WO2021139265A1 - 作文评分方法、装置、计算机设备及计算机可读存储介质 - Google Patents

作文评分方法、装置、计算机设备及计算机可读存储介质 Download PDF

Info

Publication number
WO2021139265A1
WO2021139265A1 PCT/CN2020/118737 CN2020118737W WO2021139265A1 WO 2021139265 A1 WO2021139265 A1 WO 2021139265A1 CN 2020118737 W CN2020118737 W CN 2020118737W WO 2021139265 A1 WO2021139265 A1 WO 2021139265A1
Authority
WO
WIPO (PCT)
Prior art keywords
preset
composition
feature
features
article
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2020/118737
Other languages
English (en)
French (fr)
Inventor
阮鸿涛
郑立颖
徐亮
阮晓雯
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Ping An Technology Shenzhen Co Ltd
Original Assignee
Ping An Technology Shenzhen Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Ping An Technology Shenzhen Co Ltd filed Critical Ping An Technology Shenzhen Co Ltd
Publication of WO2021139265A1 publication Critical patent/WO2021139265A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/205—Parsing
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/10—Text processing
    • G06F40/194—Calculation of difference between files

Definitions

  • This application relates to the field of artificial intelligence technology, and in particular to a composition scoring method, device, computer equipment, and computer-readable storage medium based on feature extraction and random forest.
  • AES Automated Essay Scoring
  • the more widely used systems for automatic scoring of essays include E-rater, Project Essay Grader, Intelligent EssayAssessor, IntelliMetric, Picai.com, and Tencent Cloud composition review.
  • composition automatic scoring system that has been used in the industry mainly uses linear regression modeling to predict the score. According to different test types, a dozen or more indicators that have the highest correlation with the composition score are selected to obtain the final score.
  • each essay automatic scoring system uses different indicators, resulting in the scoring results obtained by each essay automatic scoring system focusing on a certain aspect of the characteristics, so that the essay is scored accurately. The sex is low.
  • This application provides a composition scoring method, device, computer equipment and computer readable storage medium based on feature extraction and random forest, which can solve the problem of low accuracy of scoring due to incomplete evaluation of automatic scoring of essays in traditional technology The problem.
  • this application provides a composition scoring method based on feature extraction and random forest.
  • the method includes: obtaining a composition to be scored; extracting target features of the composition according to a preset feature extraction method, and the target feature Including the basic features of the preset language, the preset logic features, and the preset comprehension features.
  • the basic features of the preset language include preset vocabulary features, preset grammatical features, preset syntax features, preset length features, and preset readability.
  • the preset logical features include preset article structure features, preset article continuity features, and preset article element features, the preset comprehension features include similarity features with the preset article; and the target The feature is input into a pre-trained preset random forest model for scoring, so as to obtain a score corresponding to the composition.
  • this application also provides a composition scoring device based on feature extraction and random forest, including: an acquisition unit for acquiring the composition to be scored; and an extraction unit for extracting the composition according to a preset feature extraction method
  • the target characteristics of the target language, the target characteristics include predetermined language basic characteristics, predetermined logical characteristics, and predetermined comprehension characteristics, and the predetermined basic language characteristics include predetermined vocabulary characteristics, predetermined grammatical characteristics, predetermined syntactic characteristics, and predetermined Set length features and preset readability features.
  • the preset logic features include preset article structure features, preset article continuity features, and preset article element features. Similarity feature; a scoring unit for inputting the target feature into a pre-trained preset random forest model for scoring, so as to obtain a score corresponding to the composition.
  • the present application also provides a computer device, which includes a memory and a processor, and a computer program is stored on the memory.
  • the processor executes the computer program, the following steps are executed: obtaining a composition to be scored;
  • the target features of the composition are extracted according to a preset feature extraction method.
  • the target features include a preset language basic feature, a preset logic feature, and a preset comprehension feature.
  • the preset language basic feature includes a preset vocabulary feature, a preset comprehension feature. Set grammatical features, preset syntactic features, preset length features, and preset readability features.
  • the preset logical features include preset article structure features, preset article continuity features, and preset article element features. It is assumed that the comprehension feature includes similarity features with a preset article; the target feature is input into a pre-trained preset random forest model for scoring, so as to obtain a score corresponding to the composition.
  • the present application also provides a computer-readable storage medium, the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements the following steps: obtain The composition; extract the target features of the composition according to a preset feature extraction method, the target features include a preset language basic feature, a preset logic feature, and a preset comprehension feature, the preset language basic feature includes a preset vocabulary Features, preset grammatical features, preset syntax features, preset length features, and preset readability features, where the preset logical features include preset article structure features, preset article continuity features, and preset article element features, The preset comprehension feature includes a similarity feature with a preset article; the target feature is input into a pre-trained preset random forest model for scoring, so as to obtain a score corresponding to the composition. Perform the steps of the composition scoring method based on feature extraction and random forest.
  • This application extracts the target features of the composition according to a preset feature extraction method by acquiring the composition to be scored.
  • the target feature includes a preset language basic feature, a preset logic feature, and a preset comprehension feature. Enter the pre-trained preset random forest model for scoring to obtain the score corresponding to the composition. Due to the implementation of the feature project from the three dimensions of language basic ability, logical ability, and comprehension, it covers vocabulary, grammar, syntax, Composition length, readability, article organization and structure, comparison with high-quality articles, and other comprehensive and omni-directional features, so as to extract comprehensive feature indicators with evaluation significance, which can give explanations and explanations for the scoring results and different feature dimensions. Evaluation can improve the accuracy of the results obtained by automatic scoring of the composition.
  • FIG. 1 is a schematic flowchart of a composition scoring method based on feature extraction and random forest provided by an embodiment of the application;
  • FIG. 2 is a schematic diagram of a sub-process in the composition scoring method based on feature extraction and random forest provided by an embodiment of the application;
  • FIG. 3 is a schematic diagram of another sub-process of the composition scoring method based on feature extraction and random forest provided by an embodiment of the application;
  • FIG. 4 is a schematic block diagram of a composition scoring device based on feature extraction and random forest provided by an embodiment of the application.
  • Fig. 5 is a schematic block diagram of a computer device provided by an embodiment of the application.
  • FIG. 1 is a schematic flowchart of a composition scoring method based on feature extraction and random forest provided by an embodiment of the application. As shown in Figure 1, the method includes the following steps S101-S103:
  • target features of the composition include preset language basic features, preset logic features, and preset comprehension features
  • the preset language basic features include preset vocabulary features , Preset grammatical features, preset syntax features, preset length features, and preset readability features
  • the preset logical features include preset article structure features, preset article continuity features, and preset article element features
  • the pre-determined comprehension features include similarity features with pre-determined articles.
  • the automatic scoring method for composition based on feature extraction and random forest starts from the basic language features (used to describe the basic language ability of the composition) , Logical feature (used to describe the logical ability of composition), comprehension feature (used to describe the comprehension of the composition) three dimensions to implement feature engineering, extract the three dimensions that have evaluation significance for the composition, and use For automatic scoring of the composition.
  • composition corpus data Before automatically scoring the composition, first preprocess the composition corpus data to classify the composition corpus and divide it into different grouping categories.
  • the composition corpus can be divided into different corpus data sets according to the composition of authors of different ages, for example, Divide into different grouping categories such as high school group, junior high school group and elementary school group.
  • grouping categories such as high school group, junior high school group and elementary school group.
  • each grouping group eliminate the extreme situations of the composition word number is too long or the composition word number is too short, and then divide the composition into K according to the theme Composition theme, training essay automatic scoring model based on random forest model for essays of different age groups and different themes.
  • the feature engineering involved in feature extraction of the composition includes the following:
  • Basic language features used to describe the basic language ability of composition, including preset vocabulary features, preset grammatical features, preset syntax features, preset length features, and preset readability features. Specifically divided into the following characteristics:
  • the proportion of Unigram words is used to describe the richness of the vocabulary.
  • the number of Unigram words accounts for the proportion of the total number of words in the full text. Among them, the repeated use of words is only counted as one.
  • Unigram is a concept in natural language processing (NLP).
  • NLP natural language processing
  • the Unigram format corresponding to Xi'an Jiaotong University is: Xi/An/Jiao/Tong/Da/Xue.
  • Average number of strokes used to describe the average number of strokes of a single word, which can be obtained by counting strokes.
  • 3 The standard deviation of the number of strokes is used to describe the variance of the number of strokes of a single word.
  • the length of the composition is described by preset length characteristics, including the following: 1Sentence length distribution, used to describe the distribution of the number of words in a sentence. 2Average sentence length, used to describe the average number of words in a sentence. 3Total number of clauses, used to describe the number of full-text clauses used. 4The average number of sentences in a paragraph, used to describe the average number of sentences in each paragraph. 5 Total number of sentences, used to describe the number of full text sentences. 6 Number of paragraphs, used to describe the number of paragraphs in the full text. 7The number of words in the full text, used to describe the total number of words in the full text.
  • Readability index used to describe the evaluation index about the readability of Chinese text
  • the calculation formula is: N ⁇ average number of strokes in Chinese + (1-N) ⁇ The frequency of Chinese difficult words, where N adjusts the weight ratio between the Chinese average number of strokes and the frequency of Chinese difficult words.
  • the preset comprehension features include similarity features with preset essays, including the following: 1) Comparison of composition and high-quality essays. 1 Score of the most similar articles: Divide the corpus articles into n categories according to their scores, and take the score of the article category most similar to the article to be evaluated as an indicator. 2 Similarity with the highest quality paper: the similarity between the article to be evaluated and the article category with the highest score. 2) The degree of essay off-topic. 1 Comparing similarity with articles on the same topic: the similarity between the article to be evaluated and the article category of the same topic.
  • the target feature of the composition is extracted according to a preset feature extraction method.
  • a corresponding extraction method can be set for each feature in advance, and then when performing feature extraction, The above-mentioned features are extracted according to the feature extraction method corresponding to each of the above-mentioned features set in advance, so as to obtain the target feature of the composition, and the target feature includes a preset language basic feature, a preset logic feature, and a preset comprehension feature
  • the basic features of the preset language include preset vocabulary features, preset grammatical features, preset syntax features, preset length features, and preset readability features
  • the preset logical features include preset article structure features, presets The article coherence feature and the preset article element feature, and the preset comprehension feature includes the similarity feature with the preset article.
  • the automatic scoring method for essays based on feature extraction and random forest is based on the dimensions and features used in automatic scoring, and implements feature engineering from three dimensions: basic language ability, logical ability, and comprehension, covering vocabulary , Grammar/syntax, length of composition, readability, organization and structure of the article, comparison with high-quality articles, degree of off-topic, etc., extracting indicators with evaluation significance, can give explanations for the scoring results and different feature dimensions And evaluation, the composition automatic scoring model has high interpretability.
  • Random forest or Random forest in English, is a classifier that uses multiple trees to train and predict samples.
  • the pre-trained random forest model for automatic composition scoring is first pre-trained by pre-training composition samples.
  • the pre-training composition samples include pre-training compositions and pre-training essays obtained by scoring the pre-training compositions.
  • the training composition score where the pre-training composition sample is scored to obtain the pre-training composition score, the pre-training composition sample can be scored manually, wherein the pre-training composition score can also be performed on the pre-training composition
  • the score of the pre-training composition can be multiple different humans, respectively, according to the preset random forest model preset extraction target feature as the composition scoring index to score the pre-training composition Then, the average composition score calculated based on the scores of multiple pre-training essays, so that the manual scoring of the pre-training composition is as objective and comprehensive as possible.
  • the training The preset random forest model is also more objective, and can also improve the accuracy of the preset random forest model for automatic scoring of essays.
  • the process of pre-training the preset random forest model includes the following steps: obtaining a pre-training composition sample, the pre-training composition sample including a pre-training composition and a composition score corresponding to the pre-training composition; The sample is input to a preset random forest model; the preset target features of the pre-training composition and the composition score are extracted for learning, and the pre-training composition samples are classified to realize the pre-training of the preset random forest model.
  • the random forest model is an ensemble learning algorithm based on a decision tree as a learner, it is a classifier that uses multiple trees to train and predict samples. Therefore, when the preset random forest model is pre-trained, the preset random forest model will extract the preset target features and composition scores of the pre-training composition for learning, and learn according to the extracted pre-training composition The preset target features and composition scores classify the pre-trained composition samples to obtain different composition categories, thereby realizing the pre-trained preset random forest model.
  • the preset random forest model automatically scores the composition to be scored, it also first extracts the preset target features of the composition to be scored, and automatically converts the composition to be scored according to the preset target features of the extracted composition to be scored according to the pre-training time.
  • the learning result of the learning is to classify the composition to be scored.
  • the preset target features of the composition to be scored will be extracted and the pre-training composition is classified during pre-training.
  • the preset target features learned in each composition category are compared for similarity.
  • composition to be scored is classified into the composition category with the highest similarity, and the most similar composition that is most similar to the composition to be scored is obtained through the obtained similarity, and the score of the most similar composition is used as the basis for scoring the composition to be scored, based on the composition to be scored.
  • the similarity with the most similar composition is obtained, and the composition score of the composition to be scored is obtained, so as to realize the automatic scoring of the composition to be scored.
  • different target groups can also be classified to train different preset random forest models.
  • the composition corpus data is preprocessed to classify the composition corpus and divided into different grouping categories
  • the composition corpus can be divided into different corpus data sets according to the composition of authors of different age groups, such as high school group and junior high school
  • the composition corpus and the score corresponding to each composition are used to train the preset random forest model, so that the preset random forest model can automatically score the composition.
  • the statistical index QWK Quadrattic Weighted Kappa
  • Correlation a measure of model prediction
  • the results are consistent with the actual results
  • Accuracy the accuracy of model prediction results
  • RMSE RootMean Squared Error
  • the pre-training of the preset random forest model is completed. After the preset random forest model meets the predetermined goal, the training of the preset random forest model is completed, the composition to be scored is obtained, and after the target feature extraction of the composition is completed, the The target features are input into the random forest model, and the total score of the composition is evaluated through the above three dimensions.
  • the embodiment of the application extracts the target features of the composition according to a preset feature extraction method by acquiring the composition to be scored.
  • the target feature includes a preset language basic feature, a preset logic feature, and a preset comprehension feature.
  • the target feature is input into a pre-trained preset random forest model for scoring to obtain the score corresponding to the composition.
  • the embodiment of this application implements feature engineering from three dimensions: basic language ability, logical ability, and comprehension, it covers Vocabulary, grammar, syntax, composition length, readability, article organization and structure, and comparison with high-quality articles, and other comprehensive and comprehensive features, so as to extract comprehensive evaluation significance feature indicators, which can be based on the scoring results and differences
  • the feature dimension gives explanation and evaluation, which can improve the accuracy of the results obtained by automatic scoring of the composition.
  • the step of extracting the target features of the composition according to a preset feature extraction method includes: extracting the preset language basic features of the composition according to a preset language basic feature extraction method; and according to a preset logic
  • the feature extraction method extracts the preset logical feature of the composition; the preset comprehension feature extraction method is used to extract the preset comprehension feature of the composition.
  • the target feature includes a preset language basic feature, a preset logical feature, and a prediction.
  • the corresponding preset feature extraction method is used for feature extraction for each aspect of the composition, that is, the preset language basic feature of the composition is extracted according to the preset language basic feature extraction method, according to The preset logical feature extraction method extracts the preset logical feature of the composition, and the preset comprehension feature of the composition is extracted according to the preset comprehension feature extraction method, and the corresponding preset language foundation of the composition is obtained from Features, preset logical features, and preset comprehension features, and input the target features into a pre-trained preset random forest model for scoring, so as to obtain scores corresponding to the composition, so as to perform the composition Automatic scoring, because comprehensive evaluation-meaningful feature indicators are extracted for the composition, the scoring results and different feature dimensions of the composition can be explained and evaluated through the random forest model
  • the step of extracting the basic features of the composition according to the default language basic feature extraction method includes: using a first preset language tool to segment the composition to obtain the The phrase contained in the composition; based on the phrase, feature extraction is performed according to each preset basic sub-feature extraction method of the target language to obtain the basic sub-feature of each target language; the obtained basic sub-features of the target language are combined into a set Obtain the basic features of the preset language.
  • the first preset language tool may be a language tool that supports word segmentation, such as StanfordCoreNLP or HanLP.
  • language tools such as Stanford CoreNLP or HanLP
  • NLP tasks including tokenization, shallow analysis (sentence-character segmentation), word segmentation, sentence segmentation, segmentation, part-of-speech tagging, named entity recognition, and grammar parsing
  • tokenization shallow analysis
  • sentence segmentation word segmentation
  • sentence segmentation segmentation
  • part-of-speech tagging named entity recognition
  • grammar parsing you can The input sentence is segmented by a preset language tool to obtain phrase division, and then the phrase is labeled with part of speech, that is, the single sentence is segmented using the first preset language tool to obtain several phrases contained in the single sentence.
  • the obtained basic sub-features of the target language are grouped into a set to obtain the basic feature of the preset language. For example, if the preset basic sub-feature of the target language is the proportion of Unigram words (words), the basic sub-feature of the preset target language is obtained by counting the proportion of the number of Unigram words in the total number of full-text words.
  • the preset basic sub-feature of the target language is the proportion of Unigram words (words)
  • the basic sub-feature of the preset target language is obtained by counting the proportion of the number of Unigram words in the total number of full-text words.
  • the basic sub-feature of the preset target language is The average number of strokes is calculated by counting the average number of strokes of a single word to obtain the basic sub-features of the target language.
  • the corresponding prediction is performed according to the extraction method of the basic sub-features of each preset target language.
  • the basic sub-features of the target language are extracted to obtain the basic sub-features of the preset target language, and all the obtained basic sub-features of the target language are combined into a set to obtain the basic sub-features of the preset language.
  • the step of extracting the preset logical features of the composition according to a preset logical feature extraction method includes: cutting the composition into a plurality of sentences according to punctuation included in the composition; Based on the sentence, count the number of elements contained in the composition text corresponding to the preset article structure feature of the composition, and the number of words in all the elements corresponding to the preset article element feature Mean, the number of overlaps between adjacent sentences corresponding to the coherence feature of the preset article, where the element is the background type, argument type, main point of view, supporting point of view, or conclusion contained in the composition Among the categories, the set of sentences of the largest consecutively appearing category is called the elements of the category.
  • punctuation marks are generally used as the division of sentences. Therefore, through the recognition of punctuation marks, according to the recognized punctuation marks, it can be divided into several independent sentences, based on the above-mentioned pre-defined article structure characteristics.
  • each sentence into one of background, argument, main point of view, supporting point of view, and conclusion, and define the largest continuous set of sentences in a certain category as an element of that category, for example, 3 sentences of main point of view + 4 Sentence supporting point of view + 2 main points of sentence, corresponding to 3 elements, count the number of elements as an indicator of the composition structure, and count the overlap (ie overlap) of adjacent sentences in the composition to obtain the corresponding coherence characteristics of the preset article
  • the number of overlaps between adjacent sentences is used as a composition structure indicator, and the average value of the number of words in all the elements corresponding to the feature of the preset article element is calculated as the composition structure indicator, etc.
  • FIG. 2 is a schematic diagram of a sub-process in the composition scoring method based on feature extraction and random forest provided by an embodiment of the application.
  • the step of extracting the preset comprehension feature of the composition according to a preset comprehension feature extraction method includes:
  • the preset feature index includes all or part of the sub-features of the preset language basic feature, the preset logic feature, and the preset comprehension feature.
  • the preset article contained in the preset corpus is acquired, and the preset feature index includes all or part of the sub-features of the preset language basic feature, the preset logic feature, and the preset comprehension feature.
  • the preset article is a number of articles
  • the composition is compared with the preset article according to the preset feature index to obtain the composition
  • the method further includes: S203. Obtain the most similar article with the smallest difference in the similarity score from the composition from the preset article; S204. Obtain the most similar article with the composition.
  • the score index of similar articles belonging to the same article category is the comparison index; S205, based on the comparison index, compare the composition with the preset target article with the highest score to obtain the difference between the composition and the preset target article ⁇ similarity.
  • the preset articles contained in the preset corpus are several articles, and each article has a corresponding score.
  • Several articles in the corpus can be divided into n categories according to the scores, and obtained from the preset articles The most similar article with the smallest difference in the similarity score of the composition is obtained, and then the score index of the article category most similar to the article to be evaluated is obtained as the comparison index.
  • the composition is compared with the preset target with the highest score
  • the articles are compared to obtain the similarity between the composition and the preset target article, so as to realize the comparison between the composition to be graded and the best quality paper to obtain the similarity between the composition and the best quality paper, thereby The similarity between the composition to be evaluated and the article category with the highest score can be obtained, thereby achieving a more accurate scoring of the composition.
  • FIG. 3 is a schematic diagram of another sub-process of the composition scoring method based on feature extraction and random forest provided by an embodiment of the application.
  • the step of extracting the preset comprehension feature of the composition according to the preset comprehension feature extraction method further includes:
  • the subject corresponding to the composition is acquired, for example, the subject of the composition is acquired from the title of the composition, and articles of the same subject included in the subject in the preset corpus are acquired, wherein the article of the same subject
  • compare the similarity between the composition and the article on the same topic to obtain the category similarity of the category corresponding to the composition and the article on the same topic, so as to realize the
  • the composition is compared with the article on the same topic, and the category similarity of the comparison between the composition and the article on the same topic can be used to determine the degree of off-topic of the composition, so that the composition can be automatically scored from a more comprehensive dimension, which is quite high
  • the accuracy of the automatic scoring of the composition is quite high.
  • composition scoring method based on feature extraction and random forest described in the above embodiments can recombine the technical features included in different embodiments as needed to obtain a combined implementation plan, but they are all in Within the scope of protection required by this application.
  • FIG. 4 is a schematic block diagram of a composition scoring device based on feature extraction and random forest provided by an embodiment of the application.
  • an embodiment of the present application also provides a composition scoring device based on feature extraction and random forest.
  • the composition scoring device based on feature extraction and random forest includes a unit for executing the above-mentioned composition scoring method based on feature extraction and random forest.
  • the composition scoring device based on feature extraction and random forest can be Configured in computer equipment.
  • the composition scoring device 400 based on feature extraction and random forest includes an acquisition unit 401, an extraction unit 402 and a scoring unit 403.
  • the acquisition unit 401 is used to obtain the composition to be scored;
  • the extraction unit 402 is used to extract the target features of the composition according to a preset feature extraction method, and the target features include preset language basic features, preset logic features, and Pre-defined comprehension features,
  • the pre-defined basic language features include pre-defined vocabulary features, pre-defined grammatical features, pre-defined syntactic features, pre-defined length features, and pre-defined readability features
  • the predetermined logical features include pre-defined The article structure feature, the preset article continuity feature, and the preset article element feature, the preset comprehension feature includes the similarity feature with the preset article;
  • the scoring unit 403 is used to input the target feature into the pre-trained The random forest model is preset for scoring to obtain the score corresponding to the composition.
  • the extraction unit 402 includes: a language basic feature extraction subunit, configured to extract the preset language basic features of the composition according to a preset language basic feature extraction method; a logical feature extraction subunit, using To extract the preset logical features of the composition according to a preset logical feature extraction method; the comprehension feature extraction subunit is configured to extract the preset comprehension features of the composition according to a preset comprehension feature extraction method.
  • the language basic feature extraction subunit includes: a word segmentation subunit, configured to use a first preset language tool to segment the composition to obtain the phrases contained in the composition; a subfeature extraction unit , Used to perform feature extraction according to each preset target language basic sub-feature extraction method based on the phrase, to obtain each target language basic sub-feature; the set sub-unit is used to combine the obtained several target language basic sub-features Form a set to obtain the basic features of the preset language;
  • the logical feature extraction subunit includes: a cutting subunit for cutting the composition into a number of sentences according to the punctuation included in the composition; and a statistics subunit for cutting the composition based on the punctuation marks.
  • Sentence the number of elements contained in the composition text corresponding to the preset article structure feature of the composition, the mean value of the number of words in all the elements corresponding to the preset article element feature, and the State the number of overlaps between adjacent sentences corresponding to the coherence features of the presupposed article, where the elements are the background type, argument type, main point of view, supporting point of view, or conclusion type included in the composition,
  • the set of sentences of the largest consecutive category is called the elements of that category.
  • the comprehension feature extraction subunit includes: an acquisition subunit for acquiring a preset article contained in a preset corpus; a comparison subunit for comparing the composition with the preset article The comparison is performed according to the preset feature index to obtain the preset similarity between the composition and the preset article.
  • the preset articles are several articles
  • the comprehension feature extraction subunit further includes: a most similar article obtaining subunit, configured to obtain similarities to the composition from the preset articles The most similar article with the smallest degree score difference; a comparison index acquisition unit for acquiring a score index that belongs to the same article category as the most similar article as a comparison index; a comparison subunit for comparing the composition based on the comparison index A comparison is made with the preset target article with the highest score to obtain the similarity between the composition and the preset target article.
  • the comprehension feature extraction subunit further includes: a topic acquisition subunit for acquiring the topic corresponding to the composition; and the same topic article acquisition subunit for acquiring the information in the preset corpus Articles on the same subject included in the subject, wherein the article on the same subject is an article with the same subject as the composition; the category similarity acquisition subunit is used to compare the similarity of the composition with the article on the same subject , In order to obtain the category similarity of the category corresponding to the composition and the article of the same topic.
  • the division and connection of the units in the composition scoring device based on feature extraction and random forest are only used for illustration.
  • the composition scoring device based on feature extraction and random forest can be divided into different types as needed.
  • the units of the composition scoring device based on feature extraction and random forest can also be connected in different order and manners to complete all or part of the functions of the composition scoring device based on feature extraction and random forest.
  • composition scoring device based on feature extraction and random forest can be implemented in the form of a computer program, and the computer program can be run on a computer device as shown in FIG. 5.
  • FIG. 5 is a schematic block diagram of a computer device according to an embodiment of the present application.
  • the computer device 500 may be a computer device such as a desktop computer or a server, or may be a component or component in other devices.
  • the computer device 500 includes a processor 502, a memory, and a network interface 505 connected through a system bus 501, where the memory may include a non-volatile storage medium 503 and an internal memory 504.
  • the non-volatile storage medium 503 can store an operating system 5031 and a computer program 5032.
  • the processor 502 can execute the above-mentioned composition scoring method based on feature extraction and random forest.
  • the processor 502 is used to provide calculation and control capabilities to support the operation of the entire computer device 500.
  • the internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503.
  • the processor 502 can execute the above-mentioned composition based on feature extraction and random forest. Scoring method.
  • the network interface 505 is used for network communication with other devices.
  • the specific computer device 500 may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.
  • the computer device may only include a memory and a processor.
  • the structures and functions of the memory and the processor are the same as those of the embodiment shown in FIG. 5, which will not be repeated here.
  • the processor 502 is configured to run a computer program 5032 stored in a memory to implement the composition scoring method described in the embodiment of the present application.
  • the processor 502 may be a central processing unit (Central Processing Unit, CPU), and the processor 502 may also be other general-purpose processors, digital signal processors (Digital Signal Processors, DSPs), and special purpose processors.
  • Integrated circuit Application Specific Integrated Circuit, ASIC
  • off-the-shelf programmable gate array Field-Programmable Gate Array, FPGA
  • the general-purpose processor may be a microprocessor or the processor may also be any conventional processor.
  • the embodiment of the present application also provides a computer-readable storage medium.
  • the computer-readable storage medium may be a non-volatile computer-readable storage medium, or may be a volatile computer-readable storage medium, the computer-readable storage medium stores a computer program, and the computer program is executed by the processor When the processor executes the steps of the composition scoring method based on feature extraction and random forest described in the above embodiments.
  • the storage medium is a physical, non-transitory storage medium, such as a U disk, a mobile hard disk, a read-only memory (Read-Only Memory, ROM), a magnetic disk, or an optical disk, etc., which can store computer programs. medium.
  • a physical, non-transitory storage medium such as a U disk, a mobile hard disk, a read-only memory (Read-Only Memory, ROM), a magnetic disk, or an optical disk, etc., which can store computer programs. medium.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Machine Translation (AREA)

Abstract

一种基于特征抽取及随机森林的作文评分方法、装置、计算机设备及计算机可读存储介质，属于人工智能技术领域，该方法包括：获取待评分的作文（S101），根据预设特征抽取方式抽取作文的目标特征，目标特征包括预设语言基础特征、预设逻辑特征及预设理解力特征（S102），将目标特征输入经过预训练的预设随机森林模型进行评分，以得到作文所对应的分数（S103）。

Description

作文评分方法、装置、计算机设备及计算机可读存储介质
本申请要求于2020年07月16日提交中国专利局、申请号为202010687313.4、申请名称为“作文评分方法、装置、计算机设备及计算机可读存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及人工智能技术领域，尤其涉及一种基于特征抽取及随机森林的作文评分方法、装置、计算机设备及计算机可读存储介质。
背景技术
作文自动评分(Automated Essay Scoring，AES)近年来被广泛应用于教师批改作文、学生自助评测、家长辅助检查和考试自动评分等场景。运用作文自动评分较为广泛的系统有E-rater、Project Essay Grader、Intelligent EssayAssessor、IntelliMetric、批改网及腾讯云作文批改等。
目前工业上已经应用的作文自动评分系统主要利用线性回归方法建模进行得分预测，根据不同的考试类型选择十几个甚至更多和作文分数相关性最高的指标来获取最终评分。但是，发明人发现，传统技术中，每个作文自动评分系统采用的指标不一样，导致每个作文自动评分系统得出的评分结果较侧重于某一方面的特征，从而对作文进行评分的准确性较低。
发明内容
本申请提供了一种基于特征抽取及随机森林的作文评分方法、装置、计算机设备及计算机可读存储介质，能够解决传统技术中对对作文进行自动评分由于评价不全面而导致评分准确性较低的问题。
第一方面，本申请提供了一种基于特征抽取及随机森林的作文评分方法，所述方法包括：获取待评分的作文；根据预设特征抽取方式抽取所述作文的目标特征，所述目标特征包括预设语言基础特征、预设逻辑特征及预设理解力特征，所述预设语言基础特征包括预设词汇特征、预设语法特征、预设句法特征、预设篇幅特征及预设可读性特征，所述预设逻辑特征包括预设文章结构特征、预设文章连贯性特征及预设文章元素特征，所述预设理解力特征包括与预设文章的相似度特征；将所述目标特征输入经过预训练的预设随机森林模型进行评分，以得到所述作文所对应的分数。
第二方面，本申请还提供了一种基于特征抽取及随机森林的作文评分装置，包括：获取单元，用于获取待评分的作文；抽取单元，用于根据预设特征抽取方式抽取所述作文的目标特征，所述目标特征包括预设语言基础特征、预设逻辑特征及预设理解力特征，所述预设语言基础特征包括预设词汇特征、预设语法特征、预设句法特征、预设篇幅特征及预设可读性特征，所述预设逻辑特征包括预设文章结构特征、预设文章连贯性特征及预设文章元素特征，所述预设理解力特征包括与预设文章的相似度特征；评分单元，用于将所述目标特征输入经过预训练的预设随机森林模型进行评分，以得到所述作文所对应的分数。
第三方面，本申请还提供了一种计算机设备，其包括存储器及处理器，所述存储器上存储有计算机程序，所述处理器执行所述计算机程序时执行如下步骤：获取待评分的作文；根据预设特征抽取方式抽取所述作文的目标特征，所述目标特征包括预设语言基础特征、预设逻辑特征及预设理解力特征，所述预设语言基础特征包括预设词汇特征、预设语法特征、预设句法特征、预设篇幅特征及预设可读性特征，所述预设逻辑特征包括预设文章结构特征、预设文章连贯性特征及预设文章元素特征，所述预设理解力特征包括与预设文章的相似度特征；将所述目标特征输入经过预训练的预设随机森林模型进行评分，以得到所述作文所对应的分数。
第四方面，本申请还提供了一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，所述计算机程序被处理器执行时使所述处理器实现如下步骤：获取待评分的作文；根据预设特征抽取方式抽取所述作文的目标特征，所述目标特征包括预设语言基础特征、 预设逻辑特征及预设理解力特征，所述预设语言基础特征包括预设词汇特征、预设语法特征、预设句法特征、预设篇幅特征及预设可读性特征，所述预设逻辑特征包括预设文章结构特征、预设文章连贯性特征及预设文章元素特征，所述预设理解力特征包括与预设文章的相似度特征；将所述目标特征输入经过预训练的预设随机森林模型进行评分，以得到所述作文所对应的分数。执行所述基于特征抽取及随机森林的作文评分方法的步骤。
本申请通过获取待评分的作文，根据预设特征抽取方式抽取所述作文的目标特征，所述目标特征包括预设语言基础特征、预设逻辑特征及预设理解力特征，将所述目标特征输入经过预训练的预设随机森林模型进行评分，以得到所述作文所对应的分数，由于从语言基础能力、逻辑能力、理解力三大维度进行实施特征工程，涵盖了词汇、语法、句法、作文篇幅、可读性、文章的组织与结构、与优质文章的对比等多方面、全方位的特征，从而抽取全面的具有评价意义的特征指标，可以针对打分结果及不同特征维度给出解释和评价，能够提高作文自动评分所得到结果的准确性。
附图说明
为了更清楚地说明本申请实施例技术方案，下面将对实施例描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1为本申请实施例提供的基于特征抽取及随机森林的作文评分方法的一个流程示意图；
图2为本申请实施例提供的基于特征抽取及随机森林的作文评分方法中一个子流程的示意图；
图3为本申请实施例提供的基于特征抽取及随机森林的作文评分方法的另一个子流程示意图；
图4为本申请实施例提供的基于特征抽取及随机森林的作文评分装置的一个示意性框图；以及
图5为本申请实施例提供的计算机设备的示意性框图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
请参阅图1，图1为本申请实施例提供的基于特征抽取及随机森林的作文评分方法的一个流程示意图。如图1所示，该方法包括以下步骤S101-S103：
S101、获取待评分的作文。
S102、根据预设特征抽取方式抽取所述作文的目标特征，所述目标特征包括预设语言基础特征、预设逻辑特征及预设理解力特征，所述预设语言基础特征包括预设词汇特征、预设语法特征、预设句法特征、预设篇幅特征及预设可读性特征，所述预设逻辑特征包括预设文章结构特征、预设文章连贯性特征及预设文章元素特征，所述预设理解力特征包括与预设文章的相似度特征。
具体地，由于作文的语言基础能力包括作文中采取的词汇、语法、句法、作文篇幅及作文可读性等特征指标来体现，作文的逻辑能力涉及作文的组织与结构等特征指标，作文的理解力包括作文与优质文章的对比及跑题程度等特征指标，因此，本申请实施例所提供的基于特征抽取及随机森林的作文自动评分方法，从语言基础特征(用于描述作文的语言基础能力)、逻辑特征(用于描述作文的逻辑能力)、理解力特征(用于描述作文的理解力)三大维度实施特征工程，抽取对于作文具有评价意义的上述三大维度所包含的特征指标，用于对作文进行自动评分。
在对作文进行自动评分前，首先对作文语料数据进行预处理以对作文语料进行分类，划分为不同的分组类别，可以将作文语料按照不同年龄段作者的作文划分为不同的语料数据集， 例如划分为高中组、初中组及小学组等不同的分组类别，并在每一分组组中剔除作文单词数目过长或者作文单词数目过短等极端情形的作文，然后将作文按照主题划分为K个作文主题，针对不同年龄段及不同主题的作文分别训练基于随机森林模型的作文自动评分模型。其中，对作文进行特征抽取涉及的特征工程，具体包括如下：
(1)语言基础特征，用于描述作文的语言基础能力，包括预设词汇特征、预设语法特征、预设句法特征、预设篇幅特征及预设可读性特征。具体分为以下特征内容：
1)词汇，用词汇特征进行描述，包含以下各个子特征：
①Unigram单词(字)占比，用于描述词汇丰富度，Unigram单词个数占全文单词总数目的比例，其中，重复使用单词只计算为一个，Unigram，是自然语言处理(NLP)中的概念，用于描述一元模型，比如，西安交通大学所对应的Unigram形式为：西/安/交/通/大/学。②平均笔画数，用于描述单个单词笔画数的平均数目，可以采用统计笔画的方式获得。③笔画数的标准方差，用于描述单个单词笔画数的方差。④单词类别数(含标点或不含标点)，用于描述Unigram单词个数。⑤笔画数大于n的单词百分比(n＝6，7，8，···)。⑥Mechanics错误率，用于描述错别字、标点使用等错误次数占总单词数目的百分比。⑦词汇水平指标，通过将每个单词标记词汇难易水平，并选定作文中Index(即特征)排名倒数第五的单词的水平值作为特征。⑧成语使用数，用于描述作文中使用的成语个数。
2)语法及句法，通过预设语法特征及预设句法特征进行描述，包括以下特征指标：①词性标注占比，例如名词、动词、形容词、代词、借词、连词、虚词分别占总单词数目的百分比，可以通过预设语言工具进行分词处理，然后进行词性标注，再统计词性标注占比。②语法使用错误率(Grammar)，用于描述语法(词性及、词组等)使用错误次数占总单词数目的百分比。③过度重复使用词，用于描述过度重复使用的词或词组的使用次数占总单词数目的百分比。④句法结构解析树，通过树的深度等树结构所对应的指标以描述句法结构。⑤从句关系，包括主语从句、宾语从句、定语从句、状语从句等数目。
3)作文篇幅，通过预设篇幅特征来描述，包括以下内容：①句长分布，用于描述句子单词数目的分布。②平均句长，用于描述句子单词数目的平均数。③从句总数，用于描述全文从句使用数量。④平均段落句子数，用于描述平均每个段落的句子数目。⑤句子总数，用于描述全文句子数目。⑥段落数量，用于描述全文段落数目。⑦全文单词数，用于描述全文总单词数目。
4)可读性，通过预设可读性特征来描述，包括以下内容：①可读性指标，用于描述关于中文文本可读性的评价指标，计算公式为：N×中文平均笔画数+(1-N)×中文难词频度，其中N调整中文平均笔画数指标以及中文难词频度指标之间的权重配比。
(2)逻辑能力，通过预设逻辑特征进行描述，所述预设逻辑特征包括预设文章结构特征、预设文章连贯性特征及预设文章元素特征，包括以下内容：
1)文章的组织与结构，采用预设文章结构特征来描述，包括以下内容：①文章结构指标，：将每个句子分类为背景、论点、主要观点、支撑观点、结论中的一类，将最大连续出现的某个类别的句子集合定义为该类别的一个元素，例如3句主要观点+4句支撑观点+2句主要观点，则对应3个元素，计算元素个数作为文章结构指标。②文章连贯性，用于描述相邻句子的Overlap(即重叠)情况。③文章元素平均单词数，用于描述所有元素中单词数目的均值。
(3)作文理解力，通过所述预设理解力特征，所述预设理解力特征包括与预设文章的相似度特征，包括以下内容：1)作文与优质文章的对比。①最相似文章得分：将语料库文章按得分分为n类，取与待评价文章最相似的文章类别的得分为指标。②与最优质论文相似度：待评价文章与得分最高的文章类别的相似度。2)作文跑题程度。①与同主题文章对比相似度：待评价文章与同主题文章类别的相似度。
获取待评分的作文后，根据预设特征抽取方式抽取所述作文的目标特征，针对上述每一项所述特征，可以预先针对每一项特征设置对应的抽取方式，进而在进行特征抽取时，根据预先设置的上述特征各自所对应的特征抽取方式抽取上述各项特征，从而得到所述作文的目 标特征，所述目标特征包括预设语言基础特征、预设逻辑特征及预设理解力特征，所述预设语言基础特征包括预设词汇特征、预设语法特征、预设句法特征、预设篇幅特征及预设可读性特征，所述预设逻辑特征包括预设文章结构特征、预设文章连贯性特征及预设文章元素特征，所述预设理解力特征包括与预设文章的相似度特征。本申请实施例提供的基于特征抽取及随机森林的作文自动评分方法，针对自动评分用到的维度及特征，筛选从语言基础能力、逻辑能力、理解力三大维度进行实施特征工程，涵盖了词汇、语法/句法、作文篇幅、可读性、文章的组织与结构、与优质文章的对比、跑题程度等多方面的特征，抽取具有评价意义的指标，可以针对打分结果及不同特征维度给出解释和评价，作文自动评分模型可解释性高。
S103、将所述目标特征输入经过预训练的预设随机森林模型进行评分，以得到所述作文所对应的分数。
其中，随机森林，英文为Random forest，是利用多棵树对样本进行训练并预测的一种分类器。
具体地，首先通过预训练作文样本对用于进行作文自动评分的预设随机森林模型进行预训练，所述预训练作文样本包括预训练作文及对所述预训练作文进行评分所对应得到的预训练作文分值，其中，对预训练作文样本进行评分以得到预训练作文分值，可以通过人工对预训练作文样本进行评分，其中，所述预训练作文分值还可以是对预训练作文进行多次评分得到的预训练作文分值的作文分值均值。比如，若通过人工对预训练作文进行评分时，所述预训练作文分值可以为多位不同的人工各自根据预设随机森林模型预设提取的目标特征作为作文评分指标对预训练作文进行评分后，根据得到多个预训练作文分值计算得到的作文分值均值，以使人工对该预训练作文的评分尽可能客观和全面，由于预训作文样本数据的准确性得到提升，从而使训练出的预设随机森林模型也更具备客观性，也能提升预设随机森林模型对作文自动评分的评分准确性。对预设随机森林模型进行预训练的过程包括以下步骤：获取预训练作文样本，所述预训练作文样本包括预训练作文及所述预训练作文所对应的作文分值；将所述预训练作文样本输入至预设随机森林模型；提取所述预训练作文的预设目标特征及所述作文分值进行学习，并对所述预训练作文样本进行分类，以实现预训练所述预设随机森林模型。
具体地，由于随机森林模型是以决策树为基学习器的集成学习算法，是利用多棵树对样本进行训练并预测的一种分类器。因此，在对预设随机森林模型进行预训练时，所述预设随机森林模型会提取所述预训练作文的预设目标特征及作文分值进行学习，并根据提取的所述预训练作文的预设目标特征及作文分值对所述预训练作文样本进行分类，以得到不同的作文类别，从而实现预训练预设随机森林模型。后续，预设随机森林模型对待评分作文进行自动评分时，也先提取待评分作文的预设目标特征，并根据提取的待评分作文的预设目标特征将所述待评分作文根据预训练时自动进行学习的学习结果对待评分作文进行分类，即将提取的待评分作文的预设目标特征与预训练时将预训练作文进行分类时对于各个作文类别学习的预设目标特征进行相似性比对，将待评分作文分类至相似度最高的作文类别中，并通过得到的相似度获取与待评分最相似的最相似作文，将最相似作文的作文分值作为待评分作文评分的基础，根据待评分作文与最相似作文的相似度，得到待评分作文的作文分值，以实现对待评分作为进行自动评分。
进一步地，对预设随机森林模型进行预训练时，还可以针对不同的目标群体进行分类，以训练不同的预设随机森林模型。例如，在对作文语料数据进行预处理以对作文语料进行分类，划分为不同的分组类别，可以将作文语料按照不同年龄段作者的作文划分为不同的语料数据集，例如划分为高中组、初中组及小学组等不同的分组类别等，采用所述作文语料及每篇作文所对应的评分训练预设随机森林模型，以使预设随机森林模型能够对作文进行自动评分。
进一步地，在训练所述随机森林模型时，以广泛应用于衡量有序量表评估一致性的统计指标QWK(Quadratic Weighted Kappa)作为自动评分模型的评价指标，同时增加了Correlation (衡量了模型预测结果与实际结果的趋势一致性)、Accuracy(模型预测结果的准确度)、RMSE(RootMean Squared Error)三种常用指标作为辅助输出指标值，对所述随机森林模型进行调试，以提供所述随机森林模型进行作文自动评分的准确性。
对预设随机森林模型预训练完成，预设随机森林模型满足预定目标后，完成所述预设随机森林模型的训练，获取待评分的作文，并在对所述作文完成目标特征抽取后，将目标特征输入随机森林模型，通过以上三个维度对作文进行总分评估。
本申请实施例通过获取待评分的作文，根据预设特征抽取方式抽取所述作文的目标特征，所述目标特征包括预设语言基础特征、预设逻辑特征及预设理解力特征，将所述目标特征输入经过预训练的预设随机森林模型进行评分，以得到所述作文所对应的分数，由于本申请实施例从语言基础能力、逻辑能力、理解力三大维度进行实施特征工程，涵盖了词汇、语法、句法、作文篇幅、可读性、文章的组织与结构、与优质文章的对比等多方面、全方位的特征，从而抽取全面的具有评价意义的特征指标，可以针对打分结果及不同特征维度给出解释和评价，能够提高作文自动评分所得到结果的准确性。
在一实施例中，所述根据预设特征抽取方式抽取所述作文的目标特征的步骤包括：根据预设语言基础特征抽取方式抽取所述作文的所述预设语言基础特征；根据预设逻辑特征抽取方式抽取所述作文的所述预设逻辑特征；根据预设理解力特征抽取方式抽取所述作文的所述预设理解力特征。
具体地，由于在本申请实施例中，为了提高对作文进行评分的准确性，采用了三大维度的较为全面的目标特征，所述目标特征包括预设语言基础特征、预设逻辑特征及预设理解力特征，针对作文在每一方面的特征均采用所对应的预设特征抽取方式进行特征抽取，即根据预设语言基础特征抽取方式抽取所述作文的所述预设语言基础特征，根据预设逻辑特征抽取方式抽取所述作文的所述预设逻辑特征，根据预设理解力特征抽取方式抽取所述作文的所述预设理解力特征，从得到作文所对应的包含预设语言基础特征、预设逻辑特征及预设理解力特征的目标特征，并将将所述目标特征输入经过预训练的预设随机森林模型进行评分，以得到所述作文所对应的分数，从而对作文进行自动评分，由于针对作文抽取了全面的具有评价意义的特征指标，可以通过随机森林模型对作文的打分结果及不同特征维度给出解释和评价，能够提高作文自动评分所得到结果的准确性。
在一实施例中，所述根据预设语言基础特征抽取方式抽取所述作文的所述预设语言基础特征的步骤包括：采用第一预设语言工具对所述作文进行分词，以得到所述作文所包含的短语；基于所述短语，根据每个预设目标语言基础子特征抽取方式进行特征抽取，以得到每个目标语言基础子特征；将得到的若干个目标语言基础子特征组成集合以得到所述预设语言基础特征。
其中，第一预设语言工具可以为StanfordCoreNLP或者HanLP等支持进行分词的语言工具。
具体地，由于语言工具(例如Stanford CoreNLP或者HanLP)支持包括标记化、浅层分析(句字分块)、分词、分句、分块、词性标注、命名实体识别及语法解析等NLP任务，可以通过预设语言工具对输入语句进行分词得到短语划分，再对短语进行词性标注，即采用第一预设语言工具对所述单语句进行分词，以得到所述单语句所包含的若干个短语。再基于所述短语，根据所述语言基础特征所包含的预设目标语言基础子特征，根据上述对子特征所对应内容的描述，根据每个预设目标语言基础子特征抽取方式进行特征抽取，以得到每个目标语言基础子特征，将得到的若干个目标语言基础子特征组成集合以得到所述预设语言基础特征。例如，若预设目标语言基础子特征为Unigram单词(字)占比，通过统计Unigram单词个数占全文单词总数目的比例以得到预设目标语言基础子特征，若预设目标语言基础子特征为平均笔画数，通过统计单个单词笔画数的平均数目以得到目标语言基础子特征，对于其它的预设目标语言基础子特征，均根据每个预设目标语言基础子特征抽取方式进行所对应的预设目标语言基础子特征抽取，以得到预设目标语言基础子特征，将得到的所有目标语言基础 子特征组成集合以得到所述预设语言基础特征。
在一实施例中，所述根据预设逻辑特征抽取方式抽取所述作文的所述预设逻辑特征的步骤包括：根据所述作文中所包含的标点符号将所述作文切割为若干个句子；基于所述句子，统计出所述作文的所述预设文章结构特征所对应的所述作文本所包含的元素个数、所述预设文章元素特征所对应的所有所述元素中单词数目的均值、所述预设文章连贯性特征所对应的相邻句子之间的重叠数量，其中，所述元素为所述作文中所包含的背景类、论点类、主要观点类、支撑观点类或者结论类别中，最大连续出现的类别的句子集合称为该类别的元素。
具体地，作文中一般是以标点符号作为句子的划分，因此通过对标点符号的识别，根据识别出来的标点符号，可以将作为切割为若干个独立的句子，基于上述对预设文章结构特征的描述，结合切割获得的句子，统计出所述作文的所述预设文章结构特征所对应的所述作文本所包含的元素个数、所述预设文章元素特征所对应的所有所述元素中单词数目的均值、所述预设文章连贯性特征所对应的相邻句子之间的重叠数量，其中，所述元素为所述作文中所包含的背景类、论点类、主要观点类、支撑观点类或者结论类别中，最大连续出现的类别的句子集合称为该类别的元素，从而根据预设逻辑特征抽取方式抽取所述作文的所述预设逻辑特征。比如，将每个句子分类为背景、论点、主要观点、支撑观点、结论中的一类，将最大连续出现的某个类别的句子集合定义为该类别的一个元素，例如3句主要观点+4句支撑观点+2句主要观点，则对应3个元素，计算元素个数作为作文结构指标，统计作文中相邻句子的Overlap(即重叠)情况，以得到所述预设文章连贯性特征所对应的相邻句子之间的重叠数量作为作文结构指标，统计所述预设文章元素特征所对应的所有所述元素中单词数目的均值作为作文的结构指标等。
请参阅图2，图2为本申请实施例提供的基于特征抽取及随机森林的作文评分方法中一个子流程的示意图。在该实施例中，所述根据预设理解力特征抽取方式抽取所述作文的所述预设理解力特征的步骤包括：
S201、获取预设语料库中所包含的预设文章；S202、将所述作文与所述预设文章按照预设特征指标进行对比，以得到所述作文与所述预设文章的预设相似度。
其中，预设特征指标包括预设语言基础特征、预设逻辑特征及预设理解力特征中的全部或者部分子特征。
具体地，获取预设语料库中所包含的预设文章，预设特征指标包括预设语言基础特征、预设逻辑特征及预设理解力特征中的全部或者部分子特征。
请继续参阅图2，如图2所示，进一步地，所述预设文章为若干篇文章，所述将所述作文与所述预设文章按照预设特征指标进行对比，以得到所述作文与所述预设文章的预设相似度的步骤之后，还包括：S203、从所述预设文章中获得与所述作文的相似度分数差距最小的最相似文章；S204、获取与所述最相似文章同属一个文章类别的得分指标为对比指标；S205、基于所述对比指标，将所述作文与得分最高的预设目标文章进行对比，以得到所述作文与所述预设目标文章之间的相似度。
具体地，预设语料库中所包含的预设文章为若干篇文章，每篇文章对应有相应的得分，将语料库中的若干篇文章可以按照得分分为n类，从所述预设文章中获得与所述作文的相似度分数差距最小的最相似文章，进而获取与待评价文章最相似的文章类别的得分指标为对比指标，基于所述对比指标，将所述作文与得分最高的预设目标文章进行对比，以得到所述作文与所述预设目标文章之间的相似度，从而实现将待评分作文与最优质论文进行对比以得到所述作文与最优质论文之间的相似度，从而可以得到待评价作文与得分最高的文章类别的相似度，从而实现对所述作文进行较为准确的评分。
请参阅图3，图3为本申请实施例提供的基于特征抽取及随机森林的作文评分方法的另一个子流程示意图。在该实施例中，所述根据预设理解力特征抽取方式抽取所述作文的所述预设理解力特征的步骤还包括：
S301、获取所述作文所对应的主题；S302、获取所述预设语料库中所述主题所包含的同 主题文章，其中，所述同主题文章为与所述作文的主题相同的文章；S303、将所述作文与所述同主题文章进行相似度对比，以得到所述作文与所述同主题文章所对应类别的类别相似度。
具体地，获取所述作文所对应的主题，例如从所述作文的标题获得所述作文的主题，获取所述预设语料库中所述主题所包含的同主题文章，其中，所述同主题文章为与所述作文的主题相同的文章，将所述作文与所述同主题文章进行相似度对比，以得到所述作文与所述同主题文章所对应类别的类别相似度，从而实现将所述作文与同主题文章进行对比，通过所述作文与同主题文章的进行对比的类别相似度，可进而判断得出所述作文的跑题程度，从而从较为全面的维度对作文进行自动评分，挺高对作文自动评分的准确性。
需要说明的是，上述各个实施例所述的基于特征抽取及随机森林的作文评分方法，可以根据需要将不同实施例中包含的技术特征重新进行组合，以获取组合后的实施方案，但都在本申请要求的保护范围之内。
请参阅图4，图4为本申请实施例提供的基于特征抽取及随机森林的作文评分装置的一个示意性框图。对应于上述所述基于特征抽取及随机森林的作文评分方法，本申请实施例还提供一种基于特征抽取及随机森林的作文评分装置。如图4所示，该基于特征抽取及随机森林的作文评分装置包括用于执行上述所述基于特征抽取及随机森林的作文评分方法的单元，该基于特征抽取及随机森林的作文评分装置可以被配置于计算机设备中。具体地，请参阅图4，该基于特征抽取及随机森林的作文评分装置400包括获取单元401、抽取单元402及评分单元403。
其中，获取单元401，用于获取待评分的作文；抽取单元402，用于根据预设特征抽取方式抽取所述作文的目标特征，所述目标特征包括预设语言基础特征、预设逻辑特征及预设理解力特征，所述预设语言基础特征包括预设词汇特征、预设语法特征、预设句法特征、预设篇幅特征及预设可读性特征，所述预设逻辑特征包括预设文章结构特征、预设文章连贯性特征及预设文章元素特征，所述预设理解力特征包括与预设文章的相似度特征；评分单元403，用于将所述目标特征输入经过预训练的预设随机森林模型进行评分，以得到所述作文所对应的分数。
在一实施例中，所述抽取单元402包括：语言基础特征抽取子单元，用于根据预设语言基础特征抽取方式抽取所述作文的所述预设语言基础特征；逻辑特征抽取子单元，用于根据预设逻辑特征抽取方式抽取所述作文的所述预设逻辑特征；理解力特征抽取子单元，用于根据预设理解力特征抽取方式抽取所述作文的所述预设理解力特征。
在一实施例中，所述语言基础特征抽取子单元包括：分词子单元，用于采用第一预设语言工具对所述作文进行分词，以得到所述作文所包含的短语；子特征抽取单元，用于基于所述短语，根据每个预设目标语言基础子特征抽取方式进行特征抽取，以得到每个目标语言基础子特征；集合子单元，用于将得到的若干个目标语言基础子特征组成集合以得到所述预设语言基础特征；
在一实施例中，所述逻辑特征抽取子单元包括：切割子单元，用于根据所述作文中所包含的标点符号将所述作文切割为若干个句子；统计子单元，用于基于所述句子，统计出所述作文的所述预设文章结构特征所对应的所述作文本所包含的元素个数、所述预设文章元素特征所对应的所有所述元素中单词数目的均值、所述预设文章连贯性特征所对应的相邻句子之间的重叠数量，其中，所述元素为所述作文中所包含的背景类、论点类、主要观点类、支撑观点类或者结论类别中，最大连续出现的类别的句子集合称为该类别的元素。
在一实施例中，所述理解力特征抽取子单元包括：获取子单元，用于获取预设语料库中所包含的预设文章；对比子单元，用于将所述作文与所述预设文章按照预设特征指标进行对比，以得到所述作文与所述预设文章的预设相似度。
在一实施例中，所述预设文章为若干篇文章，所述理解力特征抽取子单元还包括：最相似文章获得子单元，用于从所述预设文章中获得与所述作文的相似度分数差距最小的最相似文章；对比指标获取单元，用于获取与所述最相似文章同属一个文章类别的得分指标为对比 指标；对比子单元，用于基于所述对比指标，将所述作文与得分最高的预设目标文章进行对比，以得到所述作文与所述预设目标文章之间的相似度。
在一实施例中，所述理解力特征抽取子单元还包括：主题获取子单元，用于获取所述作文所对应的主题；同主题文章获取子单元，用于获取所述预设语料库中所述主题所包含的同主题文章，其中，所述同主题文章为与所述作文的主题相同的文章；类别相似度获取子单元，用于将所述作文与所述同主题文章进行相似度对比，以得到所述作文与所述同主题文章所对应类别的类别相似度。
需要说明的是，所属领域的技术人员可以清楚地了解到，上述基于特征抽取及随机森林的作文评分装置和各单元的具体实现过程，可以参考前述方法实施例中的相应描述，为了描述的方便和简洁，在此不再赘述。
同时，上述基于特征抽取及随机森林的作文评分装置中各个单元的划分和连接方式仅用于举例说明，在其他实施例中，可将基于特征抽取及随机森林的作文评分装置按照需要划分为不同的单元，也可将基于特征抽取及随机森林的作文评分装置中各单元采取不同的连接顺序和方式，以完成上述基于特征抽取及随机森林的作文评分装置的全部或部分功能。
上述基于特征抽取及随机森林的作文评分装置可以实现为一种计算机程序的形式，该计算机程序可以在如图5所示的计算机设备上运行。
请参阅图5，图5是本申请实施例提供的一种计算机设备的示意性框图。该计算机设备500可以是台式机电脑或者服务器等计算机设备，也可以是其他设备中的组件或者部件。
参阅图5，该计算机设备500包括通过系统总线501连接的处理器502、存储器和网络接口505，其中，存储器可以包括非易失性存储介质503和内存储器504。
该非易失性存储介质503可存储操作系统5031和计算机程序5032。该计算机程序5032被执行时，可使得处理器502执行一种上述基于特征抽取及随机森林的作文评分方法。
该处理器502用于提供计算和控制能力，以支撑整个计算机设备500的运行。
该内存储器504为非易失性存储介质503中的计算机程序5032的运行提供环境，该计算机程序5032被处理器502执行时，可使得处理器502执行一种上述基于特征抽取及随机森林的作文评分方法。
该网络接口505用于与其它设备进行网络通信。本领域技术人员可以理解，图5中示出的结构，仅仅是与本申请方案相关的部分结构的框图，并不构成对本申请方案所应用于其上的计算机设备500的限定，具体的计算机设备500可以包括比图中所示更多或更少的部件，或者组合某些部件，或者具有不同的部件布置。例如，在一些实施例中，计算机设备可以仅包括存储器及处理器，在这样的实施例中，存储器及处理器的结构及功能与图5所示实施例一致，在此不再赘述。
其中，所述处理器502用于运行存储在存储器中的计算机程序5032，以实现本申请实施例所描述的作文评分方法。
应当理解，在本申请实施例中，处理器502可以是中央处理单元(Central ProcessingUnit，CPU)，该处理器502还可以是其他通用处理器、数字信号处理器(Digital Signal Processor，DSP)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现成可编程门阵列(Field-Programmable GateArray，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。其中，通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
本领域普通技术人员可以理解的是实现上述实施例的方法中的全部或部分流程，是可以通过计算机程序来完成，该计算机程序可存储于一计算机可读存储介质。该计算机程序被该计算机系统中的至少一个处理器执行，以实现上述方法的实施例的步骤。
因此，本申请实施例还提供一种计算机可读存储介质。该计算机可读存储介质可以为非易失性的计算机可读存储介质，也可以为易失性的计算机可读存储介质，该计算机可读存储介质存储有计算机程序，该计算机程序被处理器执行时使处理器执行以上各实施例中所描述 的所述基于特征抽取及随机森林的作文评分方法的步骤。
所述存储介质为实体的、非瞬时性的存储介质，例如可以是U盘、移动硬盘、只读存储器(Read-Only Memory，ROM)、磁碟或者光盘等各种可以存储计算机程序的实体存储介质。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、计算机软件或者二者的结合来实现，为了清楚地说明硬件和软件的可互换性，在上述说明中已经按照功能一般性地描述了各示例的组成及步骤。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
以上所述，仅为本申请的具体实施方式，但本申请明的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本申请揭露的技术范围内，可轻易想到各种等效的修改或替换，这些修改或替换都应涵盖在本申请的保护范围之内。因此，本申请的保护范围应以权利要求的保护范围为准。

Claims (20)

  1. 一种基于特征抽取及随机森林的作文评分方法，包括：
    获取待评分的作文；
    根据预设特征抽取方式抽取所述作文的目标特征，所述目标特征包括预设语言基础特征、预设逻辑特征及预设理解力特征，所述预设语言基础特征包括预设词汇特征、预设语法特征、预设句法特征、预设篇幅特征及预设可读性特征，所述预设逻辑特征包括预设文章结构特征、预设文章连贯性特征及预设文章元素特征，所述预设理解力特征包括与预设文章的相似度特征；
    将所述目标特征输入经过预训练的预设随机森林模型进行评分，以得到所述作文所对应的分数。
  2. 根据权利要求1所述基于特征抽取及随机森林的作文评分方法，其中，所述根据预设特征抽取方式抽取所述作文的目标特征的步骤包括：
    根据预设语言基础特征抽取方式抽取所述作文的所述预设语言基础特征；
    根据预设逻辑特征抽取方式抽取所述作文的所述预设逻辑特征；
    根据预设理解力特征抽取方式抽取所述作文的所述预设理解力特征。
  3. 根据权利要求2所述基于特征抽取及随机森林的作文评分方法，其中，所述根据预设语言基础特征抽取方式抽取所述作文的所述预设语言基础特征的步骤包括：
    采用第一预设语言工具对所述作文进行分词，以得到所述作文所包含的短语；
    基于所述短语，根据每个预设目标语言基础子特征抽取方式进行特征抽取，以得到每个目标语言基础子特征；
    将得到的若干个目标语言基础子特征组成集合以得到所述预设语言基础特征。
  4. 根据权利要求2所述基于特征抽取及随机森林的作文评分方法，其中，所述根据预设逻辑特征抽取方式抽取所述作文的所述预设逻辑特征的步骤包括：
    根据所述作文中所包含的标点符号将所述作文切割为若干个句子；
    基于所述句子，统计出所述作文的所述预设文章结构特征所对应的所述作文本所包含的元素个数、所述预设文章元素特征所对应的所有所述元素中单词数目的均值、所述预设文章连贯性特征所对应的相邻句子之间的重叠数量，其中，所述元素为所述作文中所包含的背景类、论点类、主要观点类、支撑观点类或者结论类别中，最大连续出现的类别的句子集合称为该类别的元素。
  5. 根据权利要求2所述基于特征抽取及随机森林的作文评分方法，其中，所述根据预设理解力特征抽取方式抽取所述作文的所述预设理解力特征的步骤包括：
    获取预设语料库中所包含的预设文章；
    将所述作文与所述预设文章按照预设特征指标进行对比，以得到所述作文与所述预设文章的预设相似度。
  6. 根据权利要求5所述基于特征抽取及随机森林的作文评分方法，其中，所述预设文章为若干篇文章，所述将所述作文与所述预设文章按照预设特征指标进行对比，以得到所述作文与所述预设文章的预设相似度的步骤之后，还包括：
    从所述预设文章中获得与所述作文的相似度分数差距最小的最相似文章；
    获取与所述最相似文章同属一个文章类别的得分指标为对比指标；
    基于所述对比指标，将所述作文与得分最高的预设目标文章进行对比，以得到所述作文与所述预设目标文章之间的相似度。
  7. 根据权利要求5所述基于特征抽取及随机森林的作文评分方法，其中，所述根据预设理解力特征抽取方式抽取所述作文的所述预设理解力特征的步骤还包括：
    获取所述作文所对应的主题；
    获取所述预设语料库中所述主题所包含的同主题文章，其中，所述同主题文章为与所述作文的主题相同的文章；
    将所述作文与所述同主题文章进行相似度对比，以得到所述作文与所述同主题文章所对应类别的类别相似度。
  8. 一种基于特征抽取及随机森林的作文评分装置，包括：
    获取单元，用于获取待评分的作文；
    抽取单元，用于根据预设特征抽取方式抽取所述作文的目标特征，所述目标特征包括预设语言基础特征、预设逻辑特征及预设理解力特征，所述预设语言基础特征包括预设词汇特征、预设语法特征、预设句法特征、预设篇幅特征及预设可读性特征，所述预设逻辑特征包括预设文章结构特征、预设文章连贯性特征及预设文章元素特征，所述预设理解力特征包括与预设文章的相似度特征；
    评分单元，用于将所述目标特征输入经过预训练的预设随机森林模型进行评分，以得到所述作文所对应的分数。
  9. 一种计算机设备，所述计算机设备包括存储器以及与所述存储器相连的处理器；所述存储器用于存储计算机程序；所述处理器用于运行所述计算机程序，以执行如下步骤：
    获取待评分的作文；
    根据预设特征抽取方式抽取所述作文的目标特征，所述目标特征包括预设语言基础特征、预设逻辑特征及预设理解力特征，所述预设语言基础特征包括预设词汇特征、预设语法特征、预设句法特征、预设篇幅特征及预设可读性特征，所述预设逻辑特征包括预设文章结构特征、预设文章连贯性特征及预设文章元素特征，所述预设理解力特征包括与预设文章的相似度特征；
    将所述目标特征输入经过预训练的预设随机森林模型进行评分，以得到所述作文所对应的分数。
  10. 根据权利要求9所述计算机设备，其中，所述根据预设特征抽取方式抽取所述作文的目标特征的步骤包括：
    根据预设语言基础特征抽取方式抽取所述作文的所述预设语言基础特征；
    根据预设逻辑特征抽取方式抽取所述作文的所述预设逻辑特征；
    根据预设理解力特征抽取方式抽取所述作文的所述预设理解力特征。
  11. 根据权利要求10所述计算机设备，其中，所述根据预设语言基础特征抽取方式抽取所述作文的所述预设语言基础特征的步骤包括：
    采用第一预设语言工具对所述作文进行分词，以得到所述作文所包含的短语；
    基于所述短语，根据每个预设目标语言基础子特征抽取方式进行特征抽取，以得到每个目标语言基础子特征；
    将得到的若干个目标语言基础子特征组成集合以得到所述预设语言基础特征。
  12. 根据权利要求10所述计算机设备，其中，所述根据预设逻辑特征抽取方式抽取所述作文的所述预设逻辑特征的步骤包括：
    根据所述作文中所包含的标点符号将所述作文切割为若干个句子；
    基于所述句子，统计出所述作文的所述预设文章结构特征所对应的所述作文本所包含的元素个数、所述预设文章元素特征所对应的所有所述元素中单词数目的均值、所述预设文章连贯性特征所对应的相邻句子之间的重叠数量，其中，所述元素为所述作文中所包含的背景类、论点类、主要观点类、支撑观点类或者结论类别中，最大连续出现的类别的句子集合称为该类别的元素。
  13. 根据权利要求10所述计算机设备，其中，所述根据预设理解力特征抽取方式抽取所述作文的所述预设理解力特征的步骤包括：
    获取预设语料库中所包含的预设文章；
    将所述作文与所述预设文章按照预设特征指标进行对比，以得到所述作文与所述预设文章的预设相似度。
  14. 根据权利要求13所述计算机设备，其中，所述预设文章为若干篇文章，所述将所述 作文与所述预设文章按照预设特征指标进行对比，以得到所述作文与所述预设文章的预设相似度的步骤之后，还包括：
    从所述预设文章中获得与所述作文的相似度分数差距最小的最相似文章；
    获取与所述最相似文章同属一个文章类别的得分指标为对比指标；
    基于所述对比指标，将所述作文与得分最高的预设目标文章进行对比，以得到所述作文与所述预设目标文章之间的相似度。
  15. 根据权利要求13所述计算机设备，其中，所述根据预设理解力特征抽取方式抽取所述作文的所述预设理解力特征的步骤还包括：
    获取所述作文所对应的主题；
    获取所述预设语料库中所述主题所包含的同主题文章，其中，所述同主题文章为与所述作文的主题相同的文章；
    将所述作文与所述同主题文章进行相似度对比，以得到所述作文与所述同主题文章所对应类别的类别相似度。
  16. 一种计算机可读存储介质，所述存储介质存储有计算机程序，所述计算机程序被处理器执行时实现如下步骤：
    获取待评分的作文；
    根据预设特征抽取方式抽取所述作文的目标特征，所述目标特征包括预设语言基础特征、预设逻辑特征及预设理解力特征，所述预设语言基础特征包括预设词汇特征、预设语法特征、预设句法特征、预设篇幅特征及预设可读性特征，所述预设逻辑特征包括预设文章结构特征、预设文章连贯性特征及预设文章元素特征，所述预设理解力特征包括与预设文章的相似度特征；
    将所述目标特征输入经过预训练的预设随机森林模型进行评分，以得到所述作文所对应的分数。
  17. 根据权利要求16所述计算机可读存储介质，其中，所述根据预设特征抽取方式抽取所述作文的目标特征的步骤包括：
    根据预设语言基础特征抽取方式抽取所述作文的所述预设语言基础特征；
    根据预设逻辑特征抽取方式抽取所述作文的所述预设逻辑特征；
    根据预设理解力特征抽取方式抽取所述作文的所述预设理解力特征。
  18. 根据权利要求17所述计算机可读存储介质，其中，所述根据预设语言基础特征抽取方式抽取所述作文的所述预设语言基础特征的步骤包括：
    采用第一预设语言工具对所述作文进行分词，以得到所述作文所包含的短语；
    基于所述短语，根据每个预设目标语言基础子特征抽取方式进行特征抽取，以得到每个目标语言基础子特征；
    将得到的若干个目标语言基础子特征组成集合以得到所述预设语言基础特征。
  19. 根据权利要求17所述计算机可读存储介质，其中，所述根据预设逻辑特征抽取方式抽取所述作文的所述预设逻辑特征的步骤包括：
    根据所述作文中所包含的标点符号将所述作文切割为若干个句子；
    基于所述句子，统计出所述作文的所述预设文章结构特征所对应的所述作文本所包含的元素个数、所述预设文章元素特征所对应的所有所述元素中单词数目的均值、所述预设文章连贯性特征所对应的相邻句子之间的重叠数量，其中，所述元素为所述作文中所包含的背景类、论点类、主要观点类、支撑观点类或者结论类别中，最大连续出现的类别的句子集合称为该类别的元素。
  20. 根据权利要求17所述计算机可读存储介质，其中，所述根据预设理解力特征抽取方式抽取所述作文的所述预设理解力特征的步骤包括：
    获取预设语料库中所包含的预设文章；
    将所述作文与所述预设文章按照预设特征指标进行对比，以得到所述作文与所述预设文 章的预设相似度。
PCT/CN2020/118737 2020-07-16 2020-09-29 作文评分方法、装置、计算机设备及计算机可读存储介质 Ceased WO2021139265A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010687313.4 2020-07-16
CN202010687313.4A CN111832281B (zh) 2020-07-16 2020-07-16 作文评分方法、装置、计算机设备及计算机可读存储介质

Publications (1)

Publication Number Publication Date
WO2021139265A1 true WO2021139265A1 (zh) 2021-07-15

Family

ID=72923310

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/118737 Ceased WO2021139265A1 (zh) 2020-07-16 2020-09-29 作文评分方法、装置、计算机设备及计算机可读存储介质

Country Status (2)

Country Link
CN (1) CN111832281B (zh)
WO (1) WO2021139265A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN121435986A (zh) * 2025-12-19 2026-01-30 中国科学院文献情报中心 一种论文价值量化评价方法与装置

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112561334A (zh) * 2020-12-16 2021-03-26 咪咕文化科技有限公司 阅读对象的评分方法、装置、电子设备及存储介质
CN113536769B (zh) * 2021-07-21 2024-10-15 深圳证券信息有限公司 文本简洁明晰程度评价方法及相关设备
CN113569560A (zh) * 2021-08-05 2021-10-29 北京师范大学 一种汉语二语作文自动评分方法
CN113836894B (zh) * 2021-09-26 2023-08-15 武汉天喻信息产业股份有限公司 多维度英语作文评分方法、装置及可读存储介质
CN115345149A (zh) * 2022-08-23 2022-11-15 华云(河北雄安)大数据科技有限公司 学生作文的评分方法、装置、计算设备及存储介质

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2005045786A1 (en) * 2003-10-27 2005-05-19 Educational Testing Service Automatic essay scoring system
CN107506360A (zh) * 2016-06-14 2017-12-22 科大讯飞股份有限公司 一种文章评分方法及系统
CN110069768A (zh) * 2018-01-22 2019-07-30 北京博智天下信息技术有限公司 一种基于篇章结构的英语议论文自动评分方法
CN110795538A (zh) * 2019-10-30 2020-02-14 腾讯科技（深圳）有限公司 一种基于人工智能的文本评分方法和相关设备
CN110851599A (zh) * 2019-11-01 2020-02-28 中山大学 一种中文作文自动评分方法及教辅系统
CN111104789A (zh) * 2019-11-22 2020-05-05 华中师范大学 文本评分方法、装置和系统

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1700200A (zh) * 2005-05-30 2005-11-23 梁茂成 英语作文自动评分系统
US10262547B2 (en) * 2014-11-10 2019-04-16 Educational Testing Service Generating scores and feedback for writing assessment and instruction using electronic process logs

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2005045786A1 (en) * 2003-10-27 2005-05-19 Educational Testing Service Automatic essay scoring system
CN107506360A (zh) * 2016-06-14 2017-12-22 科大讯飞股份有限公司 一种文章评分方法及系统
CN110069768A (zh) * 2018-01-22 2019-07-30 北京博智天下信息技术有限公司 一种基于篇章结构的英语议论文自动评分方法
CN110795538A (zh) * 2019-10-30 2020-02-14 腾讯科技（深圳）有限公司 一种基于人工智能的文本评分方法和相关设备
CN110851599A (zh) * 2019-11-01 2020-02-28 中山大学 一种中文作文自动评分方法及教辅系统
CN111104789A (zh) * 2019-11-22 2020-05-05 华中师范大学 文本评分方法、装置和系统

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN121435986A (zh) * 2025-12-19 2026-01-30 中国科学院文献情报中心 一种论文价值量化评价方法与装置

Also Published As

Publication number Publication date
CN111832281B (zh) 2024-11-26
CN111832281A (zh) 2020-10-27

Similar Documents

Publication Publication Date Title
WO2021139265A1 (zh) 作文评分方法、装置、计算机设备及计算机可读存储介质
FI125823B (en) A measure of the quality of machine translation
US9959776B1 (en) System and method for automated scoring of texual responses to picture-based items
CN116541493A (zh) 基于意图识别的交互应答方法、装置、设备、存储介质
US20130185057A1 (en) Computer-Implemented Systems and Methods for Scoring of Spoken Responses Based on Part of Speech Patterns
EP3617930A1 (en) Training data augmentation for convesational ai bots
CN105741831A (zh) 一种基于语法分析的口语评测方法和系统
US9652991B2 (en) Systems and methods for content scoring of spoken responses
CN101814066A (zh) 文本阅读难度判断设备及其方法
US10867525B1 (en) Systems and methods for generating recitation items
CN111832278A (zh) 文档流畅度的检测方法、装置、电子设备及介质
CN108280065B (zh) 一种外文文本评价方法及装置
KR101633556B1 (ko) 문법 오류 수정 장치 및 이를 이용한 문법 오류 수정 방법
Quispesaravia et al. Coh-Metrix-Esp: A complexity analysis tool for documents written in Spanish
TW201403354A (zh) 以資料降維法及非線性算則建構中文文本可讀性數學模型之系統及其方法
Walker 20 Variation analysis
CN115359799A (zh) 语音识别方法、训练方法、装置、电子设备及存储介质
CN115563959A (zh) 面向中文拼音拼写纠错的自监督预训练方法、系统及介质
Doewes et al. Individual Fairness Evaluation for Automated Essay Scoring System.
CN116030794A (zh) 听后问答评分方法、装置、终端及可读存储介质
KR102706374B1 (ko) 말하기 채점 인공지능 모델을 이용하여 문법력 및 어휘력을 평가하는 장치
KR20250008455A (ko) 말하기 능력의 평가 결과에 기반한 학습 콘텐츠 추천 장치
CN118468860A (zh) 拼写检查模型生成方法、装置、计算设备及存储介质
Nguyen et al. Predicting the proficiency level of nonnative Hebrew authors
CN115481216A (zh) 句子质量评价方法、装置、计算机设备和存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20911629

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20911629

Country of ref document: EP

Kind code of ref document: A1