

The present application relates to the technical field of task scoring, and in particular to a target text scoring system, the system comprising: a storage medium storing a preset task type set, a historical task item text list and a historical task item weight, a processor and a memory storing a computer program, wherein when the computer program is executed by the processor, each intermediate task item text in the initial task text and the historical task item text most similar to each intermediate task item text are obtained, and the weight of each intermediate task item text is calculated according to the weight of the historical task item text, and then the task item feedback text corresponding to each intermediate task item text is obtained from the received task feedback text, and the final score of the task feedback text is calculated according to the score of each task item feedback text and the weight of each intermediate task item text, which can reflect the importance of the task feedback text to a large extent and can objectively reflect the work quality and efficiency of the user.

Description

Technical Field

[0001] The present invention relates to the technical field of task scoring, and in particular to a target text scoring system. Background Art

[0002] In the working process of some enterprises and institutions, leaders or customers usually issue many work tasks, and upload these tasks to the internal work system of the enterprise or institution in the form of documents, and then distribute them to the target department or a specific user through the system. The user gives corresponding feedback based on the content of the task, and the feedback results are sent to the auditor in the form of text or briefing. The auditor scores according to the content of the feedback. However, when the auditor scores the task execution results of each user, there is a great deal of subjectivity, which can easily cause scoring errors and lead to inaccurate scoring. In addition, when it is necessary to score the feedback results of multiple users, the labor cost is high and the scoring efficiency is low. Summary of the invention

[0003] In view of the above technical problems, the technical solution adopted by the present invention is:

[0004] A target text scoring system, the system comprising: a preset task type set, a historical task item text list set A={A1, A2, ..., A i , ..., A m} and the historical task item weight set A corresponding to A 0 ={A 0 1. A 0 2, ..., A 0 i , ..., A 0 m} storage medium, processor and memory storing a computer program, the preset task type set is in a tree structure, and includes preset position types, preset work types and preset task item types from the upper level to the lower level, and the preset task item types correspond to the historical task item text list one by one, wherein A i ={A i1 , A i2 , ..., A ie , ..., A if(i)}, where A ie is the e-th historical task item text corresponding to the i-th preset task item type, i=1, 2, ..., m, m is the number of historical task item text lists, e=1, 2, ..., f(i), f(i) is the number of historical task item texts corresponding to the i-th preset task item type, A 0 i A i The corresponding historical task item weights, when the computer program is executed by the processor, implement the following steps:

[0005] S100, according to the received initial task text, obtain the intermediate task item text list B = {B1, B2, ..., B j , ..., B n}, where B j is the j-th task item text obtained from the initial task text, j=1, 2, …, n, and n is the number of task item texts obtained from the initial task text.

[0006] S200, through the target classification model and target entity recognition model, from B j The type dataset and task entity label set C = {C1, C2, ..., C p , ..., C q}; The type of data set includes B j Corresponding target position type, target work type and target task item type; C p is the pth task entity label, p=1, 2, ..., q, q is the number of task entity labels; the target position type, target work type and target task item type are respectively obtained from the preset position type, preset work type and preset task item type.

[0007] S300, obtain from A a key task item text list D corresponding to the target task item type = {D1, D2, ..., D h , ..., D k}, and according to the target entity recognition model, from D h Get D h The corresponding key entity label set D 0 h ={D 0 h1 , D 0 h2 , ..., D 0 hp , ..., D 0 hq}, where D h is the hth key task item text corresponding to the target task item type, h=1, 2, ..., k, k is the number of key task item texts corresponding to the target task item type, and the key task item text is any historical task item text corresponding to the target task item type. 0 hp D h The corresponding p-th key entity label.

[0008] S400, according to C and D 0 h , determine the target task item text from D; the target task item text refers to the text related to B jThe key task item text with the highest similarity.

[0009] S500, from A 0 The historical task item weight A of the target task item text is obtained in 0 x , and according to the historical task item weights of the target task item text corresponding to each intermediate task item text, obtain the intermediate task item weight set B corresponding to B 0 ={B 0 1, B 0 2, ..., B 0 j , ..., B 0 n}, where B 0 j For B j The corresponding intermediate task item weight.

[0010] S600, according to the received task feedback text, obtain the task item feedback text list G corresponding to B = {G1, G2, ..., G j , ..., G n}, where G j For B j The corresponding feedback text.

[0011] S700, using the preset automatic scoring model to j Score and obtain the text score list G corresponding to G 0 ={G 0 1, G 0 2, ..., G 0 j , ..., G 0 n}, where G 0 j G j The corresponding text score.

[0012] S800, according to B 0 and G 0 , get the final score W of the task feedback text.

[0013] Compared with the prior art, the present invention has obvious beneficial effects. By means of the above technical solution, a target text scoring system provided by the present invention can achieve considerable technical advancement and practicality, and has wide industrial utilization value, and has at least the following beneficial effects:

[0014] The present invention provides a target text scoring system. When the computer program is executed by a processor, each intermediate task item text in the initial task text is first obtained, and then the target task item type and task entity label corresponding to each intermediate task item text are identified, and a plurality of historical task item texts under each target task item type are obtained from the historical task item text, and the historical task item text most similar to the intermediate task item text is used as the target task item text, and then the weight of each intermediate task item text is calculated according to the weight corresponding to the target task item text, and the importance of each intermediate task item can be determined by obtaining the weight of each intermediate task item text in the initial task text; then, the task item feedback text corresponding to each intermediate task item text is obtained from the received task feedback text, and the final score of the task feedback text is obtained by weighted calculation according to the score of each task item feedback text and the weight of each intermediate task item text, which can reflect the importance of the task feedback text to a large extent, and effectively reduce the influence of subjective factors on the scoring in the scoring process, and can objectively reflect the work quality and efficiency of the user, while improving the scoring efficiency and saving labor costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0016] Figure 1 A flowchart of a target text scoring system executing a computer program provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0018] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0019] A target text scoring system, the system comprising: a preset task type set, a historical task item text list set A={A1, A2, ..., A i , ..., A m} and the historical task item weight set A corresponding to A 0 ={A 0 1. A 0 2, ..., A 0 i , ..., A 0 m} storage medium, processor and memory storing a computer program, the preset task type set is in a tree structure, and includes preset position types, preset work types and preset task item types from the upper level to the lower level, and the preset task item types correspond to the historical task item text list one by one, wherein A i ={A i1 , A i2 , ..., A ie , ..., A if(i)}, where A ie is the e-th historical task item text corresponding to the i-th preset task item type, i=1, 2, ..., m, m is the number of historical task item text lists, e=1, 2, ..., f(i), f(i) is the number of historical task item texts corresponding to the i-th preset task item type, A 0 i A i The corresponding historical task item weights, when the computer program is executed by the processor, implement the following steps, such as Figure 1 As shown:

[0020] S100, according to the received initial task text, obtain the intermediate task item text list B = {B1, B2, ..., B j , ..., B n}, where B jis the j-th task item text obtained from the initial task text, j=1, 2, …, n, and n is the number of task item texts obtained from the initial task text.

[0021] Specifically, the preset position type includes several position titles under the preset position type, and the position title is any position title pre-set according to the enterprise business system, for example: science and technology position.

[0022] Specifically, the preset work type includes several work names under the preset work type, and the work name is any work name preset according to the enterprise business system, for example: application development work under the science and technology post.

[0023] Specifically, the preset task item type includes several task item names under the preset task item type, and the task item name is any task item name preset according to the enterprise business system, for example: a development research and judgment model under application development work.

[0024] Specifically, the initial task text refers to a text including several task items issued by a target person; wherein the target person refers to the person who issues the initial task text at that time, for example, a leader who issues a task.

[0025] In a specific embodiment, in S100, B is obtained by the following steps:

[0026] S101, using an OCR character recognition model to perform text recognition on the initial task text to obtain a recognition result text.

[0027] Specifically, those skilled in the art are aware of the specific implementation process of the OCR character recognition model, which will not be described in detail here.

[0028] S102, using a text data cleaning model to clean the recognition result text to obtain an intermediate task text; this can be understood as: cleaning the garbled characters, special characters, typos, spaces, etc. in the recognition result text to obtain a text that is readable and in a regular format.

[0029] Specifically, the text data cleaning model refers to a data cleaning model trained based on a natural language processing model, wherein the natural language processing model may be a BERT model. Those skilled in the art are aware of the training process of obtaining a text data cleaning model based on BERT model training, which will not be repeated here.

[0030] S103, analyze the intermediate task text through the preset semantic model to obtain B; it can be understood as: semantic analysis is performed on the keywords in the intermediate task text, and when keywords related to the command are obtained, the task content corresponding to the keyword is used as the intermediate task item text. For example, the keywords related to the command may be "execution", "analysis", "summary", "coordination", "response", etc.

[0031] As described above, by cleaning and semantically analyzing the initial task text, several task items with smooth reading and clear themes can be extracted from the initial task text, so as to obtain the number of task items and the specific content of each task item, which is conducive to the subsequent analysis of each task item and the completion results of each task item.

[0032] S200, through the target classification model and target entity recognition model, from B j The type dataset and task entity label set C = {C1, C2, ..., C p , ..., C q}; The type of data set includes B j Corresponding target position type, target work type and target task item type; C p is the p-th task entity label, p=1, 2, …, q, q is the number of task entity labels.

[0033] Specifically, in S200, the type data set is obtained through the following steps:

[0034] S201, from B j A number of task keywords are obtained from the text. Those skilled in the art can know that any method of extracting keywords from text in the prior art falls within the protection scope of the present invention and will not be described in detail here.

[0035] S202, compare each task keyword with each preset job keyword in the preset job keyword library to obtain the target job type in the type data set; the preset job keyword refers to any keyword pre-set and corresponding to the preset job type; it can be understood that: one preset job type corresponds to multiple job keywords, when there is a task keyword that is consistent with the preset job keyword or has the same semantics, the preset job type corresponding to the preset job keyword is used as the target job type.

[0036] S203, compare each task keyword with each preset job type word in the preset job type vocabulary corresponding to the target job type, and obtain the target job type in the type data set. The acquisition method is the same as the target job type acquisition method, which will not be repeated here.

[0037] Specifically, the preset job type word library corresponding to the target job type refers to a library composed of job type words corresponding to several preset job types under the target job type.

[0038] S204, compare each task keyword with each preset task item type word in the preset task item type vocabulary corresponding to the target job type, and obtain the target task item type in the type data set. The acquisition method is consistent with the acquisition method of the target position type, which will not be repeated here.

[0039] Specifically, the target entity recognition model refers to an entity recognition model trained based on the BERT model and used to identify task entity labels in task items; those skilled in the art are aware of the training process of obtaining the target entity recognition model based on the BERT model training, which will not be repeated here.

[0040] Specifically, the task entity label refers to any entity label identified from any intermediate task item text; the task entity label set includes a task time requirement label, an output requirement label, a workload label and several task dependency data labels.

[0041] As mentioned above, through the target classification model and the target entity recognition model, the text content of the intermediate task item text can be disassembled and analyzed, and then the task entity label and type data corresponding to each intermediate task item text can be obtained, so as to map these data to the historical task text, and then obtain the historical task text with the highest similarity to each intermediate task item text.

[0042] S300, obtain from A a key task item text list D corresponding to the target task item type = {D1, D2, ..., D h , ..., D k}, and according to the target entity recognition model, from D h Get D h The corresponding key entity label set D 0 h ={D 0 h1 , D 0 h2 , ..., D 0 hp , ..., D 0 hq}, where D h is the hth key task item text corresponding to the target task item type, h=1, 2, ..., k, k is the number of key task item texts corresponding to the target task item type, and the key task item text is any historical task item text corresponding to the target task item type. 0 hp D hThe corresponding p-th key entity label.

[0043] Specifically, the method for obtaining the key entity tags is consistent with the method for obtaining the task entity tags, which will not be repeated here.

[0044] S400, according to C and D 0 h , determine the target task item text from D; the target task item text refers to the text related to B j The key task item text with the highest similarity.

[0045] In a specific embodiment, in S400, the target task item text is obtained by the following steps:

[0046] S401, sequentially concatenate the q task entity tags in C to obtain the first concatenated text, and then concatenate the q task entity tags in D 0 h The q key entity tags in are concatenated in sequence to obtain the second concatenated text.

[0047] S402, obtaining a target similarity ζ between the first concatenated text and the second concatenated text.

[0048] Specifically, the target similarity ζ meets the following conditions:

[0049] ζ=(α·β) / (||α||×||β||), where α is the word vector corresponding to the first concatenated text, and β is the word vector corresponding to the second concatenated text.

[0050] S403, when the target similarity ζ is the maximum similarity corresponding to C, D h Determine the target task item text.

[0051] In the above, the task entity labels are spliced, and the key entity label sets corresponding to each key task item text are spliced ​​separately, which realizes the combination of keywords for each text, and can calculate the similarity between the intermediate task item text and each corresponding historical task item text, and then find the historical task item text that is most similar to the intermediate task item text.

[0052] S500, from A 0 The historical task item weight A of the target task item text is obtained in 0 x , and according to the historical task item weights of the target task item text corresponding to each intermediate task item text, obtain the intermediate task item weight set B corresponding to B 0 ={B 0 1, B 0 2, ..., B 0 j , ..., B 0n}, where B 0 j For B j The corresponding intermediate task item weight.

[0053] Specifically, B 0 j Meet the following conditions:

[0054] B 0 j =0.8×ε+0.2×ζ×A 0 x , where ε is the preset initial task weight; it can be understood that when the target personnel issues a task, they will pre-set an initial task weight according to the importance of the task.

[0055] As mentioned above, when determining the historical task item texts corresponding to each intermediate task item text, the weight of the historical task item text is assigned as a parameter to the calculation of some intermediate task item weights based on the similarity between the two, and combined with the importance of the intermediate task item text itself, the final intermediate task item weight is obtained to comprehensively grasp the importance of each intermediate task item text.

[0056] S600, according to the received task feedback text, obtain the task item feedback text list G corresponding to B = {G1, G2, ..., G j , ..., G n}, where G j For B j The corresponding feedback text.

[0057] Specifically, the task feedback text refers to the text of feedback provided by the target user based on the initial task text, including the initial task text and the feedback content corresponding to each intermediate task item text in the initial task text; it can be understood that: the task feedback text includes each intermediate task item text in the initial task text, and the corresponding feedback content is recorded at the corresponding position of each intermediate task item text, and the natural language processing model can be used to obtain the feedback content corresponding to each intermediate task item text.

[0058] S700, using the preset automatic scoring model to j Score and obtain the text score list G corresponding to G 0 ={G 0 1, G 0 2, ..., G 0 j , ..., G 0 n}, where G 0 j G jThe corresponding text score.

[0059] In a specific embodiment, in S700, the automatic scoring model is obtained by the following steps:

[0060] S701, obtaining a number of historical task item feedback texts and score data of each historical task item feedback text, and preprocessing each historical task item feedback text and each score data; it can be understood as: cleaning, denoising, etc. of the historical task item feedback text and score data.

[0061] Specifically, the historical task item feedback text refers to the feedback text corresponding to the historical task item text.

[0062] S702, segmenting the processed historical task item feedback text to form a plurality of entries, and using the plurality of entries as feature data.

[0063] S703: Use a supervised machine learning model to learn the feature data and the score data to train the machine learning model to form an automatic scoring model.

[0064] As described above, based on the historical task item feedback texts and their corresponding score data as training samples, the machine learning model can be optimized to obtain an automatic scoring model with more realistic scoring, thereby obtaining a more accurate and objective text score for each task item feedback text in the task feedback text.

[0065] S800, according to B 0 and G 0 , get the final score W of the task feedback text.

[0066] Specifically, W meets the following conditions:

[0067] W=∑ n j=1 (B 0 j ×G 0 j ).

[0068] As mentioned above, the final score of the task feedback text calculated through the weight of each task item in the initial task text and the score of the task feedback text corresponding to each task item can largely reflect the importance of the task feedback text and the user's execution effect, and effectively reduce the influence of subjective factors on the scoring process, and can objectively reflect the user's work quality and efficiency, while improving the scoring efficiency and saving labor costs.

[0069] Although some specific embodiments of the present invention have been described in detail by way of example, it will be appreciated by those skilled in the art that the above examples are for illustration only and are not intended to limit the scope of the present invention. It will also be appreciated by those skilled in the art that various modifications may be made to the embodiments without departing from the scope and spirit of the present invention. The scope of the present invention is defined by the appended claims.

Claims

1. A target text scoring system, characterized in that: The system includes: a preset task type set, a historical task item text list set A={A1, A2, . . . , A i , ..., A m } and the historical task item weight set A corresponding to A 0 ={A 0 1. A 0 2, ..., A 0 i , ..., A 0 m } storage medium, processor and memory storing a computer program, the preset task type set is in a tree structure, and includes preset position types, preset work types and preset task item types from the upper level to the lower level, and the preset task item types correspond to the historical task item text list one by one, wherein A i ={A i1 , A i2 , ..., A ie , ..., A if(i) }, where A ie is the e-th historical task item text corresponding to the i-th preset task item type, i=1, 2, ..., m, m is the number of historical task item text lists, e=1, 2, ..., f(i), f(i) is the number of historical task item texts corresponding to the i-th preset task item type, A 0 i A i The corresponding historical task item weights, when the computer program is executed by the processor, implement the following steps: S100, according to the received initial task text, obtain the intermediate task item text list B = {B1, B2, ..., B j , ..., B n }, where B j is the jth task item text obtained from the initial task text, j=1, 2, ..., n, n is the number of task item texts obtained from the initial task text; S200, through the target classification model and target entity recognition model, from B j The type dataset and task entity label set C = {C1, C2, ..., C p , ..., C q }; The type of data set includes B j Corresponding target position type, target work type and target task item type; C p is the pth task entity label, p=1, 2, ..., q, q is the number of task entity labels; the target position type, target work type and target task item type are obtained from the preset position type, preset work type and preset task item type respectively; S300, obtain from A a key task item text list D corresponding to the target task item type = {D1, D2, ..., D h , ..., D k }, and according to the target entity recognition model, from D h Get D h The corresponding key entity label set D 0 h ={D 0 h1 , D 0 h2 , ..., D 0 hp , ..., D 0 hq }, where D h is the hth key task item text corresponding to the target task item type, h=1, 2, ..., k, k is the number of key task item texts corresponding to the target task item type, and the key task item text is any historical task item text corresponding to the target task item type. 0 hp D h The corresponding p-th key entity label; S400, according to C and D 0 h , determine the target task item text from D; the target task item text refers to the text related to B j The key task item text with the highest similarity; S500, from A 0 The historical task item weight A of the target task item text is obtained in 0 x , and according to the historical task item weights of the target task item text corresponding to each intermediate task item text, obtain the intermediate task item weight set B corresponding to B 0 ={B 0 1, B 0 2, ..., B 0 j , ..., B 0 n }, where B 0 j For B j The corresponding intermediate task item weights; S600, according to the received task feedback text, obtain the task item feedback text list G corresponding to B = {G1, G2, ..., G j , ..., G n }, where G j For B j Corresponding feedback text; the task feedback text refers to the text of the feedback content corresponding to the initial task text and each intermediate task item text in the initial task text, which is fed back by the target user according to the initial task text; S700, using the preset automatic scoring model to j Score and obtain the text score list G corresponding to G 0 ={G 0 1, G 0 2, ..., G 0 j , ..., G 0 n }, where G 0 j G j The corresponding text score; S800, according to B 0 and G 0 , get the final score W of the task feedback text.

2. The target text scoring system according to claim 1, characterized in that: The initial task text refers to the text published by the target person and includes several task items.

3. The target text scoring system according to claim 1, characterized in that: In S100, B is obtained by the following steps: S101, using an OCR character recognition model to perform text recognition on the initial task text to obtain a recognition result text; S102, using a text data cleaning model to perform data cleaning on the recognition result text to obtain an intermediate task text; S103, analyzing the intermediate task text through a preset semantic model to obtain B.

4. The target text scoring system according to claim 1, characterized in that: In S200, the type data set is obtained through the following steps: S201, from B j Get several task keywords from S202, comparing each task keyword with each preset position keyword in the preset position keyword library to obtain the target position type in the type data set; The preset job keyword refers to any keyword that is preset and corresponds to the preset job type; S203, comparing each task keyword with each preset job type word in a preset job type word library corresponding to the target job type, and obtaining the target job type in the type data set; S204, comparing each task keyword with each preset task item type word in a preset task item type word library corresponding to the target work type, and acquiring the target task item type in the type data set.

5. The target text scoring system according to claim 1, characterized in that: In S300, the key entity tag is obtained in the same manner as the task entity tag.

6. The target text scoring system according to claim 1, characterized in that: In S400, the target task item text is obtained through the following steps: S401, sequentially concatenate the q task entity tags in C to obtain the first concatenated text, and then concatenate the q task entity tags in D 0 h The q key entity tags in are concatenated in sequence to obtain the second concatenated text; S402, obtaining a target similarity ζ between the first concatenated text and the second concatenated text; Among them, the target similarity ζ meets the following conditions: ζ=(α·β) / (||α||×||β||), where α is the word vector corresponding to the first concatenated text, and β is the word vector corresponding to the second concatenated text; S403, when the target similarity ζ is the maximum similarity corresponding to C, D h Determine the target task item text.

7. The target text scoring system according to claim 6, characterized in that: In S500, B 0 j Meet the following conditions: B 0 j =0.8×ε+0.2×ζ×A 0 x , where ε is the preset initial task weight.

8. The target text scoring system according to claim 1, characterized in that: The task feedback text refers to the text of the feedback content corresponding to the initial task text and each intermediate task item text in the initial task text, which is fed back by the target user according to the initial task text.

9. The target text scoring system according to claim 1, characterized in that: In S700, the automatic scoring model is obtained by following the steps below: S701, obtaining a plurality of historical task item feedback texts and score data of each historical task item feedback text, and preprocessing each historical task item feedback text and each score data; S702, segmenting the processed historical task item feedback text to form a plurality of entries, and using the plurality of entries as feature data; S703: Use a supervised machine learning model to learn the feature data and the score data to train the machine learning model to form an automatic scoring model.

10. The target text scoring system according to claim 1, characterized in that: In S800, W meets the following conditions: W=∑ n j=1 (B 0 j ×G 0 j )。

Images