CN107977472A - The method that house property class news article automatically generates - Google Patents
The method that house property class news article automatically generates Download PDFInfo
- Publication number
- CN107977472A CN107977472A CN201711443090.1A CN201711443090A CN107977472A CN 107977472 A CN107977472 A CN 107977472A CN 201711443090 A CN201711443090 A CN 201711443090A CN 107977472 A CN107977472 A CN 107977472A
- Authority
- CN
- China
- Prior art keywords
- character
- data set
- article
- automatically generates
- house property
- 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.)
- Granted
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
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/951—Indexing; Web crawling techniques
-
- 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
- G06F16/95—Retrieval from the web
- G06F16/958—Organisation or management of web site content, e.g. publishing, maintaining pages or automatic linking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/10—Text processing
- G06F40/166—Editing, e.g. inserting or deleting
- G06F40/186—Templates
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention relates to a kind of method that house property class news article automatically generates, comprise the following steps：Step（1）：Data set is prepared, collection content includes article title, and summary and text three parts, three parts form sentence after collecting；Step（2）：To step（1）In data set carry out modelling and training；Step（3）：Based on step（2）After the completion of model training, text generation is carried out.It is an advantage of the invention that：Manual control writing is not required, saves manpower；Efficient work, cooperation efficiency are more times manually write；Generation is stablized, and the usual theme of context can control, and avoids the occurrence of yellow anti-sudden and violent probably content.
Description
Technical field
The present invention relates to a kind of method that house property class news article automatically generates.
Background technology
News article is issued, and can not only increase brand exposure, but also can bring flow, especially search engine long-tail flow
Collection.But directly reprint, it is original not good enough, influence SEO（Search engine optimization）, but it is completely manually original, open
Sell larger, therefore we utilize artificial intelligence technology, the news article that trained a house property field automatically generates robot.And
Content hair portion system of the Successful utilization to us.
Existing news category writing has two kinds mostly, and one kind is to reprint news, using crawler technology, is climbed with speed as soon as possible
The original article in other sources is removed, is published to the website of oneself, another kind is artificial original writing, is created by manpower, automatic
Generate article field, mainly some it is simple upset splicing, the content of generation is to people and unfriendly, merely for search engine
Retrieval drainage uses.
The mode of existing these articles issue, there is larger shortcoming respectively, and consumption manpower is compared in manually creation, and reptile is reprinted,
It is original bad, simply upset splicing, it is unfriendly to people.
The content of the invention
The defects of to overcome the prior art, the present invention provide a kind of method that house property class news article automatically generates, this hair
Bright technical solution is：
The method that house property class news article automatically generates, comprises the following steps：
Step（1）：Data set is prepared, collection content includes article title, and summary and text three parts, three parts collect
After form sentence；
Step（2）：To step（1）In data set carry out modelling and training；
Step（3）：Based on step（2）After the completion of model training, text generation is carried out.
The step（1）Specifically include：Word frequency is counted, Arabic numerals handle and bebinning character and end
Only character splices；The word frequency statistics are specially to carry out word frequency statistics to the language material of whole data set, occurring in data set
Word of the frequency less than 10 times is replaced with setting character；The Arabic numerals processing is specially the appearance concentrated to data
The digit of Arabic numerals together is marked；The bebinning character and final character splicing are specially in data set
The foremost addition mark character representation article of every article start, in every article ending addition mark character representation below
Article terminates.
The step（2）Specific method be：Build coder-decoder model, and to the encoder of the model and
Decoder is separately trained：
Encoder：Being used as search condition by the whole section of character generated to bebinning character or current time goes whole training sets to examine
Rope, obtains some related text collection, some before being taken according to Information-retrieval Algorithm；Some information of taking-up are carried out
Coding, coding result is as the sentence for being adapted to generation；
Decoder：The sentence being currently generated is encoded；
Attention mechanism comprises the following steps that：Encoder output matrix is multiplied with decoder output matrix, obtains one group of power
Weight values vector, weighted value vector are multiplied with encoder output matrix again, the information for the next character predicted, by one layer
Softmax activation primitives, obtain the term vector of the final next character of prediction.
The step（3）Specific method be：
（3-1）Input starting words and expressions randomly selects banner word；
（3-2）According to starting words and expressions, whole data set is retrieved, some input information as encoder before taking；
（3-3）The result obtained to retrieval encodes；
（3-4）Present generated statement is encoded；
（3-5）Character late is obtained by attention mechanism；
（3-6）One sentence is generated untill there is end mark by beam search methods.
It is an advantage of the invention that：Manual control writing is not required, saves manpower；Efficient work, cooperation efficiency are manually to write
More times made；Generation is stablized, and the usual theme of context can control, and avoids the occurrence of yellow anti-sudden and violent probably content.
Embodiment
The invention will now be further described with reference to specific embodiments, the advantages and features of the present invention will be with description and
It is apparent.But these embodiments are only exemplary, do not form any restrictions to the scope of the present invention.People in the art
Member it should be understood that without departing from the spirit and scope of the invention can to the details of technical solution of the present invention and form into
Row modifications or substitutions, but these modifications and replacement are each fallen within protection scope of the present invention.
The present invention relates to a kind of method that house property class news article automatically generates, comprise the following steps：
Step（1）：Data set is prepared, collection content includes article title, and summary and text three parts, three parts collect
After form sentence；
Step（2）：To step（1）In data set carry out modelling and training；
Step（3）：Based on step（2）After the completion of model training, text generation is carried out.
The step（1）Specifically include：Word frequency is counted, Arabic numerals handle and bebinning character and end
Only character splices；The word frequency statistics are specially to carry out word frequency statistics to the language material of whole data set, occurring in data set
Word of the frequency less than 10 times is replaced with setting character, such as can be substituted for word of the word frequency less than 10<unk>；It is described
Arabic numerals processing be specially that the digit of the Arabic numerals appeared together concentrated to data is marked, such as will
The Arabic numerals appeared together are substituted for<numN>, the N in angle brackets represents the digit of numeral, such as：1234 should replace
Change into<num4>, 342134 should be substituted for<num6>, the bebinning character and final character splicing are specially in data set
In the foremost addition mark character representation article of every article start, in every article ending addition mark character list below
Show that article terminates, with "<s>" character representation article starts, every article ending addition below "</s>" character representation article knot
Beam.
The step（2）Specific method be：Build coder-decoder model, and to the encoder of the model and
Decoder is separately trained：
Encoder：Being used as search condition by the whole section of character generated to bebinning character or current time goes whole training sets to examine
Rope, obtains some related text collection, some before being taken according to Information-retrieval Algorithm；Some information of taking-up are carried out
Coding, coding result is as the sentence for being adapted to generation；
Decoder：The sentence being currently generated is encoded；
Attention mechanism comprises the following steps that：Encoder output matrix is multiplied with decoder output matrix, obtains one group of power
Weight values vector, weighted value vector are multiplied with encoder output matrix again, the information for the next character predicted, then by one
Layer softmax activation primitives, obtain the term vector of the final next character of prediction.
The step（3）Specific method be：
（3-1）Input starting words and expressions randomly selects banner word；
（3-2）According to starting words and expressions, whole data set is retrieved, some input information as encoder before taking；
（3-3）The result obtained to retrieval encodes；
（3-4）Present generated statement is encoded；
（3-5）Character late is obtained by attention mechanism；
（3-6）One sentence is generated untill there is end mark by beam search methods.
Claims (4)
1. the method that house property class news article automatically generates, it is characterised in that comprise the following steps：
Step（1）：Data set is prepared, collection content includes article title, and summary and text three parts, three parts collect
After form sentence；
Step（2）：To step（1）In data set carry out modelling and training；
Step（3）：Based on step（2）After the completion of model training, text generation is carried out.
2. the method that house property class news article according to claim 1 automatically generates, it is characterised in that the step
（1）Specifically include：Word frequency is counted, Arabic numerals are handled and bebinning character and final character splicing；The word
Frequency statistics is specially to carry out word frequency statistics to the language material of whole data set, word of the frequency of occurrences in data set less than 10 times with setting
Determine character to be replaced；The Arabic numerals processing is specially the Arabic numerals appeared together concentrated to data
Digit is marked；The bebinning character and final character splicing are specially that the foremost of every article in data set adds
Mark-on is known character representation article and is started, and in every article ending, addition mark character representation article terminates below.
3. the method that house property class news article according to claim 1 automatically generates, it is characterised in that the step
（2）Specific method be：Coder-decoder model is built, and the encoder and decoder of the model are separately trained：
Encoder：Being used as search condition by the whole section of character generated to bebinning character or current time goes whole training sets to examine
Rope, obtains some related text collection, some before being taken according to Information-retrieval Algorithm；Some information of taking-up are carried out
Coding, coding result is as the sentence for being adapted to generation；
Decoder：The sentence being currently generated is encoded；
Attention mechanism comprises the following steps that：Encoder output matrix is multiplied with decoder output matrix, obtains one group of power
Weight values vector, weighted value vector are multiplied with encoder output matrix again, the information for the next character predicted, by one layer
Softmax activation primitives, obtain the term vector of the final next character of prediction.
4. the method that house property class news article according to claim 1 automatically generates, it is characterised in that the step
（3）Specific method be：
（3-1）Input starting words and expressions randomly selects banner word；
（3-2）According to starting words and expressions, whole data set is retrieved, some input information as encoder before taking；
（3-3）The result obtained to retrieval encodes；
（3-4）Present generated statement is encoded；
（3-5）Character late is obtained by attention mechanism；
（3-6）One sentence is generated untill there is end mark by beam search methods.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711443090.1A CN107977472B (en) | 2017-12-27 | 2017-12-27 | Method for automatically generating house property news articles |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711443090.1A CN107977472B (en) | 2017-12-27 | 2017-12-27 | Method for automatically generating house property news articles |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107977472A true CN107977472A (en) | 2018-05-01 |
| CN107977472B CN107977472B (en) | 2021-11-05 |
Family
ID=62007995
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711443090.1A Active CN107977472B (en) | 2017-12-27 | 2017-12-27 | Method for automatically generating house property news articles |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107977472B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109492112A (en) * | 2018-10-24 | 2019-03-19 | 北京百科康讯科技有限公司 | A kind of method of the computer aided writing scientific popular article of knowledge based map |
| CN110555198A (en) * | 2018-05-31 | 2019-12-10 | 北京百度网讯科技有限公司 | method, apparatus, device and computer-readable storage medium for generating article |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1489088A (en) * | 2003-08-28 | 2004-04-14 | 北京英业科技开发有限公司 | Computer article storing and displaying method |
| US20120053937A1 (en) * | 2010-08-31 | 2012-03-01 | International Business Machines Corporation | Generalizing text content summary from speech content |
| CN105320642A (en) * | 2014-06-30 | 2016-02-10 | 中国科学院声学研究所 | Automatic abstract generation method based on concept semantic unit |
| CN105930314A (en) * | 2016-04-14 | 2016-09-07 | 清华大学 | Text summarization generation system and method based on coding-decoding deep neural networks |
| CN106126507A (en) * | 2016-06-22 | 2016-11-16 | 哈尔滨工业大学深圳研究生院 | A kind of based on character-coded degree of depth nerve interpretation method and system |
| CN106649223A (en) * | 2016-12-23 | 2017-05-10 | 北京文因互联科技有限公司 | Financial report automatic generation method based on natural language processing |
| CN106682387A (en) * | 2016-10-26 | 2017-05-17 | 百度国际科技（深圳）有限公司 | Method and device used for outputting information |
| US20170192956A1 (en) * | 2015-12-31 | 2017-07-06 | Google Inc. | Generating parse trees of text segments using neural networks |
| CN106980683A (en) * | 2017-03-30 | 2017-07-25 | 中国科学技术大学苏州研究院 | Blog text snippet generation method based on deep learning |
| CN107145482A (en) * | 2017-03-28 | 2017-09-08 | 百度在线网络技术（北京）有限公司 | Article generation method and device, equipment and computer-readable recording medium based on artificial intelligence |
| CN107193792A (en) * | 2017-05-18 | 2017-09-22 | 北京百度网讯科技有限公司 | The method and apparatus of generation article based on artificial intelligence |
| CN107391609A (en) * | 2017-07-01 | 2017-11-24 | 南京理工大学 | A kind of Image Description Methods of two-way multi-modal Recursive Networks |
-
2017
- 2017-12-27 CN CN201711443090.1A patent/CN107977472B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1489088A (en) * | 2003-08-28 | 2004-04-14 | 北京英业科技开发有限公司 | Computer article storing and displaying method |
| US20120053937A1 (en) * | 2010-08-31 | 2012-03-01 | International Business Machines Corporation | Generalizing text content summary from speech content |
| CN105320642A (en) * | 2014-06-30 | 2016-02-10 | 中国科学院声学研究所 | Automatic abstract generation method based on concept semantic unit |
| US20170192956A1 (en) * | 2015-12-31 | 2017-07-06 | Google Inc. | Generating parse trees of text segments using neural networks |
| CN105930314A (en) * | 2016-04-14 | 2016-09-07 | 清华大学 | Text summarization generation system and method based on coding-decoding deep neural networks |
| CN106126507A (en) * | 2016-06-22 | 2016-11-16 | 哈尔滨工业大学深圳研究生院 | A kind of based on character-coded degree of depth nerve interpretation method and system |
| CN106682387A (en) * | 2016-10-26 | 2017-05-17 | 百度国际科技（深圳）有限公司 | Method and device used for outputting information |
| CN106649223A (en) * | 2016-12-23 | 2017-05-10 | 北京文因互联科技有限公司 | Financial report automatic generation method based on natural language processing |
| CN107145482A (en) * | 2017-03-28 | 2017-09-08 | 百度在线网络技术（北京）有限公司 | Article generation method and device, equipment and computer-readable recording medium based on artificial intelligence |
| CN106980683A (en) * | 2017-03-30 | 2017-07-25 | 中国科学技术大学苏州研究院 | Blog text snippet generation method based on deep learning |
| CN107193792A (en) * | 2017-05-18 | 2017-09-22 | 北京百度网讯科技有限公司 | The method and apparatus of generation article based on artificial intelligence |
| CN107391609A (en) * | 2017-07-01 | 2017-11-24 | 南京理工大学 | A kind of Image Description Methods of two-way multi-modal Recursive Networks |
Non-Patent Citations (3)
| Title |
|---|
| LOPYREV,K.: "Generating News Headlines with Recurrent Neural Networks", 《ARXIV》 * |
| 李宝程: "基于浅层语义分析的文本摘要方法研究与实现", 《中国优秀硕士学位论文全文数据库(电子期刊)》 * |
| 李庆丰: "基于主题模型的多文档自动文摘方法研究", 《中国优秀硕士学位论文全文数据库(电子期刊)》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110555198A (en) * | 2018-05-31 | 2019-12-10 | 北京百度网讯科技有限公司 | method, apparatus, device and computer-readable storage medium for generating article |
| CN110555198B (en) * | 2018-05-31 | 2023-05-23 | 北京百度网讯科技有限公司 | Method, apparatus, device and computer readable storage medium for generating articles |
| CN109492112A (en) * | 2018-10-24 | 2019-03-19 | 北京百科康讯科技有限公司 | A kind of method of the computer aided writing scientific popular article of knowledge based map |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107977472B (en) | 2021-11-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Deconvolutional paragraph representation learning | |
| CN107908671A (en) | Knowledge mapping construction method and system based on law data | |
| CN108415977A (en) | One is read understanding method based on the production machine of deep neural network and intensified learning | |
| WO2021077974A1 (en) | Personalized dialogue content generating method | |
| CN108255805A (en) | The analysis of public opinion method and device, storage medium, electronic equipment | |
| CN112463424B (en) | Graph-based end-to-end program repairing method | |
| CN109189862A (en) | A kind of construction of knowledge base method towards scientific and technological information analysis | |
| CN108549658A (en) | A kind of deep learning video answering method and system based on the upper attention mechanism of syntactic analysis tree | |
| AU2020244577A1 (en) | Slot filling with contextual information | |
| CN112084841A (en) | Cross-modal image multi-style subtitle generation method and system | |
| CN106557298A (en) | Background towards intelligent robot matches somebody with somebody sound outputting method and device | |
| CN104376842A (en) | Neural network language model training method and device and voice recognition method | |
| CN107977472A (en) | The method that house property class news article automatically generates | |
| CN115906815B (en) | Error correction method and device for modifying one or more types of error sentences | |
| CN111198966B (en) | Natural language video clip retrieval method based on multi-agent boundary perception network | |
| CN114118065A (en) | Chinese text error correction method and device in electric power field, storage medium and computing equipment | |
| CN114492407A (en) | News comment generation method, system, equipment and storage medium | |
| CN108549636A (en) | A kind of race written broadcasting live critical sentence abstracting method | |
| CN114490991A (en) | Dialog structure perception dialog method and system based on fine-grained local information enhancement | |
| Yuan et al. | Controllable video captioning with an exemplar sentence | |
| CN114238652A (en) | Industrial fault knowledge map establishing method for end-to-end scene | |
| CN115983274A (en) | Noise event extraction method based on two-stage label correction | |
| CN116737922A (en) | Tourist online comment fine granularity emotion analysis method and system | |
| CN114691858A (en) | Improved UNILM abstract generation method | |
| Tian et al. | BEBERT: Efficient and robust binary ensemble BERT |
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 | ||
| CB03 | Change of inventor or designer information | ||
| CB03 | Change of inventor or designer information |
Inventor after: Li Zuochao Inventor after: Bai Junfeng Inventor after: Zhang Wenzhan Inventor after: Liu Ziyao Inventor after: Su Weijie Inventor before: Bai Junfeng Inventor before: Zhang Wenzhan Inventor before: Liu Ziyao Inventor before: Su Weijie |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP03 | Change of name, title or address | ||
| CP03 | Change of name, title or address |
Address after: Floor 20, Building 6, Smart Valley Park, the Taihu Lake Software Industrial Park, No. 1421, Wuzhong Avenue, Yuexi Street, Suzhou Economic Development Zone, Jiangsu Province, 215000 Patentee after: Zhuge Qihang (Suzhou) Technology Co.,Ltd. Address before: No. 506, Xingang center, No. 16, Jiuxianqiao Road, Jiangtai Township, Chaoyang District, Beijing 100015 Patentee before: BEIJING ZHUGE ZHAOFANG INFORMATION TECHNOLOGY Co.,Ltd. |