CN111914566A - Automatic comment generation method - Google Patents
Automatic comment generation method Download PDFInfo
- Publication number
- CN111914566A CN111914566A CN202010754601.7A CN202010754601A CN111914566A CN 111914566 A CN111914566 A CN 111914566A CN 202010754601 A CN202010754601 A CN 202010754601A CN 111914566 A CN111914566 A CN 111914566A
- Authority
- CN
- China
- Prior art keywords
- comment
- content
- comment information
- evaluation
- information
- 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
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Abstract
The invention provides an automatic comment generation method, and relates to the field of machine learning. A comment automatic generation method comprises the steps of obtaining evaluation content, inputting the evaluation content into a GPT-2 training model to determine reply content, wherein the GPT-2 training model is obtained through machine learning training by using multiple groups of comment data, the multiple groups of comment data comprise first-class data and second-class data, each group of data of the first-class data comprises comment information and attribute identification for identifying content attributes of the comment information, and each group of data of the second-class data comprises the comment information and quality identification for identifying content quality of the multiple groups of comment information; and selecting the attribute identifier according to the industry attribute of the evaluation content to determine the selection range of the comment information, and selecting the content of the comment information according to the quality identifier to generate reply content. The method and the device can migrate information in different fields and are suitable for commenting contents.
Description
Technical Field
The invention relates to the field of machine learning, in particular to an automatic comment generation method.
Background
The construction of internet user communities is rapidly increased, and in order to increase the liveness and experience of users, the user contents need to be commented in time. In most of the conventional techniques, manual work or manual template operation is used, which is labor-consuming. And a semantic retrieval technology is also used for searching for proper comments from the existing comment libraries, and the technical mode needs to maintain a large-scale comment library, can only generate the existing comments and cannot migrate the unknown fields.
With the development of deep learning in the field of natural language, NLG (natural language generation) technology has also been greatly developed,
particularly, the GPT2 is generated with more striking effect in recent two years. However, in this model, automatic comments need to be converted into a dialogue field in a short text comment scenario, but this method is suitable for chatting, and is not suitable for comment, evaluation, and discussion-like comment content.
Disclosure of Invention
The invention aims to provide an automatic comment generation method which can migrate information in different fields and is suitable for commenting contents.
The embodiment of the invention is realized by the following steps:
the embodiment of the application provides an automatic comment generation method, which comprises the steps of obtaining evaluation content, inputting the evaluation content into a GPT-2 training model to determine reply content, wherein the GPT-2 training model is obtained by using multiple groups of comment data through machine learning training, the multiple groups of comment data comprise first-class data and second-class data, each group of data of the first-class data comprises comment information and attribute identification for identifying content attributes of the comment information, and each group of data of the second-class data comprises the comment information and quality identification for identifying content quality of the multiple groups of comment information; and selecting the attribute identification according to the industry attribute of the evaluation content to determine the selection range of the comment information, and selecting the content of the comment information according to the quality identification to generate the reply content.
In some embodiments of the present invention, the content quality evaluates the comment information according to one or more of a length, a picture, a paragraph, a title, a font, and an opinion guide of the comment information, and generates a quality indicator.
In some embodiments of the invention, the aforementioned opinion guide includes any one or more of a belief, opinion, feeling, opinion, approval, support, objection, consent, general, and general.
In some embodiments of the present invention, the industry attribute is analyzed according to the evaluation content, and the selection range of the comment information is determined according to a matching degree between the industry attribute and the attribute identifier.
In some embodiments of the present invention, the comment information is obtained through a news information website.
In some embodiments of the present invention, the news information website includes any one or more of news evaluation wording, internet social evaluation, and wikipedia.
In some embodiments of the present invention, semantic analysis is performed on the contents of the plurality of sets of comment information, and the frequency of occurrence of similar contents is calculated according to the matching degree between the plurality of sets of comment information and the evaluation content, thereby generating the reply content.
In some embodiments of the present invention, the comment information includes an evaluation term and a comment.
In some embodiments of the present invention, an evaluation term template is generated according to an appearance frequency of the evaluation term in the comment information, and the evaluation term template is stored in an evaluation term database, so that the evaluation term is called by the evaluation term template and the reply content is generated.
In some embodiments of the present invention, a comment template is generated according to the frequency of occurrence of the comment in the comment information, and the comment template is stored in a comment database, so that the comment is called by the comment wording template and the reply content is generated.
Compared with the prior art, the embodiment of the invention has at least the following advantages or beneficial effects:
a comment automatic generation method comprises the steps of obtaining evaluation content, inputting the evaluation content into a GPT-2 training model to determine reply content, wherein the GPT-2 training model is obtained through machine learning training by using multiple groups of comment data, the multiple groups of comment data comprise first-class data and second-class data, each group of data of the first-class data comprises comment information and attribute identification for identifying content attributes of the comment information, and each group of data of the second-class data comprises the comment information and quality identification for identifying content quality of the multiple groups of comment information; and selecting the attribute identifier according to the industry attribute of the evaluation content to determine the selection range of the comment information, and selecting the content of the comment information according to the quality identifier to generate reply content.
According to the method, the evaluation content is acquired and input into the GPT-2 training model to determine the reply content of the evaluation content, so that the reply of the commenting content is realized; comment information in the first type of data and attribute identification for identifying content attributes of the comment information, wherein the first type of data obtains a GPT-2 training model through machine learning, so that information in different fields is migrated, and new reply content is generated; comment information in comment data in the second type of data and quality identification for identifying content quality of a plurality of groups of comment information, wherein the second type of data obtains a GPT-2 training model through machine learning, so that reply content is generated by utilizing the content quality of the comment information, the reply content is higher in fidelity, and the reply content meets user requirements better; selecting an attribute identifier according to the industry attribute of the evaluation content to determine the selection range of the comment information, so that more appropriate reply content in the comment information is selected by utilizing the industry attribute of the evaluation content, and the accuracy of the reply content is improved; and selecting the content of the comment information according to the quality identification to generate the reply content, so that the content of the comment information is selected to be combined into new reply content, and the value of the reply content is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.
Fig. 1 is a schematic flow diagram of an automatic comment generation method according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some embodiments of the present application, but not all embodiments. The components of the embodiments of the present application, generally described and illustrated in the figures herein, can be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the present application, presented in the accompanying drawings, is not intended to limit the scope of the claimed application, but is merely representative of selected embodiments of the application. 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 application.
Some embodiments of the present application will be described in detail below with reference to the accompanying drawings. The embodiments described below and the individual features of the embodiments can be combined with one another without conflict.
Examples
Referring to fig. 1, fig. 1 is a schematic flow chart illustrating an automatic comment generation method according to an embodiment of the present application. The comment automatic generation method comprises the steps of obtaining comment contents, inputting the comment contents into a GPT-2 training model to determine reply contents, wherein the GPT-2 training model is obtained through machine learning training by using multiple groups of comment data, the multiple groups of comment data comprise first-class data and second-class data, each group of data of the first-class data comprises comment information and attribute identification for identifying content attributes of the comment information, and each group of data of the second-class data comprises the comment information and quality identification for identifying content quality of the multiple groups of comment information; and selecting the attribute identification according to the industry attribute of the evaluation content to determine the selection range of the comment information, and selecting the content of the comment information according to the quality identification to generate the reply content.
In detail, the GPT-2 training model determines the reply content according to the evaluation content, and automatically generated comments can be obtained by inputting user content Comment. The GPT-2 training model is obtained by machine learning training by using multiple groups of comment data. The GPT-2 training model is a language model that predicts the next word from the above, so it can use the knowledge that pre-training has learned to generate text, such as generating news. Other data may also be used for fine-tuning to generate text in a particular format or theme, such as poetry and drama.
The multiple sets of comment data comprise first type data and second type data. Each group of data of the first type of data comprises comment information and attribute identification for identifying content attributes of the comment information, and each group of data of the second type of data comprises the comment information and quality identification for identifying content quality of multiple groups of comment information. And selecting the attribute identifier according to the industry attribute of the evaluation content to determine the selection range of the comment information, and selecting the content of the comment information according to the quality identifier to generate reply content. In detail, the attribute identifier is added to the first type of data through the content attribute of the comment information, and the quality identifier is added to the second type of data through the content quality of the comment information, so that the comment information meeting the requirement of the evaluation content is searched by using the attribute identifier, valuable content in the comment information is screened by using the content quality of the comment information, and then reply content of the evaluation content is generated. And selecting the attribute identification through the industry attribute of the evaluation content, so that the industry attribute of the evaluation content is utilized to select the comment information with pertinence. The industry attribute may be a keyword of different application fields, topics or application modes. The industry attribute can be obtained by the evaluation content, and can also be input through a data expansion technology and manual assistance. The industry attribute can be multiple, and the same industry attribute can correspond to multiple attribute identifications.
In detail, the content of the comment information may include all the content or only the comment language. Since the proportion of the commenting content is small relative to other data, when a plurality of sets of commenting data are collected, sentences belonging to the commenting content can be screened through a specific identifier such as a quality identifier.
It is to be understood that the flow shown in fig. 1 is merely illustrative, and that the comment automatic generation method may further include more or fewer steps than shown in fig. 1, or have a different configuration than shown in fig. 1. The components of the flow chart shown in fig. 1 may be implemented in hardware, software, or a combination thereof.
In the application scenario described above, the action description of the embodiment of the present application may be executed by a processor, or may be executed by a user terminal, or may also be executed by a part of the user terminal and a part of the processor. The present application is not limited in terms of the execution subject as long as the actions disclosed in the embodiments of the present application are executed.
In some embodiments of the present invention, the content quality evaluates the comment information according to one or more of a length, a picture, a paragraph, a title, a font, and an opinion guide of the comment information, and generates a quality indicator.
In detail, the content quality can be evaluated according to the length of the comment information, whether a picture is inserted, paragraph distribution rules, whether subtitles exist between paragraphs, whether the paragraphs are special fonts and whether an opinion guide is included, so that the editing fidelity of the comment information is judged. Wherein, the special font can be blackened, italicized, bolded or colored. The rating may be based on the content quality, whereby the quality identifications are labeled according to different levels of the content quality. Therefore, whether the quality marks exist or not or the number and the grade of the quality marks are utilized to carry out screening, so that the symbolic content of the comment related field is quickly found out, and further the comment content with depth is generated.
In some embodiments of the invention, the aforementioned opinion guide includes any one or more of a belief, opinion, feeling, opinion, approval, support, objection, consent, general, and general.
In detail, the opinion guide includes words or sentences expressing viewpoints, which can be expressed by words such as "think", "feel", "agree", "support", and "idea". Subjects such as "i think" and "writer think" may also be added to quickly screen out the comments expressing the opinions in the comment information.
In some embodiments of the present invention, the industry attribute is analyzed according to the evaluation content, and the selection range of the comment information is determined according to a matching degree between the industry attribute and the attribute identifier.
In detail, the industry attribute is analyzed according to the evaluation content, so that the matching degree is calculated according to the industry attribute of the evaluation content and the attribute identification of the comment information, and the evaluation information in the same or similar fields is selected. The comment information and the evaluation information can be matched through a semantic matching technology.
In some embodiments of the present invention, the comment information is obtained through a news information website. Optionally, the news information website includes any one or more of news evaluation wording, internet social evaluation, and wikipedia.
Optionally, the evaluation terms or comments of the comment information may be obtained through a news information website. GPT2 is a large-scale language model based on a transform Decoder, requires a very large high-quality evaluation term, and is linguistically biased in consideration of the fact that an application scenario is comment generation. Therefore, in addition to Chinese evaluation terms such as news evaluation terms, Internet social evaluation, wiki and the like, the comment information of the Internet information is also captured to enrich the training evaluation terms.
In some embodiments of the present invention, the contents of the plurality of sets of comment information are subjected to semantic analysis, and the frequency of occurrence of similar contents is calculated according to the matching degree of the plurality of sets of comment information with the evaluation content, thereby generating the reply content.
In detail, semantic analysis is performed on the content of the comment information through a semantic matching technology, so that the same or similar content is matched with the evaluation content, comment information with high similarity is screened out according to the matching degree of the comment information and the evaluation content, and the comment information is selected according to similar content with high occurrence frequency, so that reply content is generated. And finding out similar content through the matching degree, and predicting the probability of the next statement according to the similar content.
In some embodiments of the present invention, the comment information includes an evaluation term and a comment. Optionally, an evaluation term template is generated according to the frequency of occurrence of the evaluation term in the comment information, and the evaluation term template is stored in an evaluation term database, so that the evaluation term is called by the evaluation term template and the reply content is generated. In detail, the comment is conveniently and directly added into the evaluation term template by calling the evaluation term template, so that the content replied by the comment is humanized. Alternatively, reviews may be washed individually with human assistance to remove low quality and advertising-like content.
In some embodiments of the present invention, a comment template is generated according to the frequency of occurrence of the comment in the comment information, and the comment template is stored in a comment database, so that the comment is called by the comment wording template and the reply content is generated. In detail, by calling the comment template, the corresponding comment is directly called to generate the reply content according to the attribute identification and is added to the evaluation term, so that the reply efficiency of the evaluation content is improved.
In summary, according to the method for automatically generating comments provided by the embodiment of the application, the comment content is acquired and input into the GPT-2 training model to determine the reply content of the comment content, so that the reply of the comment content is realized; comment information in the first type of data and attribute identification for identifying content attributes of the comment information, wherein the first type of data obtains a GPT-2 training model through machine learning, so that information in different fields is migrated, and new reply content is generated; comment information in comment data in the second type of data and quality identification for identifying content quality of a plurality of groups of comment information, wherein the second type of data obtains a GPT-2 training model through machine learning, so that reply content is generated by utilizing the content quality of the comment information, the reply content is higher in fidelity, and the reply content meets user requirements better; selecting an attribute identifier according to the industry attribute of the evaluation content to determine the selection range of the comment information, so that more appropriate reply content in the comment information is selected by utilizing the industry attribute of the evaluation content, and the accuracy of the reply content is improved; and selecting the content of the comment information according to the quality identification to generate the reply content, so that the content of the comment information is selected to be combined into new reply content, and the value of the reply content is improved. The method has the advantages that the content quality is judged through the attribute identification and the quality identification, deep and targeted comment contents can be automatically generated, meanwhile, the GPT-2 training model of large-scale evaluation expressions is benefited, and the method has the advantage of low migration cost for cross-domain content generation. The GPT-2 training model can be intuitively guided to comment on specific contents in the comment generation field, and the limitation that the deep learning field cannot be explained and fine-tuned is avoided.
In the embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. The apparatus embodiments described above are merely illustrative, and for example, the block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
In addition, functional modules in the embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.
The above-described functions, if implemented in the form of software functional modules and sold or used as a separate product, may be stored in a computer-readable storage medium. Based on such understanding, the technical solution of the present application or portions thereof that substantially contribute to the prior art may be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the above-described method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
The above description is only a preferred embodiment of the present application and is not intended to limit the present application, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present application shall be included in the protection scope of the present application.
Claims (10)
1. An automatic comment generation method is characterized by comprising the following steps: obtaining evaluation content, and inputting the evaluation content into a GPT-2 training model to determine reply content, wherein the GPT-2 training model is obtained by using multiple groups of comment data through machine learning training, the multiple groups of comment data comprise first-class data and second-class data, each group of data of the first-class data comprises comment information and an attribute identifier for identifying the content attribute of the comment information, and each group of data of the second-class data comprises comment information and a quality identifier for identifying the content quality of the multiple groups of comment information; and selecting the attribute identification according to the industry attribute of the evaluation content to determine the selection range of the comment information, and selecting the content of the comment information according to the quality identification to generate the reply content.
2. The automatic comment generation method of claim 1, wherein the content quality evaluates the comment information in terms of one or more of a length, a picture, a paragraph, a title, a font, and an opinion guide of the comment information and generates a quality indicator.
3. A method of automatically generating comments of claim 2 wherein the opinion leaders include any one or more of opinions, feelings, opinions, endorsements, supports, objections, assertions, generalizations, and generalizations.
4. The method for automatically generating comments, according to claim 1, wherein the industry attribute is analyzed according to the evaluation content, and the selection range of the comment information is determined according to the matching degree of the industry attribute and the attribute identification.
5. The method for automatically generating comments of claim 1, wherein the comment information is obtained through a news information-based website.
6. The method as claimed in claim 5, wherein the news information website contains any one or more of news evaluation wording, internet social evaluation and wikipedia.
7. The automatic comment generation method according to claim 5, wherein a plurality of sets of contents of the comment information are subjected to semantic analysis, and the frequency of occurrence of similar contents is calculated based on the degree of matching between a plurality of sets of the comment information and the comment contents, thereby generating reply contents.
8. The automatic comment generation method according to claim 7, wherein the comment information contains an evaluation wording and a comment.
9. The automatic comment generation method according to claim 8, wherein an evaluation term template is generated based on the frequency of occurrence of the evaluation term in the comment information, and the evaluation term template is stored in an evaluation term database, so that the evaluation term is called by the evaluation term template and the reply content is generated.
10. A comment automatic generation method according to claim 8 or 9, wherein a comment template is generated from the frequency of occurrence of the comment in the comment information, and the comment template is stored in a comment database, whereby the comment is called by the comment wording template and the reply content is generated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010754601.7A CN111914566A (en) | 2020-07-30 | 2020-07-30 | Automatic comment generation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010754601.7A CN111914566A (en) | 2020-07-30 | 2020-07-30 | Automatic comment generation method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111914566A true CN111914566A (en) | 2020-11-10 |
Family
ID=73286758
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010754601.7A Pending CN111914566A (en) | 2020-07-30 | 2020-07-30 | Automatic comment generation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111914566A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113515719A (en) * | 2021-07-29 | 2021-10-19 | 北京百度网讯科技有限公司 | Method and device for cold starting community products |
| CN115730030A (en) * | 2021-08-26 | 2023-03-03 | 腾讯科技（深圳）有限公司 | Comment information processing method and related device |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001325104A (en) * | 2000-05-12 | 2001-11-22 | Mitsubishi Electric Corp | Method and device for inferring language case and recording medium recording language case inference program |
| US20160292288A1 (en) * | 2015-03-31 | 2016-10-06 | Linkedin Corporation | Comments analyzer |
| CN109117470A (en) * | 2017-06-22 | 2019-01-01 | 北京国双科技有限公司 | A kind of evaluation relation extracting method and device for evaluating text information |
| CN110569334A (en) * | 2019-09-11 | 2019-12-13 | 北京搜狐新动力信息技术有限公司 | method and device for automatically generating comments |
| CN111324713A (en) * | 2020-02-18 | 2020-06-23 | 腾讯科技（深圳）有限公司 | Automatic replying method and device for conversation, storage medium and computer equipment |
| CN111353024A (en) * | 2018-12-04 | 2020-06-30 | 阿里巴巴集团控股有限公司 | Method, device and system for generating comment text |
-
2020
- 2020-07-30 CN CN202010754601.7A patent/CN111914566A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001325104A (en) * | 2000-05-12 | 2001-11-22 | Mitsubishi Electric Corp | Method and device for inferring language case and recording medium recording language case inference program |
| US20160292288A1 (en) * | 2015-03-31 | 2016-10-06 | Linkedin Corporation | Comments analyzer |
| CN109117470A (en) * | 2017-06-22 | 2019-01-01 | 北京国双科技有限公司 | A kind of evaluation relation extracting method and device for evaluating text information |
| CN111353024A (en) * | 2018-12-04 | 2020-06-30 | 阿里巴巴集团控股有限公司 | Method, device and system for generating comment text |
| CN110569334A (en) * | 2019-09-11 | 2019-12-13 | 北京搜狐新动力信息技术有限公司 | method and device for automatically generating comments |
| CN111324713A (en) * | 2020-02-18 | 2020-06-23 | 腾讯科技（深圳）有限公司 | Automatic replying method and device for conversation, storage medium and computer equipment |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113515719A (en) * | 2021-07-29 | 2021-10-19 | 北京百度网讯科技有限公司 | Method and device for cold starting community products |
| CN113515719B (en) * | 2021-07-29 | 2024-01-12 | 北京百度网讯科技有限公司 | Method and device for cold starting community products |
| CN115730030A (en) * | 2021-08-26 | 2023-03-03 | 腾讯科技（深圳）有限公司 | Comment information processing method and related device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108287858B (en) | Semantic extraction method and device for natural language | |
| RU2628431C1 (en) | Selection of text classifier parameter based on semantic characteristics | |
| CN107480143B (en) | Method and system for segmenting conversation topics based on context correlation | |
| RU2628436C1 (en) | Classification of texts on natural language based on semantic signs | |
| CN107818085B (en) | Answer selection method and system for reading understanding of reading robot | |
| US20180032508A1 (en) | Aspect-based sentiment analysis using machine learning methods | |
| US10198432B2 (en) | Aspect-based sentiment analysis and report generation using machine learning methods | |
| CN112749326B (en) | Information processing method, information processing device, computer equipment and storage medium | |
| RU2639655C1 (en) | System for creating documents based on text analysis on natural language | |
| RU2679988C1 (en) | Extracting information objects with the help of a classifier combination | |
| US20170161255A1 (en) | Extracting entities from natural language texts | |
| CN107463548B (en) | Phrase mining method and device | |
| US20190392035A1 (en) | Information object extraction using combination of classifiers analyzing local and non-local features | |
| RU2618374C1 (en) | Identifying collocations in the texts in natural language | |
| US20180113856A1 (en) | Producing training sets for machine learning methods by performing deep semantic analysis of natural language texts | |
| US11379656B2 (en) | System and method of automatic template generation | |
| CN110555440B (en) | Event extraction method and device | |
| CN107077640B (en) | System and process for analyzing, qualifying, and ingesting unstructured data sources via empirical attribution | |
| CN111125360A (en) | Emotion analysis method and device in game field and model training method and device thereof | |
| CN111914566A (en) | Automatic comment generation method | |
| CN113961823B (en) | News recommendation method, system, storage medium and equipment | |
| CN117216214A (en) | Question and answer extraction generation method, device, equipment and medium | |
| CN116151220A (en) | Word segmentation model training method, word segmentation processing method and device | |
| CN114970543A (en) | Semantic analysis method for crowdsourced design resources | |
| CN112446206A (en) | Menu title generation method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |