CN111367986A - Joint information extraction method based on weak supervised learning - Google Patents

Joint information extraction method based on weak supervised learning Download PDF

Info

Publication number
CN111367986A
CN111367986A CN202010170467.6A CN202010170467A CN111367986A CN 111367986 A CN111367986 A CN 111367986A CN 202010170467 A CN202010170467 A CN 202010170467A CN 111367986 A CN111367986 A CN 111367986A
Authority
CN
China
Prior art keywords
information
supervised learning
extraction
method based
labeling
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.)
Pending
Application number
CN202010170467.6A
Other languages
Chinese (zh)
Inventor
王岚熙
姜同强
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.)
Beijing Technology and Business University
Original Assignee
Beijing Technology and Business University
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 Beijing Technology and Business University filed Critical Beijing Technology and Business University
Priority to CN202010170467.6A priority Critical patent/CN111367986A/en
Publication of CN111367986A publication Critical patent/CN111367986A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/25—Integrating or interfacing systems involving database management systems
    • G06F16/254—Extract, transform and load [ETL] procedures, e.g. ETL data flows in data warehouses
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/906—Clustering; Classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/02—Knowledge representation; Symbolic representation
    • G06N5/022—Knowledge engineering; Knowledge acquisition

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The invention relates to the technical field of information extraction, and discloses a joint information extraction method based on weak supervised learning, which comprises the following steps: s1: collecting information to form a training corpus, matching the training corpus with an entity pair in a knowledge base to obtain a training set, classifying the information in the training set, labeling the label according to the information containing characteristics of the information, and inputting the information subjected to multi-label labeling into a combined extraction model; s2: extracting the feature labels in the training set according to the information to be extracted, and labeling all feature labels on the target after the target is obtained; s3: and (4) putting the label information obtained in the step (S2) into a joint extraction model for extraction to obtain an extraction result. The combined information extraction method based on the weak supervised learning can solve the problem that the labeling of a data set is time-consuming and labor-consuming due to the current supervised learning/semi-supervised learning mode.

Description

Joint information extraction method based on weak supervised learning
Technical Field
The invention relates to the technical field of information extraction, in particular to a joint information extraction method based on weak supervised learning.
Background
The information technology in the web2.0 era has rapidly developed, and the advent of the internet has promoted an explosive increase in data volume. As a main carrier of information dissemination, these data carry much information of interest, and how to quickly and efficiently process large-scale unstructured data and obtain structured information becomes a hot spot of research at present, which is a main task of information extraction. The entity relationship extraction is an important branch of the information extraction field, has promotion significance in the aspect of theoretical research, and also has wide application value in the field of practical engineering application. Currently, entity relationship extraction mainly stays in modes based on supervised learning/semi-supervised learning and the like, and data set labeling caused by the supervised learning/semi-supervised learning mode is time-consuming and labor-consuming.
Disclosure of Invention
Technical problem to be solved
Aiming at the defects of the prior art, the invention provides a joint information extraction method based on weak supervised learning, which has the advantage of improving learning efficiency and solves the problem that the labeling of a data set is time-consuming and labor-consuming due to the current supervised learning/semi-supervised learning mode.
(II) technical scheme
In order to achieve the purpose of improving the surface stress, the invention provides the following technical scheme: a joint information extraction method based on weak supervised learning comprises the following steps:
s1: collecting information to form a training corpus, matching the training corpus with an entity pair in a knowledge base to obtain a training set, classifying the information in the training set, labeling the label according to the information containing characteristics of the information, and inputting the information subjected to multi-label labeling into a combined extraction model;
s2: extracting the feature labels in the training set according to the information to be extracted, and labeling all feature labels on the target after the target is obtained;
s3: and (4) putting the label information obtained in the step (S2) into a joint extraction model for extraction to obtain an extraction result.
Preferably, in the step S1, in the entity relationship extraction, relationship classification is generally performed based on verb phrases between two entities, and for one entity pair (4, B) and the trigger ρ therebetween, a prediction process of the relationship is defined as f (a, B, ρ) → (a, B, R), that is, the trigger related to the entity pair is mapped to a certain relationship by the extraction system.
Preferably, in the step S1, according to the prior knowledge of the entity information to be extracted, some tag classes, tag element sets, and generalization operations of set elements are predefined:
identification and extraction such as "company name" defines a Key token CLASS Key, a set of Key token elements [ Key _ T ] T ∈ (a company, finite, factory, etc. publicly identifiable token), and corresponding generalization operations, including both quantitative generalization and type generalization, generralize _ NUM1(Key _ T) ═ Key _ T (generalizing one Key _ T to 0 or more) and generralize _ CLASS (Key _ T) ═ Key _ (generalizing Key _ T to the wildcard Key-of a set of Key token elements).
Preferably, when labeling the information, a filtering mechanism is adopted to filter the labeling result, so as to reduce the number of error labels and improve the function of the extraction system, and the method specifically comprises the following steps:
a1: given the tagged information, predicting whether the instance expresses a relationship;
a2: for each set of entity pairs, predicting whether the entity pair is tagged;
a3: the instances marked in step S2 are filtered using the set of negative examples.
Preferably, in the step a1, the learning of the parameters of the hierarchical generative model is improved by information, i.e. an instance W is givenrsIndicates whether the s-th word sequence expresses the r-th relation, WrsIs a binary variable, if Wrs1 indicates that the word sequence s expresses the relation r, whereas Wrs0; in the step A2, according to the word sequence WrsPredicting whether the ith entity pair in the set is marked according to a knowledge base; in the step A3, an entity generation set which cannot express the relationship in the knowledge base but is often marked by errors is obtained by analyzing according to the training corpus, and then the predicted relationship in the step A2 is screened by using the set, so that the wrongly marked relationship examples in the weak supervised learning are effectively reduced through the two processes.
Preferably, in the step S1, the joint extraction model includes an embedding layer of vectors for mapping words in a high-dimensional discrete space to a low-dimensional continuous space, a two-way long-short term memory network (Bi-LSTM) coding layer for capturing semantic information of each word, a Conditional Random Field (CRF) decoding layer for labeling linear data sequences, and a complex optimization for comprehensively considering three characteristics of a calculation region, used information and a structural hierarchy, and during an experimental process, system performance evaluation is performed in two ways: retention evaluation and manual evaluation, and statistics of accuracy and recall rate; and evaluating the accurate performance of the system according to the N entity pairs with the most occurrence times.
Preferably, the retention assessment is in particular: randomly dividing the training corpus, and automatically identifying all the relational entities by an extraction system and comparing the relational entities with the entities in a knowledge base; only about 56.7% of the entities in the corpus are present in the knowledge base, so the remaining entities as noise data will have noise influence on the extraction performance, and therefore, we can evaluate the most common n entity pairs in the corpus to reduce the influence of the noise data on the final result to some extent. The significance of the retention evaluation is that a rough evaluation mode is adopted to carry out a plurality of experiments to obtain the value ranges of some key parameters of the extraction system; because the retention evaluation only roughly screens the entity pairs in the corpus, the entity pairs with less occurrence times are equivalent to noise data, the accuracy rate is sharply reduced along with the increase of the number of the entity pairs, and the system performance is increasingly poor. However, by retention assessment experiments without manual selection, the determination of the range of important parameters can be performed quickly.
Preferably, the manual evaluation is specifically: manually selecting various relationships with the highest occurrence frequency for testing, and avoiding the noise problem caused by retention evaluation; the accuracy performance of manual evaluation is obviously superior to that of retention evaluation, the whole performance of the extraction system is improved in the aspect of extracting the relation of the middle-frequency entity pair and the high-frequency entity pair due to the introduction of the word vector, and the good performance is still maintained under the condition that the number of the entity pairs is small when the word class analysis is triggered.
(III) advantageous effects
Compared with the prior art, the invention provides a joint information extraction method based on weak supervised learning, which has the following beneficial effects:
according to the combined information extraction method based on the weak supervised learning, the information is subjected to feature labeling through a strategy of combining the weak supervised learning with the combined information extraction, then the information is extracted in the combined extraction model, the accuracy and the recall rate of the information extraction are improved, and meanwhile the time and the energy required to be consumed are reduced.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example 1
A joint information extraction method based on weak supervised learning comprises the following steps:
s1: collecting information to form a training corpus, matching the training corpus with an entity pair in a knowledge base to obtain a training set, classifying the information in the training set, labeling the label according to the information containing characteristics of the information, and inputting the information subjected to multi-label labeling into a combined extraction model;
s2: extracting the feature labels in the training set according to the information to be extracted, and labeling all feature labels on the target after the target is obtained;
s3: and (4) putting the label information obtained in the step (S2) into a joint extraction model for extraction to obtain an extraction result.
In the step S1, in the entity relationship extraction, relationship classification is generally performed based on verb phrases between two entities, and for one entity pair (4, B) and the trigger ρ therebetween, a prediction process of the relationship is defined as f (a, B, ρ) → (a, B, R), that is, the trigger related to the entity pair is mapped to a certain relationship by the extraction system.
In the step S1, according to the prior knowledge of the entity information to be extracted, some tag classes, tag element sets, and generalization operations of set elements are predefined:
identification and extraction such as "company name" defines a Key token CLASS Key, a set of Key token elements [ Key _ T ] T ∈ (a company, finite, factory, etc. publicly identifiable token), and corresponding generalization operations, including both quantitative generalization and type generalization, generralize _ NUM1(Key _ T) ═ Key _ T (generalizing one Key _ T to 0 or more) and generralize _ CLASS (Key _ T) ═ Key _ (generalizing Key _ T to the wildcard Key-of a set of Key token elements).
When labeling information, a filtering mechanism is adopted to filter labeling results, the number of error labels is reduced, and the function of an extraction system is improved, and the method specifically comprises the following steps:
a1: given the tagged information, predicting whether the instance expresses a relationship;
a2: for each set of entity pairs, predicting whether the entity pair is tagged;
a3: the instances marked in step S2 are filtered using the set of negative examples.
In the step A1, improved learning of hierarchical model parameters is performed through information, namely, an example W is givenrsIndicates whether the s-th word sequence expresses the r-th relation, WrsIs a binary variable, if Wrs1 indicates that the word sequence s expresses the relation r, whereas Wrs0; in the step A2, according to the word sequence WrsPredicting whether the ith entity pair in the set is marked according to a knowledge base; in the step A3, an entity generation set which cannot express the relationship in the knowledge base but is often marked by errors is obtained by analyzing according to the training corpus, and then the predicted relationship in the step A2 is screened by using the set, so that the wrongly marked relationship examples in the weak supervised learning are effectively reduced through the two processes.
In the step S1, the joint extraction model includes an embedding layer of vectors for mapping words in a high-dimensional discrete space to a low-dimensional continuous space, a bidirectional long-short term memory network (Bi-LSTM) coding layer for capturing semantic information of each word, a Conditional Random Field (CRF) decoding layer for labeling linear data sequences, and a multiple optimization layer for comprehensively considering three characteristics of a calculation region, used information and a structural hierarchy, and during an experiment, the accuracy and recall rate are counted through retention evaluation; and evaluating the accurate performance of the system according to the N entity pairs with the most occurrence times.
The retention assessment, which is in particular: randomly dividing the training corpus, and automatically identifying all the relational entities by an extraction system and comparing the relational entities with the entities in a knowledge base; only about 56.7% of the entities in the corpus are present in the knowledge base, so the remaining entities as noise data will have noise influence on the extraction performance, and therefore, we can evaluate the most common n entity pairs in the corpus to reduce the influence of the noise data on the final result to some extent. The significance of the retention evaluation is that a rough evaluation mode is adopted to carry out a plurality of experiments to obtain the value ranges of some key parameters of the extraction system; because the retention evaluation only roughly screens the entity pairs in the corpus, the entity pairs with less occurrence times are equivalent to noise data, the accuracy rate is sharply reduced along with the increase of the number of the entity pairs, and the system performance is increasingly poor. However, by retention assessment experiments without manual selection, the determination of the range of important parameters can be performed quickly.
Example 2
A joint information extraction method based on weak supervised learning comprises the following steps:
s1: collecting information to form a training corpus, matching the training corpus with an entity pair in a knowledge base to obtain a training set, classifying the information in the training set, labeling the label according to the information containing characteristics of the information, and inputting the information subjected to multi-label labeling into a combined extraction model;
s2: extracting the feature labels in the training set according to the information to be extracted, and labeling all feature labels on the target after the target is obtained;
s3: and (4) putting the label information obtained in the step (S2) into a joint extraction model for extraction to obtain an extraction result.
In the step S1, in the entity relationship extraction, relationship classification is generally performed based on verb phrases between two entities, and for one entity pair (4, B) and the trigger ρ therebetween, a prediction process of the relationship is defined as f (a, B, ρ) → (a, B, R), that is, the trigger related to the entity pair is mapped to a certain relationship by the extraction system.
In the step S1, according to the prior knowledge of the entity information to be extracted, some tag classes, tag element sets, and generalization operations of set elements are predefined:
identification and extraction such as "company name" defines a Key token CLASS Key, a set of Key token elements [ Key _ T ] T ∈ (a company, finite, factory, etc. publicly identifiable token), and corresponding generalization operations, including both quantitative generalization and type generalization, generralize _ NUM1(Key _ T) ═ Key _ T (generalizing one Key _ T to 0 or more) and generralize _ CLASS (Key _ T) ═ Key _ (generalizing Key _ T to the wildcard Key-of a set of Key token elements).
When labeling information, a filtering mechanism is adopted to filter labeling results, the number of error labels is reduced, and the function of an extraction system is improved, and the method specifically comprises the following steps:
a1: given the tagged information, predicting whether the instance expresses a relationship;
a2: for each set of entity pairs, predicting whether the entity pair is tagged;
a3: the instances marked in step S2 are filtered using the set of negative examples.
In the step A1, improved learning of hierarchical model parameters is performed through information, namely, an example W is givenrsIndicates whether the s-th word sequence expresses the r-th relation, WrsIs a binary variable, if Wrs1 indicates that the word sequence s expresses the relation r, whereas Wrs0; in the step A2, according to the word sequence WrsPredicting whether the ith entity pair in the set is marked according to a knowledge base; in the step A3, an entity generation set which cannot express the relationship in the knowledge base but is often marked by errors is obtained by analyzing according to the training corpus, and then the predicted relationship in the step A2 is screened by using the set, so that the wrongly marked relationship examples in the weak supervised learning are effectively reduced through the two processes.
In the step S1, the joint extraction model includes an embedding layer of vectors for mapping words in a high-dimensional discrete space to a low-dimensional continuous space, a bidirectional long-short term memory network (Bi-LSTM) coding layer for capturing semantic information of each word, a Conditional Random Field (CRF) decoding layer for labeling linear data sequences, and a multiple optimization layer for comprehensively considering three characteristics of a calculation region, used information and a structural hierarchy, and during an experimental process, the accuracy and recall rate are counted through manual evaluation; and evaluating the accurate performance of the system according to the N entity pairs with the most occurrence times.
The manual evaluation specifically comprises: manually selecting various relationships with the highest occurrence frequency for testing, and avoiding the noise problem caused by retention evaluation; the accuracy performance of manual evaluation is obviously superior to that of retention evaluation, the whole performance of the extraction system is improved in the aspect of extracting the relation of the middle-frequency entity pair and the high-frequency entity pair due to the introduction of the word vector, and the good performance is still maintained under the condition that the number of the entity pairs is small when the word class analysis is triggered.
In conclusion, the combined information extraction method based on the weak supervised learning combines the strategy of combining the weak supervised learning with the combined information extraction to perform feature labeling on the information and then extract the information in the combined extraction model, so that the accuracy and the recall rate of information extraction are improved, and the time and the energy required to be consumed are reduced.
It is to be noted that the term "comprises," "comprising," or any other variation thereof is intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.

Claims (9)

1. A joint information extraction method based on weak supervised learning is characterized by comprising the following steps:
s1: collecting information to form a training corpus, matching the training corpus with an entity pair in a knowledge base to obtain a training set, classifying the information in the training set, labeling the label according to the information containing characteristics of the information, and inputting the information subjected to multi-label labeling into a combined extraction model;
s2: extracting the feature labels in the training set according to the information to be extracted, and labeling all feature labels on the target after the target is obtained;
s3: and (4) putting the label information obtained in the step (S2) into a joint extraction model for extraction to obtain an extraction result.
2. The joint information extraction method based on weak supervised learning as claimed in claim 1, wherein: in the step S1, in the entity relationship extraction, relationship classification is generally performed based on verb phrases between two entities, and for one entity pair (4, B) and the trigger ρ therebetween, a prediction process of the relationship is defined as f (a, B, ρ) → (a, B, R), that is, the trigger related to the entity pair is mapped to a certain relationship by the extraction system.
3. The joint information extraction method based on weak supervised learning as claimed in claim 1, wherein: in the step S1, some mark classes, mark element sets, and generalization operations of set elements are predefined according to the prior knowledge of the entity information to be extracted.
4. The joint information extraction method based on weak supervised learning as claimed in claim 1, wherein: when labeling information, a filtering mechanism is adopted to filter labeling results, the number of error labels is reduced, and the function of an extraction system is improved, and the method specifically comprises the following steps:
a1: given the tagged information, predicting whether the instance expresses a relationship;
a2: for each set of entity pairs, predicting whether the entity pair is tagged;
a3: the instances marked in step S2 are filtered using the set of negative examples.
5. The joint information extraction method based on weak supervised learning as claimed in claim 4, wherein: in the step A1, improved learning of hierarchical model parameters is performed through information, namely, an example W is givenrsIndicates whether the s-th word sequence expresses the r-th relation, WrsIs a binary variable, if Wrs1 indicates that the word sequence s expresses the relation r, whereas Wrs0; in the step A2, according to the word sequence WrsPredicting whether the ith entity pair in the set is marked according to a knowledge base; in the step A3, an entity generation set which cannot express the relationship in the knowledge base but is often marked by errors is obtained by analyzing according to the training corpus, and then the predicted relationship in the step A2 is screened by using the set, so that the wrongly marked relationship examples in the weak supervised learning are effectively reduced through the two processes.
6. The joint information extraction method based on weak supervised learning as claimed in claim 1, wherein: in the step S1, the joint extraction model includes an embedding layer of vectors for mapping words in a high-dimensional discrete space to a low-dimensional continuous space, a two-way long-short term memory network (Bi-LSTM) encoding layer for capturing semantic information of each word, a Conditional Random Field (CRF) decoding layer for labeling linear data sequences, and a composite attention layer for comprehensively considering three parts of computation regions, information used and structural hierarchy characteristics.
7. The joint information extraction method based on weak supervised learning as claimed in claim 1, wherein: during the experiment, system performance evaluation was performed in two ways: and (4) retention evaluation and manual evaluation, and counting the accuracy and recall rate.
8. The joint information extraction method based on weak supervised learning as claimed in claim 7, wherein: the retention assessment, which is in particular: and randomly dividing the training corpus, and automatically identifying all the relational entities by an extraction system and comparing the relational entities with the entities in the knowledge base.
9. The joint information extraction method based on weak supervised learning as claimed in claim 7, wherein: the manual evaluation specifically comprises: and a plurality of relationships with the highest occurrence frequency are manually selected for testing, so that the noise problem caused by retention evaluation is avoided.
CN202010170467.6A 2020-03-12 2020-03-12 Joint information extraction method based on weak supervised learning Pending CN111367986A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010170467.6A CN111367986A (en) 2020-03-12 2020-03-12 Joint information extraction method based on weak supervised learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010170467.6A CN111367986A (en) 2020-03-12 2020-03-12 Joint information extraction method based on weak supervised learning

Publications (1)

Publication Number Publication Date
CN111367986A true CN111367986A (en) 2020-07-03

Family

ID=71208854

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010170467.6A Pending CN111367986A (en) 2020-03-12 2020-03-12 Joint information extraction method based on weak supervised learning

Country Status (1)

Country Link
CN (1) CN111367986A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112364174A (en) * 2020-10-21 2021-02-12 山东大学 Patient medical record similarity evaluation method and system based on knowledge graph
CN113505229A (en) * 2021-09-09 2021-10-15 北京道达天际科技有限公司 Entity relationship extraction model training method and device

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112364174A (en) * 2020-10-21 2021-02-12 山东大学 Patient medical record similarity evaluation method and system based on knowledge graph
CN113505229A (en) * 2021-09-09 2021-10-15 北京道达天际科技有限公司 Entity relationship extraction model training method and device

Similar Documents

Publication Publication Date Title
CN107229668B (en) Text extraction method based on keyword matching
CN106991085B (en) Entity abbreviation generation method and device
CN110458324B (en) Method and device for calculating risk probability and computer equipment
CN111400499A (en) Training method of document classification model, document classification method, device and equipment
CN110287292B (en) Judgment criminal measuring deviation degree prediction method and device
Das et al. A graph based clustering approach for relation extraction from crime data
CN111367986A (en) Joint information extraction method based on weak supervised learning
CN111737477A (en) Intellectual property big data-based intelligence investigation method, system and storage medium
CN114398891B (en) Method for generating KPI curve and marking wave band characteristics based on log keywords
CN114494982B (en) Live video big data accurate recommendation method and system based on artificial intelligence
CN112784580A (en) Financial data analysis method and device based on event extraction
CN109543038B (en) Emotion analysis method applied to text data
CN105808602B (en) Method and device for detecting junk information
CN115794803B (en) Engineering audit problem monitoring method and system based on big data AI technology
CN117391084A (en) Data management method and system based on DCMM system and deep learning
CN116821903A (en) Detection rule determination and malicious binary file detection method, device and medium
CN111400606B (en) Multi-label classification method based on global and local information extraction
CN109739840A (en) Data processing empty value method, apparatus and terminal device
CN110941713A (en) Self-optimization financial information plate classification method based on topic model
CN114154829A (en) Method, device, terminal and storage medium for determining industrial chain nodes of enterprise
CN114579468A (en) Source item selection software defect prediction method based on semantic metric value
CN113822069A (en) Emergency early warning method and device based on meta-knowledge and electronic device
CN113962216A (en) Text processing method and device, electronic equipment and readable storage medium
Xie et al. Pattern mining in visual concept streams
CN104933185B (en) Wikipedia quality of entry evaluation method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
WD01 Invention patent application deemed withdrawn after publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20200703