WO2008067749A1 - Système et procédé de gestion de contenu média - Google Patents

Système et procédé de gestion de contenu média Download PDF

Info

Publication number
WO2008067749A1
WO2008067749A1 PCT/CN2007/071133 CN2007071133W WO2008067749A1 WO 2008067749 A1 WO2008067749 A1 WO 2008067749A1 CN 2007071133 W CN2007071133 W CN 2007071133W WO 2008067749 A1 WO2008067749 A1 WO 2008067749A1
Authority
WO
WIPO (PCT)
Prior art keywords
content
media
information
subtitle
segment
Prior art date
Application number
PCT/CN2007/071133
Other languages
English (en)
French (fr)
Inventor
Fangshan Wang
Qi Fang
Yinyan Tan
Jieping Zhong
Original Assignee
Huawei Technologies Co., Ltd.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Huawei Technologies Co., Ltd. filed Critical Huawei Technologies Co., Ltd.
Publication of WO2008067749A1 publication Critical patent/WO2008067749A1/zh
Priority to US12/479,066 priority Critical patent/US8200597B2/en

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/60Information retrieval; Database structures therefor; File system structures therefor of audio data
    • G06F16/68Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
    • G06F16/683Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
    • G06F16/685Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using automatically derived transcript of audio data, e.g. lyrics
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/70Information retrieval; Database structures therefor; File system structures therefor of video data
    • G06F16/78Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
    • G06F16/783Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
    • G06F16/7844Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using original textual content or text extracted from visual content or transcript of audio data

Definitions

  • the present invention relates to the field of communications technologies, and in particular, to a media content management system and method.
  • Multimedia information is the main way for humans to perceive nature and to understand society. With the development of the Internet and the popularity of computer applications, multimedia information on the Internet has exploded, which has brought new problems in the management and concentration of information.
  • multimedia information such as video and music is different from ordinary text files, especially in the management of media content information.
  • multimedia content such as news and sports events
  • domain ontology domain ontology
  • Embodiments of the present invention provide a media content management system and method.
  • An embodiment of the present invention provides a media content management system, including:
  • a text classifier configured to classify the subtitle information according to the defined theme content, and obtain a plurality of content segments having different themes
  • a media content annotation processing unit configured to mark specific play time information of each content segment having different themes after being classified by the text classifier, to obtain a plurality of content segments having different themes with specific inter-frame information, and A plurality of content segments having different themes with specific information are matched with concepts in the ontology library, and the content segments are labeled with words defined in the ontology library.
  • An embodiment of the present invention provides a media content management method, including:
  • Inter-segment segmentation is performed to obtain a plurality of content segments having different themes; and the specific playing time information of each of the content segments in the media is marked according to the inter-day information of the media content inter-segment segments, and the specific inter-day information is obtained.
  • a plurality of content segments having different themes according to the theme and the ontology library of the plurality of content segments having different themes with specific information The concept is matched, and the content segment is labeled with a vocabulary defined in the ontology library.
  • the embodiment of the present invention analyzes the subtitle file corresponding to the media, and divides the media content into different content segments according to the time, and each content segment is The content is associated with the ontology concept and records where the content segment appears in the media. This is to describe the content of the media using standard vocabulary, which is conducive to the unification of the content description information, making it possible to retrieve the content of the media.
  • semantically related retrieval of media content can be provided.
  • the user wants to retrieve the content of a certain aspect of his or her interest, and uses the vocabulary to describe and mark it.
  • semantic search can be used to perform the association search. For example, when a news or other multimedia clip is marked as "basketball", in the ontology concept, the relationship "basketball” is a subclass of "sports", and an inheritance relationship is inferred, so that when the user searches for " Sports "related program content, will also find out the segment or the entire media corresponding to the segment. This has enriched the scope of media content queries to a certain extent. Recording the location of the clip in the media allows the user to easily locate the content of their interest.
  • related media content editing can be performed using the methods and systems provided by the present invention.
  • a user wants to find content related to a terrorist attack that he is interested in in a large amount of multimedia content
  • using the results of the system provided by the present invention it is easy to write an application, allowing the computer to find related topics according to ontology reasoning, and then According to the theme, from a large amount of media content, the corresponding content is edited according to the start and end, so that only the content of interest is found out. This greatly facilitates the manual processing workload to a certain extent.
  • FIG. 1 is a schematic structural view of an embodiment of the system according to the present invention.
  • FIG. 3 is a flow chart of an embodiment of content segmentation and positioning according to the method of the present invention.
  • FIG. 4 is a schematic diagram of an embodiment of subtitle content classification
  • FIG. 5 is a schematic diagram of an embodiment of media content associated with an ontology library.
  • An embodiment of the present invention provides a media content management system, and a schematic structural diagram of an embodiment of the system is as follows.
  • the system includes: an ontology library, a media library, and a media subtitle library, a text classifier, a media content annotation processing unit, a media content registration information library, and the like attached to the media library.
  • Ontology Libraries The ontology library defines several concepts, including several terms, and the relationships between words. These words are descriptions of specific transactions, each with a unique resource identifier.
  • the role of the ontology library is to acquire and store knowledge in the relevant fields of the media, to provide a common understanding of the knowledge in the field, to identify commonly recognized vocabulary in the field, and to give these terms (terms) from different levels of formalization patterns and A clear definition of the relationship between vocabulary.
  • the languages or standards currently describing ontology include: OWL (Web Ontology Language, ⁇ ), KIF (Knowledge Interchange)
  • the media library stores specific media content, such as video content, audio content, and so on. Each specific media has a unique identifier.
  • the media identifier may be a file name of the media without a duplicate name, such as "2006-9-27 news broadcast.wmv", or an index specially assigned to the media, such as "451235 8", etc., and possibly other Any number or text, alphabetic symbol sequence, URL (Uniform Resource Locator), or URI (Uniform Resource Identifier) that uniquely identifies the media.
  • the subtitle file is divided into an embedded subtitle file and an external subtitle file.
  • the embedded subtitle file is directly integrated into the media file and cannot be modified and edited; and the external subtitle file needs an additional independent file, which records the daytime Subtitles such as voices appearing one after another.
  • Video subtitle files include, but are not limited to, .txt, .srt, .subs .ssa .smi file formats.
  • the subtitle information includes, in addition to the subtitle text information, that is, the subtitle content information, the following: the dice code information of the subtitle occurrence, and the subtitle file Media identification information.
  • the suffix portion of the file name of the subtitle file is identical to the suffix portion of the media content file name, and accordingly, the correspondence between the two can be directly determined.
  • the media library can also be placed with the media subtitle library, or the media and media subtitle files can be placed together.
  • [39] Media Content Annotation Processing Unit It is used to mark the specific playing time information of each content segment with different themes obtained by the text classifier (see the function description of the text classifier below). a plurality of content segments having different themes of the information, and each piece of the content having different themes marked with a specific playing time is marked with a vocabulary defined in the ontology library, thereby performing content with the content in the ontology library Association.
  • the media content annotation processing unit includes three subunits:
  • the media content extracting unit comprises: obtaining a media identifier of the media content to be marked from the media library, obtaining corresponding subtitle information in the media caption library according to the media identifier, and identifying the subtitle information in the subtitle information
  • the subtitle content information is identified by the subtitle content information in the order of the subtitles, and a plurality of subtitle content interleaving segments, that is, media content interleaf segments are formed.
  • the main functions include labeling the specific playing time information of the content segments with different themes classified by the text classifier, that is, marking the beginning and ending points of each content segment. A plurality of content segments having different themes with specific inter-temporal information are obtained.
  • Labeling Adaptation Unit The main functions include matching the classification information differentiated by the content classification and positioning unit with the concept in the ontology library, and labeling the content segment with the vocabulary defined in the ontology library to generate the content annotation.
  • the information, the content annotation information includes but is not limited to: the identifier of the media to which the segment belongs, the ontology concept identifier corresponding to the content segment, and the start and end point description information of the segment.
  • Text classifier The subtitle content information used in the plurality of independent subtitle information acquired by the media content extraction unit is classified according to the defined theme content.
  • the text classifier generally has a plurality of subject words or topic sentences set in advance and logic and algorithms for determining which subject the text content belongs to.
  • the input is a plurality of independent text information
  • the output is a classification of the text information according to the theme, and the classification obtains a plurality of content segments having different themes.
  • Media Content Registration Information Library Used to record the content segments marked by the tag adaptation unit.
  • An embodiment of the present invention provides a media content management method, and an implementation process of the method is as shown in the following figure.
  • Step 1 Obtain the media identifier of the media content to be marked
  • the media content extraction unit obtains a media identifier of the media content to be labeled from the media library, and the media identifier may be a file name of the media file or specifically for the media
  • the identification information such as the cable bow I created by the document.
  • Step 2 Obtain a corresponding subtitle file according to the obtained media identifier
  • the subtitle file refers to a file that describes the text for each conversation or other voice or interpretation in the media.
  • a media identifier can uniquely correspond to a subtitle file.
  • the subtitle files are divided into embedded subtitle files and external subtitle files.
  • the embedded subtitle files are directly integrated into the media files and cannot be modified and edited; the external subtitle files need a separate independent file, which records Subtitles such as voices that appear one after another.
  • Subtitles file formats include but are not limited to: .txt, .srt, .sub ssa, .smi the like.
  • the subtitle files in the subtitles of these file formats include at least: subtitle content information and diurnal code information (beginning and ending) of the subtitles, and media identification information corresponding to the subtitle files.
  • the inter-day information appears in the subtitle in the format of the standard format of the time code, and the format is XX: XX: XX, and the three fields respectively represent small ⁇ , minute, and second.
  • Step 3 extracting the subtitle content in the subtitle file obtained by the foregoing, and identifying the subtitle content information according to the sequence of the subtitles, forming a plurality of interleaving segments of the media content, and classifying according to the defined theme content. Obtaining a plurality of content segments having different themes, and marking the specific playing time information of each content segment in the media according to the inter-day information of the media content segments, and obtaining a plurality of different themes having specific inter-temporal information Fragment of content;
  • FIG. 3 An embodiment implementation process is shown in FIG. 3, and includes the following steps:
  • Step 30 Read the contents of the subtitle file, record each logo with the content of the subtitles at the beginning and ending, and the daytime
  • the subtitle content information may be a subtitle sentence, and each has a time code
  • the caption statement records an identifier.
  • the content extraction result is shown in the following example: Identify subtitles
  • Each type of subtitle file has a fixed format, with mature subtitle content and format extraction tools, such as professional VOBSUB subtitle recognition software, which can extract subtitle information in multiple formats; and for text format such as .txt Subtitles, their inter-time and subtitle information are fixed formats, and regular expressions can be used to extract information that satisfies the conditions.
  • This extraction technique is prior art and will not be described in detail in the present invention.
  • Step 31 classify according to the defined theme in units of the identified caption sentences, and form a plurality of content segments including one or more representing different themes;
  • Frequency inverse text frequency
  • Bayesian algorithm Bayesian algorithm
  • Rocchio similarity calculation method
  • KNN K-nearest neighbor
  • Bayes simple Bayes
  • the various information classification methods described are capable of classifying different text contents input to the classifier according to different topics.
  • the subject matter includes a knowledge classification that is manually customized in advance or a keyword directory structure in which machine learning is performed in the process of classification.
  • the subtitle content classification according to the present invention is based on the content of the entire subtitle file, and each subtitle sentence that can be individually identified is a unit.
  • Figure 4 is a schematic diagram of a process of classification.
  • the classification process for the subtitle file uses one of the above existing techniques, and the input data can be generated by the media content extraction unit in the system of the present invention, and the receiving component of the output data can be a content classification and positioning unit.
  • each content segment contains one or more subtitle sentences, ie, topics, that can be independently distinguished in the media. Fragments and fragments may have cross-over or containment relationships between them and the subtitles. And these segments containing subtitles representing content of a certain subject content correspond to media segments of a certain segment of the media. For some news, sports commentary programs and other media, the content reflected by the subtitles is itself the content that the media can understand.
  • Step 32 Labeling the specific playing time information of each content segment in the media according to the inter-day information of the media content segment, and obtaining a plurality of content segments having different themes with specific information ;
  • each content segment contains one or more topics, each topic appears differently.
  • the content segment here is corresponding to a certain media segment in the media. It is necessary to mark the range of occurrences of the content segments based on the occurrence of each topic.
  • the labeling method comprises: segmenting a plurality of topics in a content segment in which the inter-turn interval exceeds the threshold according to an inter-turn threshold (which may be preset or determined according to a media utilization algorithm) Multiple pieces of content with the same theme. For example, when a certain content segment includes three caption sentences 1001, 1002, and 1003, and the inter-turn threshold is set to 3 minutes, the start of the caption 1003 differs from the other two captions by more than 3 minutes. The content segment is then divided into two different content segments of the same theme consisting of 1001, 1002, and 1003.
  • an inter-turn threshold which may be preset or determined according to a media utilization algorithm
  • the method of labeling the segments of each content segment played in the media includes: determining that the earliest statement in each content segment is the start statement and the end of the last statement is the end sentence, taking the content segment The beginning of the start statement is the beginning of the content clip at the beginning of the media play, and the end of the end statement is the end of the content clip.
  • Step 4 Match the content of the above classification processing with the ontology concept, and mark the words in the ontology Note the content fragment;
  • each segment has one or several topics (which may be keywords or sentences) representing its content.
  • topics which may be keywords or sentences
  • the adaptation refers to finding a concept in the ontology library that is close to or the same as the meaning of the topic.
  • a traditional word fuzzy matching algorithm may be used to find the closest concept to the matched topic word in the ontology, or to modify according to other topic words of the segment, and finally one or more ontology words may be matched to identify the word.
  • the subject matter of the clip may be used to find the closest concept to the matched topic word in the ontology, or to modify according to other topic words of the segment.
  • the traditional word fuzzy matching method can be used.
  • a specific method is as follows: The ontology concept is used as a common vocabulary, and the "like" function in the data query is used to find a word containing some or all of the vocabulary to be searched in the ontology. If you use the "like" matching method, you can find out that the concept of "terror” matching in the ontology is "terrorism"; and when there are multiple ontology concepts on the matching, you can match the word in the concept to the proportion of words in the concept. The method determines the degree of matching to determine the closest ontology concept.
  • Other ontology concept matching algorithms including the introduction of ontology reasoning and correlation matching algorithms, can provide more accurate and efficient matching methods.
  • mapping relationship For some given ontology libraries and a given domain knowledge classification knowledge base, there is a mapping relationship between the concepts themselves, which shows the vocabulary of a subject word or topic sentence and ontology concept. The mapping relationship is shown in the following table:
  • the execution process includes: first selecting a topic word of the content segment; and then searching for an ontology URI corresponding to the topic word in the mapping table as an ontology concept for labeling the content segment.
  • An embodiment in which a specific content segment is associated with an ontology library is shown in FIG. 5;
  • Step 5 Generate and store the annotation information based on the matching information described above.
  • the annotation information includes recording each of the classified content segments, including but not limited to : The media identifier to which the content segment belongs, the media concept resource identifier corresponding to the content segment, and the start and end of the content segment in the media.
  • the above-mentioned stored annotation information serves as the basis for managing the media content.
  • Step 6 Determine if there is any media to be marked
  • the above embodiment is a preferred embodiment of the present invention. After finding the theme of the content of the media segment, such as "tennis", it matches the ontology in the body, and finding the vocabulary of the ontology to mark the content of the segment can be omitted.
  • the embodiment of the present invention analyzes the subtitle file corresponding to the media, and divides the media content into different content segments according to the time, and associates the content in each segment with the ontology concept, and records the The location where the content clip appears in the media.
  • Step 1 Obtain the media identifier of the media content to be marked
  • the media content extraction unit obtains a media identifier of the media content to be labeled from the media library, where the media identifier may be a file name of the media file or specifically for the media
  • the identification information such as the cable bow I created by the document.
  • Step 2 Acquire a corresponding subtitle file according to the obtained media identifier
  • the subtitle file refers to a file that describes the text for each conversation or other voice or interpretation in the media.
  • a media identifier can uniquely correspond to a subtitle file.
  • the subtitle files are divided into embedded subtitle files and external subtitle files.
  • the embedded subtitle files are directly integrated into the media files and cannot be modified and edited; the external subtitle files need a separate independent file, which records Subtitles such as voices that appear one after another.
  • Subtitles file formats include but are not limited to: .txt, .srt, .sub ssa, .smi the like.
  • the subtitle files in the subtitles of these file formats include at least: subtitle content information and diurnal code information (beginning and ending) of the subtitles, and media identification information corresponding to the subtitle files.
  • the inter-day information is in a standard format in the subtitles.
  • the format of the inter-code appears in the format XX: XX: XX, and the three fields represent small ⁇ , minute, and second respectively.
  • Step 3 extracting the subtitle content in the subtitle file obtained by the foregoing, and identifying the subtitle content information according to the sequence of the subtitles, forming a plurality of intermediaries of the media content, and classifying according to the defined theme content. Obtaining a plurality of content segments having different themes, and marking the specific playing time information of each content segment in the media according to the inter-day information of the media content segments, and obtaining a plurality of different themes having specific inter-temporal information Fragment of content.
  • Step 4 ⁇ Match the content of the above classification processing with the ontology concept, and mark the content segment with the vocabulary in the ontology;
  • each segment has one or several topics (which may be keywords or sentences) representing its content.
  • topics which may be keywords or sentences
  • the adaptation refers to finding a concept in the ontology library that is close to or the same as the meaning of the topic.
  • a traditional word fuzzy matching algorithm may be used to find the closest concept to the matched topic word in the ontology, or to modify according to other topic words of the segment, and finally one or more ontology words may be matched to identify the word.
  • the subject matter of the clip may be used to find the closest concept to the matched topic word in the ontology, or to modify according to other topic words of the segment.
  • the traditional word fuzzy matching method can be utilized.
  • a specific method is as follows: The ontology concept is used as a common vocabulary, and the "like" function in the data query is used to find a word containing some or all of the vocabulary to be searched in the ontology. If you use the "like" matching method, you can find out that the concept of "terror” matching in the ontology is "terrorism"; and when there are multiple ontology concepts on the matching, you can match the word in the concept to the proportion of words in the concept. The method determines the degree of matching to determine the closest ontology concept.
  • Other ontology concept matching algorithms including the introduction of ontology reasoning and correlation matching algorithms, can provide more accurate and efficient matching methods.
  • mapping relationship For some given ontology libraries and a given domain knowledge classification knowledge base, there is a mapping relationship between the concepts themselves, which shows the vocabulary of a subject word or topic sentence and ontology concept. The mapping relationship is shown in the following table: [92]
  • the execution process includes: first selecting a topic word of the content segment; and then searching for an ontology URI corresponding to the topic word in the mapping table as an ontology concept for labeling the content segment.
  • An embodiment in which a specific content segment is associated with an ontology library is shown in FIG. 5;
  • Step 5 Generate and store the annotation information based on the matching information described above.
  • the annotation information includes recording each of the classified content segments, including but not limited to: the media identifier to which the content segment belongs, the media concept resource identifier corresponding to the content segment, and the start of the content segment in the media. Days and terminations, etc.
  • the above stored annotation information serves as the basis for managing the media content.
  • Step 6 Determine if there is any media to be labeled

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Library & Information Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • General Health & Medical Sciences (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Description

说明书 一种媒体内容管理系统及方法
[1] 技术领域
[2] 本发明涉及通信技术领域, 尤其涉及一种媒体内容管理系统及方法。
[3] 发明背景
[4] 多媒体信息是人类感知自然, 认识社会的主要途径。 伴随着互联网络的发展和 计算机应用的普及, 网络上的多媒体信息呈爆炸式增长, 这就在信息的管理和 釆集中带来了新的问题。
[5] 在各种多媒体信息中, 由于视频、 音乐等多媒体信息有别于一般的文本文件, 尤其体现在媒体内容信息管理方面。 对于一些新闻、 体育赛事等方面的媒体内 容, 由于不同的吋间播放的内容不完全一样, 因此, 需要将媒体中不同吋间段 的内容进行统一标注和管理。 以便在大量的媒体文件中检索需要的内容吋可以 方便快捷的找到相应的部分, 甚至可以直接利用电脑进行多个媒体内容的裁减
[6] 目前, 在解决上述问题方面大多吋候釆用人为的手工协助, 才能进行媒体内容 管理, 然而, 这是一个耗吋耗力而又效果不好的方法。
[7] 另外的一些方法, 通过利用一些文本信息对媒体内容进行描述, 从而对媒体内 容进行管理。 具体实现为: 定义一个本体库 (ontology) , 作为描述物与物之间 关联的概念架构, 其可由计算机所分享与理解。
[8] 然而, 一个构建完善的本体库通常可被搜寻引擎、 知识管理、 电子商务等应用 软件所运用, 用以增加搜寻的效率, 或增进文件处理能力。 由于在一些领域内
, 其词汇是有限的, 并且通常情况下有一些大家公认的词汇, 相对而言, 建立 本体库的难度要小一些, 因此, 目前基于本体库的应用主要是一些领域本体 (d omain
ontology) 的应用, 如在基因领域以及一些大的企业的内部信息管理等方面。 因 而在媒体内容管理的领域内如何定义一个完善的本体库是一个亟待解决的问题 , 目前还没有一个完善的本体库。 另外, 如何将本体库与媒体内容进行关联也是建立媒体库吋需要考虑的因素。 目前对于媒体内容的识别和记录, 大多数釆用图像识别或是人工标识的方法。 前者一般应用在专业领域, 如在足球比赛中, 捕获射门或进球的视频; 而后者 一般通过人工对媒体进行识别进行手工描述, 然后对本体库进行关联。
由于目前图像识别的准确性不高, 处理速度慢, 因此, 利用图像识别的技术来 进行媒体内容标注应用很少, 效果不理想。
而手工描述和标注的方法主观性强, 而且工作量大, 在实际应用中效果也很差 因此, 基于目前的状况, 如何对媒体内容进行有效管理仍是一个亟待解决的问 题。
发明内容
本发明实施例提供一种媒体内容管理系统及方法。
本发明实施例是通过以下技术方案实现的:
本发明实施例提供一种媒体内容管理系统, 包括:
文本分类器, 用于对字幕信息按照定义好的主题内容进行分类, 得到多个具有 不同主题的内容片段;
媒体内容标注处理单元, 用于标注经文本分类器分类后具有不同主题的每一个 内容片段的具体播放吋间信息, 得到具有具体吋间信息的多个具有不同主题的 内容片段, 并将所述具有具体吋间信息的多个具有不同主题的内容片段与本体 库中的概念进行匹配, 并以本体库中定义的词汇来标注所述内容片段。
本发明实施例提供一种媒体内容管理方法, 包括:
根据待标注媒体内容的媒体标识获取对应的字幕信息;
提取上述获取的字幕信息中的字幕内容信息, 对所述字幕内容信息按照吋间先 后顺序进行标识, 形成多个媒体内容吋间片段, 并按照定义好的主题内容对所 述多个媒体内容吋间片段进行分类, 得到具有不同主题的多个内容片段; 根据所述媒体内容吋间片段的吋间信息标注所述每个内容片段在媒体中的具体 播放吋间信息, 得到具有具体吋间信息的多个具有不同主题的内容片段; 根据所述具有具体吋间信息的多个具有不同主题的内容片段的主题与本体库中 的概念进行匹配, 以本体库中定义的词汇来标注所述内容片段。
[24] 由上述本发明实施例提供的技术方案可以看出, 本发明实施例通过分析与媒体 对应的字幕文件, 对媒体内容按吋间区分成不同的内容片段, 并对每个内容片 段里面的内容与本体概念进行关联, 记录了内容片段在媒体中出现的位置。 这 样是将媒体的内容利用标准的词汇进行描述, 利于对内容描述信息的统一, 使 得对媒体的内容检索成为可能。
[25] 另外, 利用本发明实施例提供的方法和系统, 可以提供对媒体内容进行语义相 关检索。 在很多应用中, 用户希望检索自己感兴趣的某方面的内容, 而利用本 体词汇进行描述和标注, 在提供普通的内容检索的基础上, 利用语义推理, 可 以进行关联搜索。 例如, 当某个新闻或其他多媒体片段被标注成"篮球"吋, 在本 体概念中, 可以通过关系 "篮球 "是"体育"的子类, 进行一个继承关系的推理, 从 而, 当用户搜索"体育"相关节目内容吋, 也会把该片段或该片段对应的整个媒体 找出来。 这在一定程度上丰富了媒体内容査询的范围。 而记录了片段在媒体中 的位置, 可以让用户很方便的定位自己关注的内容。
[26] 在另外的一些场合, 可以利用本发明提供的方法和系统进行相关媒体内容剪辑 。 如, 当用户希望在大量的多媒体内容中找到他所关注的恐怖袭击相关的内容 吋, 利用本发明提供的系统的结果, 可以很容易的编写应用程序, 让电脑根据 本体推理找到相关的主题, 再根据主题从大量的媒体内容中, 根据起止吋间将 对应的内容进行剪辑, 从而只将关注的内容找出来。 这在一定程度上大大方便 了人工处理的工作量。
[27] 附图简要说明
[28] 图 1为本发明所述系统一种实施例结构示意图;
[29] 图 2为本发明所述方法一种实施例流程图;
[30] 图 3为本发明所述方法内容分段、 定位一种实施例流程图;
[31] 图 4为字幕内容分类一种实施例示意图;
[32] 图 5为媒体内容与本体库关联一种实施例示意图。
[33] 实施本发明的方式
[34] 本发明实施例提供一种媒体内容管理系统, 所述系统一种实施例结构示意图如 图 1所示, 本系统包括: 本体库、 媒体库、 及媒体库附带的媒体字幕库、 文本分 类器、 媒体内容标注处理单元、 媒体内容注册信息库等。 下面对各实体的功能 及各实体间的关联作详细介绍:
[35] 本体库: 本体库中定义了若干概念, 包括若干词汇, 以及词汇之间的关系。 这 些词汇是对具体事务的描述, 每一词汇都有唯一的资源标识。 建立本体库的作 用在于获取并存储媒体相关领域的知识, 提供对该领域知识的共同理解, 确定 该领域内共同认可的词汇, 并从不同层次的形式化模式上给出这些词汇 (术语 ) 和词汇之间相互关系的明确定义。 目前描述本体的语言或标准包括: OWL ( Web Ontology Language , ^ ^禾中本体语言) 、 KIF (Knowledge Interchange
Format, 一种数据交换标准)、 OCML (Operational Conceptual Modelling
Language , ^ ^禾中本体语言) 、 FLogic (Frame
Logic, 框架逻辑) 、 SHOE (Simple HTML Ontology
Extensions , 一种本体语言) 、 XOL (Ontology Exchange
Language , ^ ^禾中本体语言) 、 OIL (Ontology Inference Layer/Ontology
Interchange Language , 本体推理层 /本体交换语言) 、 DAML (DARPA Agent Markup Language , 一禾中本体语言) 以及 RDF (Resource Description
Framework, 资源描述框架) 及其 RDF Schema (RDF的扩展) 等。
[36] 媒体库: 媒体库中保存的是具体的媒体内容, 如视频内容、 音频内容等。 每个 具体的媒体有一个唯一的标识。 所述的媒体标识可以是媒体无重名的文件名, 如" 2006-9-27新闻联播 .wmv", 或者是为媒体专门分配的索引等标识, 如" 451235 8"等, 还可能是其他的任何能够唯一标识该媒体的数字或文字、 字母符号序列、 URL (统一资源定位器) 或 URI (统一资源标识) 。
[37] 媒体字幕库: 记录的是与媒体库中的媒体内容对应的对媒体附带的字幕信息。
目前, 字幕文件分嵌入式字幕文件和外挂字幕文件, 嵌入式字幕文件直接融入 在媒体文件中, 是不可修改和编辑的; 而外挂字幕文件需要一个另外的独立的 文件, 里面记录了按照吋间先后出现的话音等的字幕。 视频外挂字幕文件包括 但不限于 .txt、 .srt、 .subs .ssa .smi几种文件格式。 字幕信息中除了包括字幕文 本信息, 即字幕内容信息外, 还包括: 字幕出现的吋间码信息、 字幕文件所对 应的媒体标识信息。 一般, 字幕文件的文件名中不带后缀部分与媒体内容文件 名的不带后缀部分一致, 据此可以直接判断二者的对应关系。
[38] 所述媒体库也可以和媒体字幕库放在一起, 或者将媒体和媒体字幕文件放在一 起。
[39] 媒体内容标注处理单元: 用于标注经文本分类器 (具体功能见下面文本分类器 的功能介绍) 分类后得到的每个具有不同主题的内容片段的具体播放吋间信息 , 得到具有具体吋间信息的多个具有不同主题的内容片段, 并将所述标注了具 体播放吋间的具有不同主题的每个内容片段以本体库中定义的词汇进行标注, 从而与本体库中的内容进行关联。
[40] 所述媒体内容标注处理单元包括三个子单元:
[41] 媒体内容提取单元: 主要功能包括从媒体库中获取待标注的媒体内容的媒体标 识, 根据所述的媒体标识在媒体字幕库中获取对应的字幕信息, 并识别所述字 幕信息中的字幕内容信息, 对字幕内容信息按照吋间先后顺序进行标识, 形成 多个字幕内容吋间片段, 即媒体内容吋间片段。
[42] 内容分类定位单元: 主要功能包括标注经文本分类器进行分类后的具有不同主 题的内容片段的具体播放吋间信息, 即标注每一个内容片段的开始吋间点和结 束吋间点, 得到具有具体吋间信息的多个具有不同主题的内容片段。
[43] 标注适配单元: 主要功能包括将经内容分类定位单元按吋间区分的分类信息与 本体库中的概念进行匹配, 并以本体库中定义的词汇来标注该内容片段, 生成 内容标注信息, 内容标注信息包括但不限于: 该片段所属的媒体的标识、 内容 片段对应的本体概念标识、 片段的起止吋间点描述信息等。
[44] 本实施例所述媒体内容标注处理单元的特征可以应用于本发明的其他实施例中
[45] 文本分类器: 用于对媒体内容提取单元获取的若干独立的字幕信息中的字幕内 容信息按照定义好的主题内容进行分类。 文本分类器中一般有事先设定的若干 主题词或主题语句以及判断文本内容是属于哪个主题的逻辑及算法。 其输入是 多个独立的文本信息, 而输出是按照主题对这些文本信息的分类, 分类后得到 多个具有不同主题的内容片段。 [46] 媒体内容注册信息库: 用于记录经标注适配单元标注好的内容片段。
[47] 本发明实施例提供一种媒体内容管理方法, 所述方法一种实施例实现流程如图
2所示, 包括如下步骤:
[48] 步骤 1 : 获取待标注媒体内容的媒体标识;
[49] 对于给定的媒体库, 存放至少一个媒体文件, 媒体内容提取单元从媒体库中获 取待标注的媒体内容的媒体标识, 所述媒体标识可能是媒体文件的文件名或是 专门为媒体文件建立的索弓 I等标识信息。
[50] 步骤 2: 根据所述获取的媒体标识获取对应的字幕文件;
[51] 所述的字幕文件是指为每个媒体中的对话或者其他的话音、 解释进行文字描述 的文件。 一个媒体标识可以唯一对应一份字幕文件。
[52] 目前, 字幕文件分嵌入式字幕文件和外挂字幕文件, 嵌入式字幕文件直接融入 在媒体文件中, 是不可修改和编辑的; 而外挂字幕文件需要一个另外的独立的 文件, 里面记录了按照吋间先后出现的话音等的字幕。 外挂字幕文件的格式包 括但不限于: .txt、 .srt、 .sub .ssa、 .smi等。 这些文件格式的字幕里面的字幕文 件至少包括: 字幕内容信息及字幕出现的吋间码信息 (开始吋间、 结束吋间) 、 字幕文件所对应的媒体标识信息。 所述的吋间信息在字幕中以标准格式的吋 间码的格式出现, 其格式为 XX: XX: XX, 三个字段分别表示小吋、 分、 秒。
[53] 步骤 3: 提取上述获取的字幕文件中的字幕内容, 对所述字幕内容信息按照吋 间先后顺序进行标识, 形成多个媒体内容吋间片段, 并按照定义好的主题内容 进行分类, 得到具有不同主题的多个内容片段, 根据所述媒体内容吋间片段的 吋间信息标注每个内容片段在媒体中的具体播放吋间信息, 得到具有具体吋间 信息的多个具有不同个主题的内容片段;
[54] 一种实施例具体实现过程如图 3所示, 包括如下步骤:
[55] 步骤 30: 读取字幕文件内容, 记录每个标有起止吋间的字幕内容的标识及吋间
I口自te!、.,
[56] 对于给定的字幕文件进行识别, 提取字幕文件中所有出现的字幕内容信息以及 字幕内容对应的吋间码信息, 所述的字幕内容信息可以为字幕语句, 对于每个 有吋间码的字幕语句, 记录一个标识, 内容提取结果如下实例所示: 标识 字幕语句 吋间
1001 本拉登的生死成为大家关注的一个焦点 00:25: 17-00:25:25
1002 美国反恐发言人称无证据证明本拉登已死亡 00:25:30-00:25:33
1003 目前国际恐怖活动依然猖獗 00:30:39-00:30:45
1004 现在让我们来看看体育方面的消息 00:30:45-00:30:50
1005 今天, 是中国网球公开赛的第四天 00:31: 15-00:31:20
[58] 其中每个吋间信息分为该语句在媒体播放吋的开始吋间和结束吋间, "-"前面的 部分为开始吋间, "-"后面的部分为结束吋间。
[59] 每类格式的字幕文件均有固定的格式, 有成熟的字幕内容、 格式提取工具, 如 专业的 VOBSUB字幕识别软件, 能够提取多种格式的字幕信息; 而对于 .txt等的 文本格式的字幕, 其吋间和字幕信息都是固定格式, 利用正则表达式可以提取 满足条件的信息。 该提取技术为现有技术, 本发明对此不作详细描述。
[60] 步骤 31 : 以有标识的字幕语句为单位按照定义好的主题进行分类, 形成多个包 括一个或多个代表不同主题的内容片段;
[61] 目前基于文本信息的信息分类有多种方法, 有相对成熟的现有技术。 如 TF/ID
F (term frequency , 词汇步员率 /inverse document
frequency, 逆文本频率) 在信息分类、 检索中已成为公认的方法, 还包括: 贝 叶斯算法、 Rocchio (相似度计算方法) 、 KNN (K-nearest neighbor
K近邻方法) 、 Naive
Bayes (朴素贝叶斯) 等。 所述的各种信息分类方法均能够将输入给分类器的不 同的文本内容按照不同主题进行分类。 所述的主题包括事先人为定制的知识分 类或者在分类的过程中进行机器学习的关键词目录结构等。
[62] 本发明所述的对字幕内容分类是以整个字幕文件的内容为对象, 每一个能够单 独标识的字幕语句为单位。 如图 4是分类的一个过程示意图。 其中对字幕文件的 分类过程釆用上述现有的技术之一, 而输入数据可由本发明所述系统中的媒体 内容提取单元产生, 而输出数据的接收部件可为内容分类定位单元。
[63] 进行内容分类后, 整个字幕文件分成若干个不同主题的分类信息, 即本发明所 述的内容片段, 每个内容片段包含一个或多个能够独立区分的在媒体中有起止 出现吋间的字幕语句, 即主题。 而片段与片段在吋间或包含字幕语句上可能有 交叉或包含关系。 而这些包含某一个主题内容的以字幕来代表内容的片段对应 着媒体的某一个吋间段的媒体片段。 对于一些新闻、 体育解说节目等媒体来说 , 这些由字幕反映的内容, 本身就是媒体所展现的人能够理解的内容。
[64] 步骤 32: 根据所述媒体内容吋间片段的吋间信息标注所述每个内容片段在媒体 中的具体播放吋间信息, 得到具有具体吋间信息的多个具有不同主题的内容片 段;
[65] 由于每个内容片段包含一个或多个主题, 而每个主题出现的吋间均不同。 而这 里的内容片段是对应媒体中的某一个媒体片段的。 需要根据每个主题出现的吋 间来标注内容片段出现的吋间范围。
[66] 标注方法包括: 根据一个吋间阈值 (可以是事先设定的或是根据媒体利用算法 确定) 来对吋间间隔超过所述阈值的一个内容片段内的多个主题进行切分, 成 为多个同主题的内容片段。 例如, 当某个内容片段中, 包含三个字幕语句 1001 、 1002、 1003 , 而吋间阈值设为 3分钟, 字幕 1003的开始吋间与其他的两个字幕 的吋间段相差 3分钟以上, 则将该内容片段分成分别由 1001、 1002以及 1003组成 的两个主题相同的不同的内容片段。
[67] 标注每个内容片段在媒体中播放的吋间段的方法包括: 确定每个内容片段中出 现吋间最早的语句为开始语句和结束吋间最晚的语句为结束语句, 取内容片段 的开始语句的开始吋间为内容片段在媒体播放吋的开始吋间, 结束语句的结束 吋间为内容片段的结束吋间。
[68]
Figure imgf000011_0001
步骤 4: 对上述分类处理后的内容与本体概念进行匹配, 以本体中的词汇来标 注该内容片段;
[70] 对于上述过程中进行字幕内容分类后得到的片段, 每个片段有一个代表其内容 的一个或若干主题 (可以是关键词或语句) 。 为了将该片段的内容与本体中概 念进行关联, 需要对主题与本体概念进行适配, 找到与内容片段对应的本体概 念。 所述的适配是指在本体库中找到与主题意义接近或相同的概念。 具体实现 上有多种现有方法。 例如, 可以利用传统的词语模糊匹配算法, 在本体中找到 与所匹配的主题词最接近的概念, 或是根据片段的其他的主题词进行修正, 最 后可以匹配一个或多个本体词汇来标识该片段的主题内容。
[71] 对于一些简单的关键词与本体概念匹配中, 可以利用传统的词语模糊匹配方法 。 具体方法一种实施例为: 将本体概念作为普通的词汇, 利用数据査询中的" like "函数, 找出本体中包含一部分或全部待査词汇的词。 如利用 "like"匹配方法, 可 以找出 "恐怖 "在本体中匹配的概念为"恐怖主义"; 而当有多个匹配上的本体概念 吋, 可以通过匹配上的字占概念中字数比例等方法来判断匹配度, 从而确定最 接近的本体概念。 其他的一些本体概念匹配算法, 包括引入本体推理、 相关性 匹配算法等, 能够提供更精确和效率更高的匹配方法。
[72] 而对于一些给定的本体库和给定的领域知识分类知识库, 其概念之间本身就存 在映射关系, 这种映射关系显示了某个主题词或主题语句与本体概念中词汇的 映射关系如下表所示:
[73]
Figure imgf000012_0001
[74] 对于这种有映射关系的主题, 执行过程包括: 首先选取内容片段的主题词; 之 后査找映射表中该主题词对应的本体 URI, 作为标注该内容片段的本体概念。 如 图 5所示为一个具体的内容片段与本体库关联的实施例;
[75] 步骤 5: 根据上述的匹配信息生成并存储标注信息。
[76] 标注信息包括对每个进行分类的内容片段进行记录, 记录的内容包括但不限于 : 内容片段所属的媒体标识、 内容片段所对应的媒体概念资源标识、 内容片段 在媒体中的起始吋间和终止吋间等。 上述存储的标注信息作为对媒体内容进行 管理的基础。
[77] 步骤 6: 判断是否存在待标注媒体;
[78] 如果不存在, 则结束; 如果存在, 则重复执行上述步骤 1至步骤 5的操作。
[79] 上述实施例为本发明最佳实施例, 其中找到媒体片段内容的主题如"网球"后, 对其在本体中进行匹配, 找到本体的词汇来标注该片段内容可以省略。
[80] 综上所述, 本发明实施例通过分析与媒体对应的字幕文件, 对媒体内容按吋间 区分成不同的内容片段, 并对每个片段里面的内容与本体概念进行关联, 记录 了内容片段在媒体中出现的位置。 这样将媒体的内容利用标准的词汇进行描述 , 利于对内容描述信息的统一, 使得对媒体的内容检索成为可能。
[81] 本领域普通技术人员可以理解实现上述实施例方法中的全部或部分步骤是可以 通过程序指令相关的硬件来完成, 所述的程序可以存储于一计算机可读存储介 质中 (所述的存储介质, 如: ROM/RAM 磁盘、 光盘等) , 该程序在执行吋, 包括如下步骤:
[82] 步骤 1 : 获取待标注媒体内容的媒体标识;
[83] 对于给定的媒体库, 存放至少一个媒体文件, 媒体内容提取单元从媒体库中获 取待标注的媒体内容的媒体标识, 所述媒体标识可能是媒体文件的文件名或是 专门为媒体文件建立的索弓 I等标识信息。
[84] 步骤 2: 根据所述获取的媒体标识获取对应的字幕文件;
[85] 所述的字幕文件是指为每个媒体中的对话或者其他的话音、 解释进行文字描述 的文件。 一个媒体标识可以唯一对应一份字幕文件。
[86] 目前, 字幕文件分嵌入式字幕文件和外挂字幕文件, 嵌入式字幕文件直接融入 在媒体文件中, 是不可修改和编辑的; 而外挂字幕文件需要一个另外的独立的 文件, 里面记录了按照吋间先后出现的话音等的字幕。 外挂字幕文件的格式包 括但不限于: .txt、 .srt、 .sub .ssa、 .smi等。 这些文件格式的字幕里面的字幕文 件至少包括: 字幕内容信息及字幕出现的吋间码信息 (开始吋间、 结束吋间) 、 字幕文件所对应的媒体标识信息。 所述的吋间信息在字幕中以标准格式的吋 间码的格式出现, 其格式为 XX: XX: XX, 三个字段分别表示小吋、 分、 秒。
[87] 步骤 3 : 提取上述获取的字幕文件中的字幕内容, 对所述字幕内容信息按照吋 间先后顺序进行标识, 形成多个媒体内容吋间片段, 并按照定义好的主题内容 进行分类, 得到具有不同主题的多个内容片段, 根据所述媒体内容吋间片段的 吋间信息标注每个内容片段在媒体中的具体播放吋间信息, 得到具有具体吋间 信息的多个具有不同个主题的内容片段。
[88] 步骤 4·· 对上述分类处理后的内容与本体概念进行匹配, 以本体中的词汇来标 注该内容片段;
[89] 对于上述过程中进行字幕内容分类后得到的片段, 每个片段有一个代表其内容 的一个或若干主题 (可以是关键词或语句) 。 为了将该片段的内容与本体中概 念进行关联, 需要对主题与本体概念进行适配, 找到与内容片段对应的本体概 念。 所述的适配是指在本体库中找到与主题意义接近或相同的概念。 具体实现 上有多种现有方法。 例如, 可以利用传统的词语模糊匹配算法, 在本体中找到 与所匹配的主题词最接近的概念, 或是根据片段的其他的主题词进行修正, 最 后可以匹配一个或多个本体词汇来标识该片段的主题内容。
[90] 对于一些简单的关键词与本体概念匹配中, 可以利用传统的词语模糊匹配方法 。 具体方法一种实施例为: 将本体概念作为普通的词汇, 利用数据査询中的" like "函数, 找出本体中包含一部分或全部待査词汇的词。 如利用 "like"匹配方法, 可 以找出 "恐怖 "在本体中匹配的概念为"恐怖主义"; 而当有多个匹配上的本体概念 吋, 可以通过匹配上的字占概念中字数比例等方法来判断匹配度, 从而确定最 接近的本体概念。 其他的一些本体概念匹配算法, 包括引入本体推理、 相关性 匹配算法等, 能够提供更精确和效率更高的匹配方法。
[91] 而对于一些给定的本体库和给定的领域知识分类知识库, 其概念之间本身就存 在映射关系, 这种映射关系显示了某个主题词或主题语句与本体概念中词汇的 映射关系如下表所示: [92]
主题词 对应的本体 URI 备注
1 恐怖分子 http://www.xinhua.com/terns/
恐怖主义
2 体育 http://www. xinhua.com/terns/体育
[93] 对于这种有映射关系的主题, 执行过程包括: 首先选取内容片段的主题词; 之 后査找映射表中该主题词对应的本体 URI, 作为标注该内容片段的本体概念。 如 图 5所示为一个具体的内容片段与本体库关联的实施例;
[94] 步骤 5: 根据上述的匹配信息生成并存储标注信息。
[95] 标注信息包括对每个进行分类的内容片段进行记录, 记录的内容包括但不限于 : 内容片段所属的媒体标识、 内容片段所对应的媒体概念资源标识、 内容片段 在媒体中的起始吋间和终止吋间等。 上述存储的标注信息作为对媒体内容进行 管理的基础。
[96] 步骤 6: 判断是否存在待标注媒体;
[97] 如果不存在, 则结束; 如果存在, 则重复执行上述步骤 1至步骤 5的操作。
[98] 以上所述, 仅为本发明较佳的具体实施方式, 但本发明的保护范围并不局限于 此, 任何熟悉本技术领域的技术人员在本发明揭露的技术范围内, 可轻易想到 的变化或替换, 都应涵盖在本发明的保护范围之内。 因此, 本发明的保护范围 应该以权利要求的保护范围为准。

Claims

权利要求书
[1] 1、 一种媒体内容管理系统, 其特征在于, 包括:
文本分类器, 用于对字幕信息按照定义好的主题内容进行分类, 得到多个 具有不同主题的内容片段;
媒体内容标注处理单元, 用于标注经文本分类器分类后具有不同主题的每 一个内容片段的具体播放吋间信息, 得到具有具体吋间信息的多个具有不 同主题的内容片段, 并将所述具有具体吋间信息的多个具有不同主题的内 容片段与本体库中的概念进行匹配, 并以本体库中定义的词汇来标注所述 内容片段。
[2] 2、 如权利要求 1所述的系统, 其特征在于, 所述系统还包括:
本体库, 用于存储若干概念, 包括媒体相关领域共同认可的词汇, 及所述 词汇与词汇之间相互关系;
媒体库, 用于保存具有媒体标识的具体的媒体内容;
媒体字幕库, 用于保存与媒体库中的媒体内容对应的对媒体附带的字幕信 息。
[3] 3、 如权利要求 1或 2所述的系统, 其特征在于, 所述媒体内容标注处理单元 进一步包括:
媒体内容提取单元, 用于获取待标注的媒体内容的媒体标识, 根据所述的 媒体标识获取对应的字幕信息, 并识别所述字幕信息中的字幕内容信息, 对字幕内容信息按照吋间先后顺序进行标识, 形成多个媒体内容吋间片段 内容分类定位单元, 用于根据所述媒体内容吋间片段的吋间信息标注经文 本分类器分类后具有不同主题的每一个内容片段的具体播放吋间信息, 得 到具有具体吋间信息的多个具有不同主题的内容片段;
标注适配单元, 用于将经内容分类定位单元标注了具体吋间信息且具有不 同主题的内容片段与本体库中的概念进行匹配, 并以本体库中定义的词汇 来标注所述内容片段。
[4] 4、 如权利要求 3所述的系统, 其特征在于, 所述以本体库中定义的词汇来 标注所述内容片段的标注信息包括: 每类所属的媒体标识、 每类内容对应 的本体概念标识、 每类内容的起止吋间点描述信息。
[5] 5、 如权利要求 1所述的系统, 其特征在于, 所述系统进一步包括:
媒体内容注册信息库, 用于记录与本体库中定义的词汇来标注的内容片段
[6] 6、 如权利要求 1或 2所述的系统, 其特征在于, 所述本体库中每一词汇具有 唯一的资源标识。
[7] 7、 如权利要求 2所述的系统, 其特征在于, 所述媒体标识是媒体无重名的 文件名、 或为媒体专门分配的索引、 或任何能够唯一标识该媒体的数字或 文字、 字母符号序列、 统一资源定位器 URL、 或统一资源标识 URI。
[8] 8、 如权利要求 2所述的系统, 其特征在于, 所述字幕信息包括:
字幕内容信息、 字幕出现的吋间码信息、 和 /或字幕文件所对应的媒体标识
Ι π Λ∑!、。
[9] 9、 一种媒体内容管理方法, 其特征在于, 包括:
根据待标注媒体内容的媒体标识获取对应的字幕信息;
提取上述获取的字幕信息中的字幕内容信息, 对所述字幕内容信息按照吋 间先后顺序进行标识, 形成多个媒体内容吋间片段, 并按照定义好的主题 内容对所述多个媒体内容吋间片段进行分类, 得到具有不同主题的多个内 容片段;
根据所述媒体内容吋间片段的吋间信息标注所述每个内容片段在媒体中的 具体播放吋间信息, 得到具有具体吋间信息的多个具有不同主题的内容片 段;
根据所述具有具体吋间信息的多个具有不同主题的内容片段的主题与本体 库中的概念进行匹配, 以本体库中定义的词汇来标注所述内容片段。
[10] 10、 如权利要求 9所述的方法, 其特征在于, 所述的字幕信息是为每个媒体 中的对话或者其他的话音、 解释进行文字描述的文件, 包括: 字幕内容信 息、 字幕出现的吋间码信息、 和 /或字幕文件所对应的媒体标识信息。
[11] 11、 如权利要求 9所述的方法, 其特征在于, 所述根据媒体内容吋间片段的 吋间信息标注每个内容片段在媒体中的具体播放吋间信息的方法具体包括 提取所述获取的字幕信息中的字幕内容信息, 记录每个标有起止吋间的字 幕内容的标识及吋间信息;
以有标识的字幕内容为单位按照定义好的主题进行分类, 形成多个包括一 个或多个主题的内容片段;
根据所述吋间信息标注所述每个内容片段在媒体中具体播放吋间信息。
[12] 12、 如权利要求 11所述的方法, 其特征在于, 所述根据吋间信息标注每个 内容片断在媒体中具体播放吋间信息的方法包括:
根据每个内容片段的主题在内容片段中出现的吋间标注内容片段的吋间信 息。
[13] 13、 如权利要求 12所述的方法, 其特征在于, 所述标注内容片段的吋间信 息的方法具体包括:
根据事先设定的或根据媒体利用算法确定的吋间阈值将吋间间隔超过所述 设定的阈值的一个内容片段内的多个主题切分成多个同主题的内容片段; 确定每个内容片段中出现吋间最早的语句为开始语句和结束吋间最晚的语 句为结束语句, 取内容片段的开始语句的开始吋间为内容片段在媒体播放 吋的开始吋间, 结束语句的结束吋间为内容片段的结束吋间。
[14] 14、 如权利要求 9所述的方法, 其特征在于, 所述方法还包括:
记录以本体库中定义的词汇来标注的内容片段。
PCT/CN2007/071133 2006-12-06 2007-11-27 Système et procédé de gestion de contenu média WO2008067749A1 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US12/479,066 US8200597B2 (en) 2006-12-06 2009-06-05 System and method for classifiying text and managing media contents using subtitles, start times, end times, and an ontology library

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CNB200610164838XA CN100449547C (zh) 2006-12-06 2006-12-06 一种媒体内容管理系统及方法
CN200610164838.X 2006-12-06

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US12/479,066 Continuation US8200597B2 (en) 2006-12-06 2009-06-05 System and method for classifiying text and managing media contents using subtitles, start times, end times, and an ontology library

Publications (1)

Publication Number Publication Date
WO2008067749A1 true WO2008067749A1 (fr) 2008-06-12

Family

ID=38692589

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2007/071133 WO2008067749A1 (fr) 2006-12-06 2007-11-27 Système et procédé de gestion de contenu média

Country Status (3)

Country Link
US (1) US8200597B2 (zh)
CN (1) CN100449547C (zh)
WO (1) WO2008067749A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102088568A (zh) * 2009-12-08 2011-06-08 新奥特(北京)视频技术有限公司 一种字幕制作系统

Families Citing this family (46)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7490092B2 (en) 2000-07-06 2009-02-10 Streamsage, Inc. Method and system for indexing and searching timed media information based upon relevance intervals
AU2002327677A1 (en) 2001-09-19 2003-04-01 Meta Tv, Inc. Interactive user interface for television applications
US8042132B2 (en) 2002-03-15 2011-10-18 Tvworks, Llc System and method for construction, delivery and display of iTV content
US7703116B1 (en) 2003-07-11 2010-04-20 Tvworks, Llc System and method for construction, delivery and display of iTV applications that blend programming information of on-demand and broadcast service offerings
US8220018B2 (en) 2002-09-19 2012-07-10 Tvworks, Llc System and method for preferred placement programming of iTV content
US8578411B1 (en) 2003-03-14 2013-11-05 Tvworks, Llc System and method for controlling iTV application behaviors through the use of application profile filters
US11381875B2 (en) 2003-03-14 2022-07-05 Comcast Cable Communications Management, Llc Causing display of user-selectable content types
US8819734B2 (en) 2003-09-16 2014-08-26 Tvworks, Llc Contextual navigational control for digital television
US7818667B2 (en) 2005-05-03 2010-10-19 Tv Works Llc Verification of semantic constraints in multimedia data and in its announcement, signaling and interchange
CN100449547C (zh) * 2006-12-06 2009-01-07 华为技术有限公司 一种媒体内容管理系统及方法
US11832024B2 (en) 2008-11-20 2023-11-28 Comcast Cable Communications, Llc Method and apparatus for delivering video and video-related content at sub-asset level
US8713016B2 (en) 2008-12-24 2014-04-29 Comcast Interactive Media, Llc Method and apparatus for organizing segments of media assets and determining relevance of segments to a query
US20100161441A1 (en) * 2008-12-24 2010-06-24 Comcast Interactive Media, Llc Method and apparatus for advertising at the sub-asset level
US9442933B2 (en) 2008-12-24 2016-09-13 Comcast Interactive Media, Llc Identification of segments within audio, video, and multimedia items
US11531668B2 (en) 2008-12-29 2022-12-20 Comcast Interactive Media, Llc Merging of multiple data sets
US8176043B2 (en) 2009-03-12 2012-05-08 Comcast Interactive Media, Llc Ranking search results
US8533223B2 (en) 2009-05-12 2013-09-10 Comcast Interactive Media, LLC. Disambiguation and tagging of entities
US9892730B2 (en) 2009-07-01 2018-02-13 Comcast Interactive Media, Llc Generating topic-specific language models
CN102087649B (zh) * 2009-12-08 2015-04-29 新奥特(北京)视频技术有限公司 一种字幕素材文件和伴侣文件的管理方法和装置
CN102024009A (zh) * 2010-03-09 2011-04-20 李平辉 视频场景库生成方法及系统、搜索视频场景的方法及系统
CN102136008A (zh) * 2011-04-02 2011-07-27 天脉聚源(北京)传媒科技有限公司 一种音视频数据标引的方法和系统
CN102739997A (zh) * 2011-05-10 2012-10-17 新奥特(北京)视频技术有限公司 一种字幕插件分段渲染方法
CN102377914A (zh) * 2011-09-22 2012-03-14 宇龙计算机通信科技(深圳)有限公司 终端和多媒体文件管理方法
US8620917B2 (en) * 2011-12-22 2013-12-31 Telefonaktiebolaget L M Ericsson (Publ) Symantic framework for dynamically creating a program guide
CN102752541B (zh) * 2012-02-28 2018-05-04 新奥特(北京)视频技术有限公司 一种比赛现场字幕的生成方法
US8577671B1 (en) * 2012-07-20 2013-11-05 Veveo, Inc. Method of and system for using conversation state information in a conversational interaction system
US9465833B2 (en) 2012-07-31 2016-10-11 Veveo, Inc. Disambiguating user intent in conversational interaction system for large corpus information retrieval
CN103838751A (zh) * 2012-11-23 2014-06-04 鸿富锦精密工业(深圳)有限公司 视频内容搜索系统及方法
US10880609B2 (en) 2013-03-14 2020-12-29 Comcast Cable Communications, Llc Content event messaging
US9251125B2 (en) * 2013-03-15 2016-02-02 International Business Machines Corporation Managing text in documents based on a log of research corresponding to the text
DK2994908T3 (da) 2013-05-07 2019-09-23 Veveo Inc Grænseflade til inkrementel taleinput med realtidsfeedback
US10380253B2 (en) * 2014-03-04 2019-08-13 International Business Machines Corporation Natural language processing with dynamic pipelines
KR102244298B1 (ko) 2014-04-30 2021-04-23 삼성전자주식회사 의미를 기반으로 웹 페이지 접근 기록을 구조화하는 장치 및 방법
US11783382B2 (en) 2014-10-22 2023-10-10 Comcast Cable Communications, Llc Systems and methods for curating content metadata
US9852136B2 (en) 2014-12-23 2017-12-26 Rovi Guides, Inc. Systems and methods for determining whether a negation statement applies to a current or past query
US9854049B2 (en) 2015-01-30 2017-12-26 Rovi Guides, Inc. Systems and methods for resolving ambiguous terms in social chatter based on a user profile
CN104883584A (zh) * 2015-05-19 2015-09-02 福建宏天信息产业有限公司 一种远程解析字幕的方法及系统
US10747801B2 (en) * 2015-07-13 2020-08-18 Disney Enterprises, Inc. Media content ontology
US9544704B1 (en) * 2015-07-16 2017-01-10 Avaya Inc. System and method for evaluating media segments for interestingness
KR101983493B1 (ko) * 2017-04-10 2019-05-29 한국과학기술원 자연어 텍스트로부터 학습된 객체 표상에 포함된 특성 해석 및 시각화 방법 그리고 시스템
CN107908762A (zh) * 2017-11-17 2018-04-13 广州慧睿思通信息科技有限公司 一种自定义关键词串并历史数据的方法及系统
CN109408626B (zh) * 2018-11-09 2021-09-21 思必驰科技股份有限公司 对自然语言进行处理的方法及装置
US11163953B2 (en) * 2019-06-12 2021-11-02 International Business Machines Corporation Natural language processing and candidate response evaluation
US11429789B2 (en) 2019-06-12 2022-08-30 International Business Machines Corporation Natural language processing and candidate response identification
CN112329423A (zh) * 2020-11-05 2021-02-05 上海钐昆网络科技有限公司 Icp备案公司分类方法、装置、电子设备及计算机存储介质
CN112765460A (zh) * 2021-01-08 2021-05-07 北京字跳网络技术有限公司 会议信息查询方法、装置、存储介质、终端设备和服务器

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH1021217A (ja) * 1996-07-02 1998-01-23 Fuji Xerox Co Ltd ハイパーメディア装置
US20020059303A1 (en) * 2000-10-27 2002-05-16 Yoshihiro Ohmori Multimedia data management system
CN1851705A (zh) * 2006-05-30 2006-10-25 南京大学 基于本体的主题式网络爬虫系统构建方法
CN1863278A (zh) * 2006-01-09 2006-11-15 华为技术有限公司 一种实现字幕功能的方法及系统
CN101000617A (zh) * 2006-12-06 2007-07-18 华为技术有限公司 一种媒体内容管理系统及方法

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030115598A1 (en) * 2001-03-23 2003-06-19 Pantoja William E. System and method for interactively producing a web-based multimedia presentation
CN1445684A (zh) * 2002-03-20 2003-10-01 文化传信科技(澳门)有限公司 媒体管理方法以及系统
US20040032486A1 (en) * 2002-08-16 2004-02-19 Shusman Chad W. Method and apparatus for interactive programming using captioning
CN1430166A (zh) * 2003-01-07 2003-07-16 财团法人资讯工业策进会 建立影片检索数据库的方法及记录媒体

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH1021217A (ja) * 1996-07-02 1998-01-23 Fuji Xerox Co Ltd ハイパーメディア装置
US20020059303A1 (en) * 2000-10-27 2002-05-16 Yoshihiro Ohmori Multimedia data management system
CN1863278A (zh) * 2006-01-09 2006-11-15 华为技术有限公司 一种实现字幕功能的方法及系统
CN1851705A (zh) * 2006-05-30 2006-10-25 南京大学 基于本体的主题式网络爬虫系统构建方法
CN101000617A (zh) * 2006-12-06 2007-07-18 华为技术有限公司 一种媒体内容管理系统及方法

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102088568A (zh) * 2009-12-08 2011-06-08 新奥特(北京)视频技术有限公司 一种字幕制作系统

Also Published As

Publication number Publication date
CN100449547C (zh) 2009-01-07
US8200597B2 (en) 2012-06-12
CN101000617A (zh) 2007-07-18
US20090240650A1 (en) 2009-09-24

Similar Documents

Publication Publication Date Title
WO2008067749A1 (fr) Système et procédé de gestion de contenu média
US10445359B2 (en) Method and system for classifying media content
JP3936243B2 (ja) 音声注釈を使用した、画像におけるイベントを区分及び識別するための方法及びシステム
US20100274667A1 (en) Multimedia access
US20050038814A1 (en) Method, apparatus, and program for cross-linking information sources using multiple modalities
US20070112839A1 (en) Method and system for expansion of structured keyword vocabulary
Carrive et al. Transdisciplinary analysis of a corpus of French newsreels: The ANTRACT Project
US20080208872A1 (en) Accessing multimedia
JP2010524280A (ja) 第1のメディアアイテム及び第2のメディアアイテムの同時再生を可能とする方法及び装置
Repp et al. Towards to an automatic semantic annotation for multimedia learning objects
de Jong et al. Multimedia search without visual analysis: the value of linguistic and contextual information
Lindsay et al. Representation and linking mechanisms for audio in MPEG-7
Turcu et al. Video transcript indexing and retrieval procedure
US20090106243A1 (en) System for obtaining of transcripts of non-textual media
Gonsalves et al. ML-Based Indexing of Media Libraries for Insights and Search
Gravier et al. Exploiting speech for automatic TV delinearization: From streams to cross-media semantic navigation
Angeles et al. Discovering Metadata Inconsistencies.
Alan et al. Ontological video annotation and querying system for soccer games
Heiden et al. Transdisciplinary Analysis of a Corpus of French Newsreels: The ANTRACT Project
Jacob et al. An innovative Method of Accessing Digital Video Archives through Video Indexing
Moutselakis et al. Semantic Web multimedia Metadata retrieval: A music approach
Chaudhuri Structured Models for Audio Content Analysis
Papageorgiou et al. Multimedia Indexing and Retrieval Using Natural Language, Speech and Image Processing Methods
Zewide et al. Audio data model for multi-criteria query formulation and retrieval
Allasia et al. Audio-visual content analysis in P2P networks: the SAPIR approach

Legal Events

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

Ref document number: 07817323

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 07817323

Country of ref document: EP

Kind code of ref document: A1