CN112597765A - Automatic movie and television topic generation method based on multi-mode features - Google Patents
Automatic movie and television topic generation method based on multi-mode features Download PDFInfo
- Publication number
- CN112597765A CN112597765A CN202011565739.9A CN202011565739A CN112597765A CN 112597765 A CN112597765 A CN 112597765A CN 202011565739 A CN202011565739 A CN 202011565739A CN 112597765 A CN112597765 A CN 112597765A
- Authority
- CN
- China
- Prior art keywords
- movie
- television
- text
- word
- vector
- 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
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/284—Lexical analysis, e.g. tokenisation or collocates
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Biomedical Technology (AREA)
- Databases & Information Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to the field of movie and television topics, in particular to an automatic movie and television topic generation method based on multi-mode features.
Description
Technical Field
The invention relates to the field of film and television topics, in particular to a film and television topic automatic generation method based on multi-mode characteristics.
Background
The method for automatically generating the film and television topics generally adopts a traditional machine learning algorithm, and unsupervised clustering is performed by utilizing the text characteristics of the film and television, such as directors, actors, subject matters, ages, regions and the like to generate the film and television topics.
Disclosure of Invention
The technical problems solved by the invention are as follows: the method for automatically generating the film and television special topic based on the multi-mode features solves the problems of non-uniform visual style of film and television posters in the film and television special topic and low film and television association degree.
The invention adopts the technical scheme for solving the technical problems that: the method for automatically generating the film and television special topic based on the multi-modal characteristics comprises the following steps:
s01, preprocessing poster pictures of the movie and television into a uniform size, and extracting graphic feature vectors of the movie and television posters by utilizing the representation learning capacity of a convolutional neural network;
s02, performing word segmentation and extraction on the text information of the movie to construct a keyword lexicon, calculating the inverse document frequency IDF of each keyword, performing one-hot coding on the keywords in the movie information to generate a text vector of the movie, and weighting the text vector by tf-IDF to obtain a weighted text feature vector;
s03, integrating user behaviors into segmented corpus sets, integrating corpus sets of all users to generate word2vec word vector model training samples, establishing a word vector model, training the word vector model by using a skip-gram, and generating a word feature vector of each movie;
s04, splicing the graphic feature vectors, the text feature vectors and the word feature vectors to generate movie and television multi-mode vectors, and performing cluster analysis by using an unsupervised clustering algorithm;
and S05, acquiring the film and television special topic with the multi-modal characteristics.
Further, in step S02, the text information of the movie includes a name, a brief introduction, a subject, a comment, a year, a region, a director, and an actor.
Further, in step S03, the user behavior includes a continuous movie browsing behavior, a movie searching behavior, and a watching behavior.
Further, in step S04, the unsupervised clustering algorithm is a K-Means algorithm.
The invention has the beneficial effects that: the method for automatically generating the film and television special topics based on the multi-mode features utilizes the convolutional neural network to learn the film and television posters, achieves the purpose of unifying visual styles, combines the graphic feature vectors, the text feature vectors and the word feature vectors to generate the film and television special topics through the unsupervised clustering algorithm, improves the association degree of the film and television special topics, and enhances the experience of a user when browsing the content of the film and television special topics.
Drawings
FIG. 1 is a flow chart of the method for automatically generating film and television titles based on multi-modal features.
FIG. 2 is a neural parameter diagram of a convolutional neural network of the automatic movie topic generation method based on multi-modal features.
Detailed Description
The invention provides a method for automatically generating film and television topics based on multi-modal characteristics, which utilizes a convolutional neural network to learn film and television posters to achieve the purpose of unifying visual styles, combines a graphic feature vector, a text feature vector and a word feature vector to generate the film and television topics through an unsupervised clustering algorithm to improve the association degree of the film and television topics, and comprises the following steps:
s01, preprocessing poster pictures of the movie and television into a uniform size, and extracting graphic feature vectors of the movie and television posters by utilizing the representation learning capacity of a convolutional neural network;
s02, performing word segmentation and extraction on the text information of the movie to construct a keyword lexicon, calculating the inverse document frequency IDF of each keyword, performing one-hot coding on the keywords in the movie information to generate a text vector of the movie, and weighting the text vector by tf-IDF to obtain a weighted text feature vector;
s03, integrating user behaviors into segmented corpus sets, integrating corpus sets of all users to generate word2vec word vector model training samples, establishing a word vector model, training the word vector model by using a skip-gram, and generating a word feature vector of each movie;
s04, splicing the graphic feature vectors, the text feature vectors and the word feature vectors to generate movie and television multi-mode vectors, and performing cluster analysis by using an unsupervised clustering algorithm;
and S05, acquiring the film and television special topic with the multi-modal characteristics.
Further, in step S02, the text information of the movie includes a name, a brief introduction, a subject, a comment, a year, a region, a director, and an actor.
Further, in step S03, the user behavior includes a continuous movie browsing behavior, a movie searching behavior, and a watching behavior.
Further, in step S04, the unsupervised clustering algorithm is a K-Means algorithm.
In particular, as shown in figure 1.
Movie graphics characteristic vector: the method comprises the steps of firstly downloading an original movie poster from an open source website, preprocessing the movie poster into a square size with the size of 224 x 224 to obtain a standard poster, then importing the standard poster into a convolutional neural network, wherein the convolutional neural network comprises 13 convolutional layers and 3 fully-connected layers, and convolutional layer network neuron parameters are shown in figure 2 to obtain a movie graphic characteristic vector.
Movie text characteristic vector: the movie text information comprises a name, a brief introduction, a subject, a comment, a year, a region, a director, an actor and the like, the name, the brief introduction and the comment text information are subjected to word segmentation to extract keywords, a complete movie text keyword dictionary base is generated by combining the subject, the region, the director and the actor, the anti-document frequency idf of each word in the dictionary base is calculated, then an 0/1 vector of the movie is generated by using one-hot coding through the keywords in the movie text information, and a weighted text characteristic vector is generated by combining tf-idf weights of the keywords in the movie text information to obtain a movie text characteristic vector.
Movie word feature vector: the user behaviors comprise user browsing, user searching, user watching and the like, firstly, the behaviors of a user in a continuous conversation period are constructed into a continuous array, movie objects of each behavior are represented by unique IDs of the behaviors, a section of text corpora separated by blank spaces is generated, users with the IDs less than 10 in the corpora are filtered, the text corpora of the user behaviors meeting conditions are integrated to generate word2vec word vector model training samples, a word vector model is established, and word vector models are trained by utilizing skip-grams to generate word feature vectors of each movie.
And splicing the image feature vectors, the text feature vectors and the word feature vectors to generate multi-mode movie vectors, and performing clustering analysis by using a K-Means algorithm to obtain movie titles with multi-mode features.
Claims (4)
1. The method for automatically generating the film and television special topic based on the multi-modal characteristics is characterized by comprising the following steps of:
s01, preprocessing poster pictures of the movie and television into a uniform size, and extracting graphic feature vectors of the movie and television posters by utilizing the representation learning capacity of a convolutional neural network;
s02, performing word segmentation and extraction on the text information of the movie to construct a keyword lexicon, calculating the inverse document frequency IDF of each keyword, performing one-hot coding on the keywords in the movie information to generate a text vector of the movie, and weighting the text vector by tf-IDF to obtain a weighted text feature vector;
s03, integrating user behaviors into segmented corpus sets, integrating corpus sets of all users to generate word2vec word vector model training samples, establishing a word vector model, training the word vector model by using a skip-gram, and generating a word feature vector of each movie;
s04, splicing the graphic feature vectors, the text feature vectors and the word feature vectors to generate movie and television multi-mode vectors, and performing cluster analysis by using an unsupervised clustering algorithm;
and S05, acquiring the film and television special topic with the multi-modal characteristics.
2. The method of claim 1, wherein in step S02, the text information of the movie comprises name, introduction, subject, comment, year, region, director and actor.
3. The method for automatically generating a movie theme based on multi-modal features according to claim 1, wherein in step S03, the user behavior comprises a continuous movie browsing behavior, a movie searching behavior and a watching behavior.
4. The method for automatically generating a movie theme based on multi-modal features according to claim 1, wherein in step S04, the unsupervised clustering algorithm is a K-Means algorithm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011565739.9A CN112597765A (en) | 2020-12-25 | 2020-12-25 | Automatic movie and television topic generation method based on multi-mode features |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011565739.9A CN112597765A (en) | 2020-12-25 | 2020-12-25 | Automatic movie and television topic generation method based on multi-mode features |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112597765A true CN112597765A (en) | 2021-04-02 |
Family
ID=75202234
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011565739.9A Pending CN112597765A (en) | 2020-12-25 | 2020-12-25 | Automatic movie and television topic generation method based on multi-mode features |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112597765A (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104123363A (en) * | 2014-07-21 | 2014-10-29 | 北京奇虎科技有限公司 | Method and device for extracting main image of webpage |
| US20170300782A1 (en) * | 2016-04-18 | 2017-10-19 | International Business Machines Corporation | Methods and systems of personalized photo albums based on social media data |
| CN109062995A (en) * | 2018-07-05 | 2018-12-21 | 北京工业大学 | A kind of social activity plan opens up the personalized recommendation algorithm of drawing board (Board) cover on network |
| CN110321473A (en) * | 2019-05-21 | 2019-10-11 | 山东省计算中心（国家超级计算济南中心） | Diversity preference information method for pushing, system, medium and equipment based on multi-modal attention |
| CN110401873A (en) * | 2019-06-17 | 2019-11-01 | 北京奇艺世纪科技有限公司 | Video clipping method, device, electronic equipment and computer-readable medium |
| US20200089802A1 (en) * | 2018-09-13 | 2020-03-19 | Microsoft Technology Licensing, Llc | Inferring topics with entity linking and ontological data |
| CN111698573A (en) * | 2020-06-24 | 2020-09-22 | 四川长虹电器股份有限公司 | Movie and television special topic creating method and device |
| US10803318B1 (en) * | 2016-05-18 | 2020-10-13 | Educational Testing Service | Automated scoring of video clips using extracted physiological features |
-
2020
- 2020-12-25 CN CN202011565739.9A patent/CN112597765A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104123363A (en) * | 2014-07-21 | 2014-10-29 | 北京奇虎科技有限公司 | Method and device for extracting main image of webpage |
| US20170300782A1 (en) * | 2016-04-18 | 2017-10-19 | International Business Machines Corporation | Methods and systems of personalized photo albums based on social media data |
| US10803318B1 (en) * | 2016-05-18 | 2020-10-13 | Educational Testing Service | Automated scoring of video clips using extracted physiological features |
| CN109062995A (en) * | 2018-07-05 | 2018-12-21 | 北京工业大学 | A kind of social activity plan opens up the personalized recommendation algorithm of drawing board (Board) cover on network |
| US20200089802A1 (en) * | 2018-09-13 | 2020-03-19 | Microsoft Technology Licensing, Llc | Inferring topics with entity linking and ontological data |
| CN110321473A (en) * | 2019-05-21 | 2019-10-11 | 山东省计算中心（国家超级计算济南中心） | Diversity preference information method for pushing, system, medium and equipment based on multi-modal attention |
| CN110401873A (en) * | 2019-06-17 | 2019-11-01 | 北京奇艺世纪科技有限公司 | Video clipping method, device, electronic equipment and computer-readable medium |
| CN111698573A (en) * | 2020-06-24 | 2020-09-22 | 四川长虹电器股份有限公司 | Movie and television special topic creating method and device |
Non-Patent Citations (1)
| Title |
|---|
| 吴兴宇: "基于深度学习的个性化电影推荐方法研究", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Visual to text: Survey of image and video captioning | |
| Torabi et al. | Learning language-visual embedding for movie understanding with natural-language | |
| CN109508400B (en) | Method for generating image-text abstract | |
| US12001474B2 (en) | Information determining method and apparatus, computer device, and storage medium | |
| CN112104919B (en) | Content title generation method, device, equipment and computer readable storage medium based on neural network | |
| CN109697239B (en) | Method for generating teletext information | |
| Li et al. | Residual attention-based LSTM for video captioning | |
| CN112163122A (en) | Method and device for determining label of target video, computing equipment and storage medium | |
| CN111258995B (en) | Data processing method, device, storage medium and equipment | |
| CN111488931A (en) | Article quality evaluation method, article recommendation method and corresponding devices | |
| EP4310695A1 (en) | Data processing method and apparatus, computer device, and storage medium | |
| CN111985243B (en) | Emotion model training method, emotion analysis device and storage medium | |
| CN114461804B (en) | Text classification method, classifier and system based on key information and dynamic routing | |
| CN112836702B (en) | Text recognition method based on multi-scale feature extraction | |
| CN115964560A (en) | Information recommendation method and equipment based on multi-mode pre-training model | |
| CN111680190A (en) | Video thumbnail recommendation method fusing visual semantic information | |
| CN112528989B (en) | Description generation method for semantic fine granularity of image | |
| CN117746143A (en) | AIGC-based image description text generation method and device and storage medium | |
| CN113486143A (en) | User portrait generation method based on multi-level text representation and model fusion | |
| CN113407766A (en) | Visual animation display method and related equipment | |
| CN115984842A (en) | Multi-mode-based video open tag extraction method | |
| CN113297485B (en) | Method for generating cross-modal representation vector and cross-modal recommendation method | |
| CN112597765A (en) | Automatic movie and television topic generation method based on multi-mode features | |
| CN112749553B (en) | Text information processing method and device for video file and server | |
| CN114547435A (en) | Content quality identification method, device, equipment and readable storage medium |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210402 |