CN112989118B - Video recall method and device - Google Patents
Video recall method and device Download PDFInfo
- Publication number
- CN112989118B CN112989118B CN202110156410.5A CN202110156410A CN112989118B CN 112989118 B CN112989118 B CN 112989118B CN 202110156410 A CN202110156410 A CN 202110156410A CN 112989118 B CN112989118 B CN 112989118B
- Authority
- CN
- China
- Prior art keywords
- tag
- recalled
- target
- video
- 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.)
- Active
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
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/78—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/7867—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using information manually generated, e.g. tags, keywords, comments, title and artist information, manually generated time, location and usage information, user ratings
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/73—Querying
- G06F16/735—Filtering based on additional data, e.g. user or group profiles
Abstract
The embodiment of the invention provides a video recall method and a device, wherein the method comprises the following steps: acquiring a target tag set of a target video and a recall tag set of a recall video; obtaining a target tag vector corresponding to a target tag in the target tag set, and obtaining a to-be-recalled tag vector corresponding to a to-be-recalled tag in the to-be-recalled tag set; summing the target tag vectors to obtain target tag set vectors, and summing the tag vectors to be recalled to obtain tag set vectors to be recalled; calculating the similarity between the target tag set vector and the tag set vector to be recalled; and recalling videos from the videos to be recalled corresponding to the label set vectors to be recalled, wherein the similarity of the videos to be recalled meets a preset condition. According to the embodiment of the invention, the video recall is performed by taking the tag set as a retrieval unit, so that the correlation between the recall video and the target video can be improved, and the quality of the recall video is improved.
Description
Technical Field
The present invention relates to the field of computer technology, and in particular, to a video recall method, a video recall device, an electronic apparatus, and a computer readable storage medium.
Background
With the rapid development of informatization, various internet contents are layered endlessly, and browsing videos has become an important leisure mode for people, but users are usually interested in only part of the videos, and users are individually recommended with the aim of saving time. Specifically, the video recommendation process is to recall similar videos based on target videos browsed by a user first, and then select videos from the recalled similar videos to recommend to the user.
However, when similar videos are recalled, it is sometimes difficult to acquire similar videos to the target video, resulting in degradation of quality of the recalled videos.
Disclosure of Invention
The embodiment of the invention aims to provide a video recall method, which is based on a video tag set of a target video to recall videos and improves the quality of the recall videos. The specific technical scheme is as follows:
in a first aspect of the present invention, there is provided a video recall method, the method comprising:
acquiring a target tag set of a target video and a recall tag set of a recall video;
obtaining a target tag vector corresponding to a target tag in the target tag set, and obtaining a to-be-recalled tag vector corresponding to a to-be-recalled tag in the to-be-recalled tag set; the method comprises the steps that a tag vector is obtained from a vector set, and the vector set is generated according to a video tag which has a co-occurrence relationship and replaces a synonym with a specified synonym corresponding to the synonym;
Summing the target tag vectors to obtain target tag set vectors, and summing the tag vectors to be recalled to obtain tag set vectors to be recalled;
calculating the similarity between the target tag set vector and the tag set vector to be recalled;
and recalling videos from the videos to be recalled corresponding to the label set vectors to be recalled, wherein the similarity of the videos to be recalled meets a preset condition.
Optionally, the obtaining the target tag vector corresponding to the target tag in the target tag set and obtaining the recall-to-be-obtained tag vector corresponding to the recall-to-be-obtained tag in the recall-to-be-obtained tag set includes:
acquiring a vector set; the vector set comprises label vectors corresponding to the video labels;
obtaining a target tag vector corresponding to a target tag in the target tag set from the vector set, and obtaining a to-be-recalled tag vector corresponding to a to-be-recalled tag in the to-be-recalled tag set from the vector set.
Optionally, before the acquiring the set of vectors, the method further comprises:
acquiring video tags with co-occurrence relations, wherein the co-occurrence relations are used for representing the probability of the simultaneous occurrence of the video tags;
When the synonyms exist in the video tags, replacing the synonyms in the video tags with specified synonyms corresponding to the synonyms;
and inputting a preset word vector model to obtain a corresponding tag vector by adopting the video tag after replacing the synonym, and storing the tag vector into a vector set.
Optionally, when the synonym exists in the video tag, replacing the synonym in the video tag with the specified synonym corresponding to the synonym includes:
and when the matched synonyms exist in the preset synonym table in the video tag, replacing the synonyms in the video tag with specified synonyms corresponding to the synonyms.
Optionally, the obtaining, from the vector set, a target tag vector corresponding to a target tag in the target tag set, and obtaining, from the vector set, a recall-to-tag vector corresponding to a recall-to-tag in the recall-to-tag set, includes:
when the target label vector corresponding to the target label in the target label set does not exist in the vector set, the target label is segmented to obtain a sub-word, the target sub-label vector corresponding to the sub-word of the target label is obtained from the vector set, the target sub-label vector is summed to obtain a target label vector, and
When the recall-to-be-labeled vector corresponding to the recall-to-be-labeled in the recall-to-be-labeled set does not exist in the vector set, word segmentation is carried out on the recall-to-be-labeled to obtain sub words, recall-to-be-labeled sub-vectors corresponding to the sub words of the recall-to-be-labeled are obtained from the vector set, and summation is carried out on the recall-to-be-labeled sub-vectors to obtain the recall-to-be-labeled vector.
Optionally, the recalling the video from the to-be-recalled videos corresponding to the to-be-recalled tag set vector with the similarity meeting a preset condition includes:
acquiring a target to-be-recalled tag set which is larger than a preset similarity threshold and corresponds to the to-be-recalled tag set vector of the previous N in sequence, wherein N is a positive integer; the target to-be-recalled tag set comprises a target tag set of the target video;
creating a similar video index list according to the target to-be-recalled tag set;
acquiring the associated video to be recalled as a target video to be recalled according to the similar video index list;
determining similar videos from the target video to be recalled based on a preset posterior index; the posterior indicator includes click rate.
Optionally, the target video and the video to be recalled have channel identifications, and the obtaining the video to be recalled associated with the target tag set to be recalled as the target video to be recalled includes:
and acquiring the video to be recalled, which is associated with the target video to be recalled tag set and has the same channel identification as the channel identification of the target video, as a target video to be recalled.
In a second aspect of the present invention, there is also provided a video recall device, the device comprising:
the tag set acquisition module is used for acquiring a target tag set of a target video and a to-be-recalled tag set of a to-be-recalled video;
the tag vector acquisition module is used for acquiring a target tag vector corresponding to a target tag in the target tag set and acquiring a to-be-recalled tag vector corresponding to a to-be-recalled tag in the to-be-recalled tag set; the method comprises the steps that a tag vector is obtained from a vector set, and the vector set is generated according to a video tag which has a co-occurrence relationship and replaces a synonym with a specified synonym corresponding to the synonym;
the vector summation module is used for summing the target tag vectors to obtain target tag set vectors, and summing the tag vectors to be recalled to obtain tag set vectors to be recalled;
The similarity calculation module is used for calculating the similarity between the target tag set vector and the to-be-recalled tag set vector;
and the video recall module is used for recalling videos from the videos to be recalled corresponding to the label set vectors to be recalled, wherein the similarity of the videos to be recalled meets a preset condition.
In yet another aspect of the present invention, there is also provided a computer readable storage medium having instructions stored therein that, when executed on a computer, cause the computer to perform any of the video recall methods described above.
In yet another aspect of the invention, there is also provided a computer program product containing instructions that, when run on a computer, cause the computer to perform any of the video recall methods described above.
According to the video recall method provided by the embodiment of the invention, the target label set of the target video and the to-be-recalled label set of the to-be-recalled video are obtained, the target label vector corresponding to the target label in the target label set and the to-be-recalled label vector corresponding to the to-be-recalled label in the to-be-recalled label set are obtained, then the target label vector is summed to obtain the target label set vector, and the to-be-recalled label set vector is summed, the similarity between the target label set vector and the to-be-recalled label set vector is calculated, the recall video in the to-be-recalled video corresponding to the to-be-recalled label set vector with the similarity meeting the preset condition is recalled by taking the label set as a retrieval unit, so that the correlation between the recall video and the target video can be improved, and the quality of the recall video is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below.
FIG. 1 is a flowchart illustrating steps of a video recall method according to an embodiment of the present invention;
FIG. 2 is a flowchart illustrating steps for generating a tag vector according to an embodiment of the present invention;
FIG. 3 is a flowchart illustrating steps of another video recall method provided in an embodiment of the present invention;
FIG. 4 is a block diagram of a video recall device according to an embodiment of the present invention;
fig. 5 is a block diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be described below with reference to the accompanying drawings in the embodiments of the present invention.
Referring to fig. 1, a flowchart of steps of a video recall method according to an embodiment of the present invention is shown in fig. 1, where the method specifically includes the following steps:
step 101, acquiring a target tag set of a target video and a recall tag set of a recall video.
In the embodiment of the invention, the target video can refer to a video which is browsed on a user video website in a historical way or is browsed currently. The video to be recalled may refer to all videos on the video website, or may be videos screened out by a preset screening condition, where the screening condition may be time, on-demand amount, browsing duration, and the like, or may be a video type not interested by a user browsing the target video, for example, if the user is not interested in the horror video, the horror video may be removed in advance, so that the video to be recalled does not include the horror video. For example, if the user is browsing video a on a video website, then video a is the target video.
The tag set comprises one or more video tags, wherein the video tags are a set of one or more video tags, and the video tags are used for disassembling keywords of video content in different dimensions, so that description information of the video is formed, for example, the video tags can be classification, partition, title, introduction and the like, and the video can be efficiently searched and accurately positioned through the video tags. For example, for a child animation, the video tags in its tag set may include: cartoon, juvenile, daily, family and friend, for a scout movie, the video tags in its tag set may include: movies, case breaking, brain burning, love and comedy.
In the embodiment of the invention, each video has a corresponding tag set, each video is represented by a unique tag set, and the same tag set can represent a plurality of videos. For example, in the embodiment of the present invention, the target video and the video to be recalled have their corresponding unique tag sets, respectively, the target tag set and the tag set to be recalled.
Step 102, obtaining a target tag vector corresponding to a target tag in the target tag set, and obtaining a recall tag vector corresponding to a recall tag in the recall tag set.
In the embodiment of the invention, the video tags in the tag set can be represented by tag vectors, specifically, the tag vectors are a group of data for mapping the video tags into real numbers, and the video tags are converted into a group of data which can be identified by a machine, so that the subsequent analysis and processing are facilitated. Specifically, a text matching algorithm, such as word2vec, TFIDF, LDA, and the like, may be used to obtain a target tag vector corresponding to a target tag in the target tag set, and a recall tag vector corresponding to a recall tag in the recall tag set.
The target label vector and the label vector to be recalled are label vectors, the label vectors are obtained from a vector set, and the vector set is generated according to a video label which has a co-occurrence relationship and replaces a synonym with a specified synonym corresponding to the synonym. Specifically, the co-occurrence relationship is used for representing the probability of the simultaneous occurrence of the video tags, for example, the probability of the simultaneous occurrence of two video tags, such as a student and a teacher, is high, so that the two video tags can be regarded as having the co-occurrence relationship, in addition, in consideration of the fact that the video tags in the video tag set may be some near-synonyms, in order to avoid processing too many similar words, in the embodiment of the invention, the video tags which are synonyms are uniformly determined to be designated synonyms, for example, the video tags "company" and "enterprise" are synonyms, and the two video tags are uniformly determined to be designated synonyms "company".
And step 103, summing the target tag vectors to obtain target tag set vectors, and summing the to-be-recalled tag vectors to obtain to-be-recalled tag set vectors.
In the embodiment of the invention, after the tag vector corresponding to the video tag in the tag set is obtained, the tag vector summation can be based on the tag vector summation, so that the tag set vector corresponding to the video is generated, and one video can be represented by one tag set vector.
Specifically, the target tag vectors in the target tag sets are summed to obtain target tag set vectors of the target video, and the target tag set vectors of the video to be recalled are summed to obtain the target tag set vectors of the video to be recalled.
Step 104, calculating the similarity between the target label set vector and the label set vector to be recalled.
Step 105, recall the video from the to-be-recalled videos corresponding to the to-be-recalled tag set vector with the similarity meeting a preset condition.
In the embodiment of the invention, the similarity between the target tag set and the to-be-recalled tag set can be determined according to the similarity between the target tag set vector and the to-be-recalled tag set vector, and when the similarity is larger than the preset similarity threshold, the target tag set and the to-be-recalled tag set are indicated to have certain similarity, so that the video is recalled from the to-be-recalled video based on the to-be-recalled tag set with the similarity larger than the preset similarity threshold, and the video with higher similarity to the target video can be recalled.
In the video recall method, the target tag set of the target video and the to-be-recalled tag set of the to-be-recalled video are obtained, the target tag vector corresponding to the target tag in the target tag set and the to-be-recalled tag vector corresponding to the to-be-recalled tag in the to-be-recalled tag set are obtained, then the target tag vector is summed to obtain the target tag set vector, and the to-be-recalled tag set vector is summed to obtain the to-be-recalled tag set vector, the similarity between the target tag set vector and the to-be-recalled tag set vector is calculated, and the recall video is recalled from the to-be-recalled video corresponding to the to-be-recalled tag set vector with the tag set as a retrieval unit, so that the correlation between the recall video and the target video can be improved, and the quality of the recall video is improved.
In an exemplary embodiment of the present invention, the step 102 of obtaining a target tag vector corresponding to a target tag in the target tag set, and obtaining a recall-to-tag vector corresponding to a recall-to-tag in the recall-to-tag set includes:
acquiring a vector set; the vector set comprises label vectors corresponding to the video labels;
Obtaining a target tag vector corresponding to a target tag in the target tag set from the vector set, and obtaining a to-be-recalled tag vector corresponding to a to-be-recalled tag in the to-be-recalled tag set from the vector set.
The word vector model may be word2vec, TFIDF or LDA algorithm. In the embodiment of the invention, the video label is input into the word vector model in advance for training, so that the label vector corresponding to the video label is generated, and then the label vector is stored into the vector set. Then, a label vector corresponding to the video label can be obtained from the vector set.
Specifically, after the target tag set of the target video and the to-be-recalled tag set of the to-be-recalled video are obtained, a vector set is obtained, then a target tag vector corresponding to a target tag in the target tag set can be obtained from the vector set, and a to-be-recalled tag vector corresponding to the to-be-recalled tag in the to-be-recalled tag set is obtained from the vector set.
In the above exemplary embodiment, the vector set is obtained by training the word vector model with the video tag, so that the target tag vector corresponding to the target tag in the target tag set and the to-be-recalled tag vector corresponding to the to-be-recalled tag in the to-be-recalled tag set are obtained from the vector set, so that the similarity is calculated for recalling the video, and the relevance and accuracy of the recalled video can be improved.
In an exemplary embodiment of the invention, before the acquiring the set of vectors, the method further comprises:
step 201, obtaining a video tag with a co-occurrence relationship, wherein the co-occurrence relationship is used for representing the probability of the simultaneous occurrence of the video tag.
The co-occurrence relationship (tag sequence) is used for representing the probability that video tags appear simultaneously, for example, the video tags of "children" and "cartoons" often appear simultaneously, so that the video tags of "children" and "cartoons" can be regarded as having the co-occurrence relationship, and the video tags of "game roles" and "blood volume values" can be regarded as having the co-occurrence relationship. In the embodiment of the invention, the existing video tags are acquired, and the video tags which do not contain the co-occurrence relationship are removed.
And 202, when the synonym exists in the video tag, replacing the synonym in the video tag with a specified synonym corresponding to the synonym.
In a specific implementation, a large number of synonyms may exist in the video tag, and the synonyms refer to the same meaning, if the tag vector of the video tag is constructed without distinction, it cannot be guaranteed that the optimal video will be reserved by the similar video serving as the target video, and the quality of the recalled similar video is reduced, so that the synonyms existing in the video tag are replaced in the embodiment of the invention.
In an exemplary embodiment of the present invention, when there is a synonym in the video tag, the step 202 of replacing the synonym in the video tag with a specified synonym corresponding to the synonym includes: and when the matched synonyms exist in the preset synonym table in the video tag, replacing the synonyms in the video tag with specified synonyms corresponding to the synonyms.
Specifically, the embodiment of the invention can determine whether the video tag includes the synonym according to the preset synonym table. For example, assuming that "prize and reward", "company treatment", "reward" are included in the video tag, are determined to be synonyms according to the preset synonym table, and the specified synonyms of these synonyms are set to "reward" in the preset synonym table, the "company treatment" and "reward" may be replaced with "company reward" and "reward". In the above exemplary embodiment, synonyms in the video tag are normalized to the same kind of representation form, so that the accuracy of tag vector expression is improved.
And 203, adopting the video tag after replacing the synonyms, inputting a preset word vector model to obtain a corresponding tag vector, and storing the tag vector into a vector set.
In the embodiment of the invention, the video tags which do not contain the co-occurrence relationship are removed, synonyms existing in the video tags are normalized into the same expression form, and then the tag vectors are trained in a word vector model, such as word2vec, according to the video tags and stored in a vector set for standby.
In the above exemplary embodiment, the label vector is obtained by training by using the word vector model with the co-occurrence relationship and the synonym normalized to the same expression form, and the label vector with strong correlation and high expression accuracy can be obtained.
In an exemplary embodiment of the present invention, the obtaining, from the vector set, a target tag vector corresponding to a target tag in the target tag set, and obtaining, from the vector set, a recall-to-tag vector corresponding to a recall-to-tag in the recall-to-tag set, includes:
when the target label vector corresponding to the target label in the target label set does not exist in the vector set, the target label is segmented to obtain a sub-word, the target sub-label vector corresponding to the sub-word of the target label is obtained from the vector set, the target sub-label vector is summed to obtain a target label vector, and
When the recall-to-be-labeled vector corresponding to the recall-to-be-labeled in the recall-to-be-labeled set does not exist in the vector set, word segmentation is carried out on the recall-to-be-labeled to obtain sub words, recall-to-be-labeled sub-vectors corresponding to the sub words of the recall-to-be-labeled are obtained from the vector set, and summation is carried out on the recall-to-be-labeled sub-vectors to obtain the recall-to-be-labeled vector.
The video tag may include a long tail tag. Specifically, the long tail tag refers to a video tag containing keywords, and is usually composed of 2-3 phrases or a phrase, and is usually used for further explanation and definition of the keywords. Such as: the keyword 'red bean' can be used as a video tag, and the 'nutritional value of red bean' can also be used as a long tail tag containing the keyword. In practice, an accurate video may be located by the long tail tag, but there is a problem in that any He Shipin may not be located by the long tail tag.
In the embodiment of the invention, when the video tags are trained based on the word vector model to obtain the tag vectors, a reasonable minimum word frequency is set for the word vectors, so that some video tags with fewer occurrence times, such as long tail tags, can be filtered out, and therefore, vector tags with no or only a small number of long tail tags may be contained in the vector set.
In view of the above problems, in the embodiment of the present invention, if a tag vector corresponding to a video tag exists in a vector set, the tag vector is directly obtained, and if a tag vector corresponding to a video tag does not exist in the vector set, it is indicated that the video tag may be a long-tail tag, then the video tag may be segmented to obtain a sub-word, then a sub-tag vector corresponding to the sub-word of the video tag is obtained from the vector set, and the sub-tag vectors are summed, thereby obtaining a tag vector corresponding to the long-tail tag.
In the above exemplary embodiment, the video tag that is the long tail tag is segmented, so as to obtain the sub-tag vector corresponding to the long tail tag, and sum the sub-tag vectors to obtain the tag vector corresponding to the long tail tag, so that the video tag can obtain the corresponding tag vector, and the overall coverage of the recall video can be improved.
In an exemplary embodiment of the present invention, the step 105 of recalling videos from the to-be-recalled videos corresponding to the to-be-recalled tag set vector whose similarity satisfies a preset condition includes:
acquiring a target to-be-recalled tag set which is larger than a preset similarity threshold and corresponds to the to-be-recalled tag set vector of the previous N in sequence, wherein N is a positive integer; the target to-be-recalled tag set comprises a target tag set of the target video;
Creating a similar video index list according to the target to-be-recalled tag set;
acquiring the associated video to be recalled as a target video to be recalled according to the similar video index list;
determining similar videos from the target video to be recalled based on a preset posterior index; the posterior indicator includes click rate.
In the embodiment of the invention, the similarity may be cosine similarity, the preset similarity threshold may be set to 0.8, and n may be set to 5. In the embodiment of the invention, the cosine similarity between the target tag set vector and the to-be-recalled tag set vector is calculated, the cosine similarity is reserved to be more than 0.8 according to descending order, the to-be-recalled tag sets corresponding to the to-be-recalled tag set vectors in the previous 5 are ordered to be used as target to-be-recalled tag sets, then the to-be-recalled video associated with the target to-be-recalled tag sets is used as target to-be-recalled video, and similar videos are further determined from the target to-be-recalled video based on posterior indexes, so that video recall is completed.
The posterior index may be a Click-through Rate (CTR), specifically, the Click Rate refers to a ratio of the number of times a video is clicked to the number of times displayed on the video website, i.e. clicks/views, which is a percentage, reflecting the attention degree of a video on the video website. For example, assuming that a video window of video B is displayed on a certain web page of a video website, if one of 10 persons clicks on the video window of video B on the opened web page, the click rate of the video B is 10%. Of course, in addition to the click rate, a collection rate, comment rate, sharing rate, forwarding rate, or the like may be selected as the posterior index, which is not limited in the embodiment of the present invention.
It should be noted that, the video to be recalled may include the target video, so the target set of tags to be recalled may include the target tag set of the target tag.
In the embodiment of the invention, one or more target to-be-recalled tag sets can be combined into one target to-be-recalled tag set list, the target to-be-recalled tag set list can comprise a target tag set of target videos, and then a similar video index list can be reversely created according to the target to-be-recalled tag set list, so that the similar videos finally recalled are determined based on the similar video index list. Specifically, the similar video index list may include a video identifier, where the video identifier may be a name or a number of the video to be recalled, so that the video to be recalled may be obtained through the similar video index list.
Specifically, firstly, a video set V0 (target to-be-recalled video) corresponding to a tag set T0 in a target to-be-recalled tag set list is obtained, wherein T0 is the target tag set of target videos, v_i represents a similar video index list (target videos), and X videos with highest posterior indexes (such as CTR) in the V0 are taken as target videos to be written into the v_i; if the number of videos in v_i is less than Y at this time, sequentially taking out the video set V1 corresponding to the tag set T1 after T0 in the target to-be-recalled tag set list, taking the X videos with the highest CTR, writing the X videos into v_i, and repeating the steps until the number degree of the videos in v_i meets Y.
In the above exemplary embodiment, a target recall tag set corresponding to the recall tag set vector with similarity greater than a preset similarity threshold and ranked in the previous N is obtained, then an associated target recall video is determined based on the target recall tag set, finally a similar video for recall of the target video is determined based on a posterior index, and the recall video is recalled through a plurality of combined screening conditions, so that the correlation of the recall video is improved.
In an exemplary embodiment of the present invention, the target video and the video to be recalled have channel identifications, and the obtaining the video to be recalled associated with the target set of tags to be recalled as the target video to be recalled includes:
and acquiring the video to be recalled, which is associated with the target video to be recalled tag set and has the same channel identification as the channel identification of the target video, as a target video to be recalled.
Wherein each video may have a corresponding channel identification, such as "sports," "entertainment," "newsletter," and so forth. In the embodiment of the invention, when the target video to be recalled is determined, the target video to be recalled needs to be associated with the target video to be recalled and also needs to have the same channel label as the channel identification of the target video. For example, assuming that the channel of the target video is identified as "entertainment", only the video to be recalled whose channel is identified as "entertainment" can be taken as the target video to be recalled.
In the above-described exemplary embodiment, the video to be recalled having the same channel identification as that of the target video is taken as the target video to be recalled, thereby ensuring the correlation between the final recall similar video and the target video.
Referring to fig. 3, a flowchart illustrating steps of another video recall method according to an embodiment of the present invention, as shown in fig. 3, may specifically include the following steps:
step 301, obtaining a video tag with a co-occurrence relationship, wherein the co-occurrence relationship is used for representing the probability of the simultaneous occurrence of the video tag;
step 302, when a matched synonym exists in a preset synonym table in the video tag, replacing the synonym in the video tag with a specified synonym corresponding to the synonym;
step 303, inputting a preset word vector model to obtain a corresponding tag vector by adopting the video tag after replacing the synonym, and storing the tag vector into a vector set;
step 304, obtaining a target tag set of a target video and a to-be-recalled tag set of a to-be-recalled video;
step 305, obtaining a target tag vector corresponding to a target tag in the target tag set from the vector set, and obtaining a to-be-recalled tag vector corresponding to a to-be-recalled tag in the to-be-recalled tag set from the vector set;
Step 306, summing the target tag vectors to obtain a target tag set vector, and summing the to-be-recalled tag vectors to obtain a to-be-recalled tag set vector;
step 307, calculating the similarity between the target tag set vector and the to-be-recalled tag set vector;
step 308, recall the video from the to-be-recalled videos corresponding to the to-be-recalled tag set vector whose similarity satisfies a preset condition.
In the video recall method, video tags with co-occurrence relations are obtained, synonyms existing in the video tags are normalized to be in the same representation form, then the video tags are used as training data, word2vec or other algorithms are used for obtaining tag vectors and storing the tag vectors into a vector set, then when similar videos of target videos are recalled, target tag vectors corresponding to target tags in the target tag set can be obtained from the vector set and summed to obtain target tag set vectors, and to-be-recalled tag vectors corresponding to-be-recalled tags in the to-be-recalled tag set are obtained from the vector set and summed to obtain to-be-recalled tag set vectors, similarity between the target tag set vectors and the to-be-recalled tag set vectors is calculated, and finally the similar videos of the target videos are determined based on to-be-recalled tag set vectors with similarity meeting preset conditions. Therefore, the embodiment of the invention improves the overall coverage of the label vector of the video label, thereby improving the accuracy of the recall result of the label vector, simultaneously providing synonym replacement by considering the phenomenon of a large number of video labels with a plurality of synonyms, taking the label set as a similarity index list of a retrieval unit, reversely creating similar videos with video granularity according to business strategies such as a posterior index and the like, improving the posterior index of the recalled similar videos on the basis of ensuring the correlation, and achieving the aim of improving the quality of the recalled videos in all aspects.
It should be noted that, for simplicity of description, the method embodiments are shown as a series of acts, but it should be understood by those skilled in the art that the embodiments are not limited by the order of acts, as some steps may occur in other orders or concurrently in accordance with the embodiments. Further, those skilled in the art will appreciate that the embodiments described in the specification are presently preferred embodiments, and that the acts are not necessarily required by the embodiments of the invention.
Referring to fig. 4, which is a block diagram of a video recall device according to an embodiment of the present invention, as shown in fig. 4, the device 40 may specifically include the following modules:
the tag set obtaining module 401 is configured to obtain a target tag set of a target video and a to-be-recalled tag set of a to-be-recalled video;
a tag vector obtaining module 402, configured to obtain a target tag vector corresponding to a target tag in the target tag set, and obtain a recall tag vector corresponding to a recall tag in the recall tag set; the method comprises the steps that a tag vector is obtained from a vector set, and the vector set is generated according to a video tag which has a co-occurrence relationship and replaces a synonym with a specified synonym corresponding to the synonym;
The vector summation module 403 is configured to sum the target tag vectors to obtain a target tag set vector, and sum the tag vectors to be recalled to obtain a tag set vector to be recalled;
a similarity calculating module 404, configured to calculate a similarity between the target tag set vector and the to-be-recalled tag set vector;
and the video recall module 405 is configured to recall a video from the to-be-recalled videos corresponding to the to-be-recalled tag set vector whose similarity satisfies a preset condition.
In an exemplary embodiment of the present invention, the tag vector obtaining module 402 is configured to obtain a vector set; the vector set comprises label vectors corresponding to the video labels; obtaining a target tag vector corresponding to a target tag in the target tag set from the vector set, and obtaining a to-be-recalled tag vector corresponding to a to-be-recalled tag in the to-be-recalled tag set from the vector set.
In an exemplary embodiment of the invention, the apparatus further comprises: a vector set generating module; the vector set generating module is used for acquiring video tags with co-occurrence relations, wherein the co-occurrence relations are used for representing the probability of the simultaneous occurrence of the video tags; when the synonyms exist in the video tags, replacing the synonyms in the video tags with specified synonyms corresponding to the synonyms; and inputting a preset word vector model to obtain a corresponding tag vector by adopting the video tag after replacing the synonym, and storing the tag vector into a vector set.
In an exemplary embodiment of the present invention, the vector set generating module is configured to replace, when a matching synonym exists in a preset synonym table in the video tag, the synonym in the video tag with a specified synonym corresponding to the synonym.
In an exemplary embodiment of the present invention, the tag vector obtaining module 402 is configured to, when a target tag vector corresponding to a target tag in the target tag set does not exist in the vector set, segment the target tag to obtain a sub-word, obtain a target sub-tag vector corresponding to the sub-word of the target tag from the vector set, sum the target sub-tag vectors to obtain a target tag vector, and when a to-be-recalled tag vector corresponding to the to-be-recalled tag in the to-be-recalled tag set does not exist in the vector set, segment the to-be-recalled tag to obtain a sub-word, obtain a to-be-recalled sub-tag vector corresponding to the sub-word of the to-be-recalled tag from the vector set, and sum the to-be-recalled sub-tag vectors to obtain a to-be-recalled tag vector.
In an exemplary embodiment of the present invention, the video recall module 405 is configured to obtain a target to-be-recalled tag set corresponding to the to-be-recalled tag set vector that has the similarity greater than a preset similarity threshold and is ranked in a preceding N, where N is a positive integer; the target to-be-recalled tag set comprises a target tag set of the target video; creating a similar video index list according to the target to-be-recalled tag set; acquiring the associated video to be recalled as a target video to be recalled according to the similar video index list; determining similar videos from the target video to be recalled based on a preset posterior index; the posterior indicator includes click rate.
In an exemplary embodiment of the present invention, the video recall module 405 is configured to obtain the video to be recalled associated with the target set of tags to be recalled, and the video to be recalled having the same channel identification as the channel identification of the target video is taken as the target video to be recalled.
In the embodiment of the invention, the target label set of the target video and the to-be-recalled label set of the to-be-recalled video are obtained, the target label vector corresponding to the target label in the target label set and the to-be-recalled label vector corresponding to the to-be-recalled label in the to-be-recalled label set are obtained, then the target label vector is summed to obtain the target label set vector, and the to-be-recalled label set vector is summed to obtain the to-be-recalled label set vector, the similarity between the target label set vector and the to-be-recalled label set vector is calculated, and the to-be-recalled video corresponding to the to-be-recalled label set vector with the label set as a retrieval unit is used for recall of the video, so that the correlation between the recall video and the target video can be improved, and the quality of the recall video is improved.
For the above-described device embodiments, the description is relatively simple, as it is substantially similar to the method embodiments, with reference to the description of the method embodiments in part.
The embodiment of the invention also provides an electronic device, as shown in fig. 5, which comprises a processor 501, a communication interface 502, a memory 503 and a communication bus 504, wherein the processor 501, the communication interface 502 and the memory 503 complete communication with each other through the communication bus 504,
a memory 503 for storing a computer program;
the processor 501 is configured to execute the program stored in the memory 503, and implement the following steps:
acquiring a target tag set of a target video and a recall tag set of a recall video;
obtaining a target tag vector corresponding to a target tag in the target tag set, and obtaining a to-be-recalled tag vector corresponding to a to-be-recalled tag in the to-be-recalled tag set; the method comprises the steps that a tag vector is obtained from a vector set, and the vector set is generated according to a video tag which has a co-occurrence relationship and replaces a synonym with a specified synonym corresponding to the synonym;
summing the target tag vectors to obtain target tag set vectors, and summing the tag vectors to be recalled to obtain tag set vectors to be recalled;
calculating the similarity between the target tag set vector and the tag set vector to be recalled;
And recalling videos from the videos to be recalled corresponding to the label set vectors to be recalled, wherein the similarity of the videos to be recalled meets a preset condition.
Optionally, the obtaining the target tag vector corresponding to the target tag in the target tag set and obtaining the recall-to-be-obtained tag vector corresponding to the recall-to-be-obtained tag in the recall-to-be-obtained tag set includes:
acquiring a vector set; the vector set comprises label vectors corresponding to the video labels;
obtaining a target tag vector corresponding to a target tag in the target tag set from the vector set, and obtaining a to-be-recalled tag vector corresponding to a to-be-recalled tag in the to-be-recalled tag set from the vector set.
Optionally, before the acquiring the set of vectors, the method further comprises:
acquiring video tags with co-occurrence relations, wherein the co-occurrence relations are used for representing the probability of the simultaneous occurrence of the video tags;
when the synonyms exist in the video tags, replacing the synonyms in the video tags with specified synonyms corresponding to the synonyms;
and inputting a preset word vector model to obtain a corresponding tag vector by adopting the video tag after replacing the synonym, and storing the tag vector into a vector set.
Optionally, when the synonym exists in the video tag, replacing the synonym in the video tag with the specified synonym corresponding to the synonym includes:
and when the matched synonyms exist in the preset synonym table in the video tag, replacing the synonyms in the video tag with specified synonyms corresponding to the synonyms.
Optionally, the obtaining, from the vector set, a target tag vector corresponding to a target tag in the target tag set, and obtaining, from the vector set, a recall-to-tag vector corresponding to a recall-to-tag in the recall-to-tag set, includes:
when the target label vector corresponding to the target label in the target label set does not exist in the vector set, the target label is segmented to obtain a sub-word, the target sub-label vector corresponding to the sub-word of the target label is obtained from the vector set, the target sub-label vector is summed to obtain a target label vector, and
when the recall-to-be-labeled vector corresponding to the recall-to-be-labeled in the recall-to-be-labeled set does not exist in the vector set, word segmentation is carried out on the recall-to-be-labeled to obtain sub words, recall-to-be-labeled sub-vectors corresponding to the sub words of the recall-to-be-labeled are obtained from the vector set, and summation is carried out on the recall-to-be-labeled sub-vectors to obtain the recall-to-be-labeled vector.
Optionally, the recalling the video from the to-be-recalled videos corresponding to the to-be-recalled tag set vector with the similarity meeting a preset condition includes:
acquiring a target to-be-recalled tag set which is larger than a preset similarity threshold and corresponds to the to-be-recalled tag set vector of the previous N in sequence, wherein N is a positive integer; the target to-be-recalled tag set comprises a target tag set of the target video;
creating a similar video index list according to the target to-be-recalled tag set;
acquiring the associated video to be recalled as a target video to be recalled according to the similar video index list;
determining similar videos from the target video to be recalled based on a preset posterior index; the posterior indicator includes click rate.
Optionally, the target video and the video to be recalled have channel identifications, and the obtaining the video to be recalled associated with the target tag set to be recalled as the target video to be recalled includes:
and acquiring the video to be recalled, which is associated with the target video to be recalled tag set and has the same channel identification as the channel identification of the target video, as a target video to be recalled.
The communication bus mentioned by the above terminal may be a peripheral component interconnect standard (Peripheral Component Interconnect, abbreviated as PCI) bus or an extended industry standard architecture (Extended Industry Standard Architecture, abbreviated as EISA) bus, etc. The communication bus may be classified as an address bus, a data bus, a control bus, or the like. For ease of illustration, the figures are shown with only one bold line, but not with only one bus or one type of bus.
The communication interface is used for communication between the terminal and other devices.
The memory may include random access memory (Random Access Memory, RAM) or non-volatile memory (non-volatile memory), such as at least one disk memory. Optionally, the memory may also be at least one memory device located remotely from the aforementioned processor.
The processor may be a general-purpose processor, including a central processing unit (Central Processing Unit, CPU), a Network Processor (NP), etc.; but also digital signal processors (Digital Signal Processing, DSP for short), application specific integrated circuits (Application Specific Integrated Circuit, ASIC for short), field-programmable gate arrays (Field-Programmable Gate Array, FPGA for short) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
In yet another embodiment of the present invention, a computer readable storage medium having instructions stored therein that, when executed on a computer, cause the computer to perform the video recall method of any of the above embodiments is also provided.
In yet another embodiment of the present invention, there is also provided a computer program product containing instructions that, when run on a computer, cause the computer to perform the video recall method of any of the above embodiments.
In the above embodiments, it may be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed on a computer, produces a flow or function in accordance with embodiments of the present invention, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions may be stored in or transmitted from one computer-readable storage medium to another, for example, by wired (e.g., coaxial cable, optical fiber, digital Subscriber Line (DSL)), or wireless (e.g., infrared, wireless, microwave, etc.). The computer readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. that contains an integration of one or more available media. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid State Disk (SSD)), etc.
It is noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
In this specification, each embodiment is described in a related manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments. In particular, for system embodiments, since they are substantially similar to method embodiments, the description is relatively simple, as relevant to see a section of the description of method embodiments.
The foregoing description is only of the preferred embodiments of the present invention and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.
Claims (9)
1. A video recall method, the method comprising:
acquiring a target tag set of a target video and a recall tag set of a recall video;
obtaining a target tag vector corresponding to a target tag in the target tag set, and obtaining a to-be-recalled tag vector corresponding to a to-be-recalled tag in the to-be-recalled tag set; the method comprises the steps that a tag vector is obtained from a vector set, and the vector set is generated according to a video tag which has a co-occurrence relationship and replaces a synonym with a specified synonym corresponding to the synonym; when a target label vector corresponding to a target label in the target label set does not exist in the vector set, word segmentation is carried out on the target label to obtain a sub-word, a target sub-label vector corresponding to the sub-word of the target label is obtained from the vector set, summation is carried out on the target sub-label vector to obtain a target label vector, and when a to-be-recalled label vector corresponding to the to-be-recalled label in the to-be-recalled label set does not exist in the vector set, word segmentation is carried out on the to-be-recalled label to obtain a sub-word, a to-be-recalled sub-label vector corresponding to the sub-word of the to-be-recalled label is obtained from the vector set, and summation is carried out on the to-be-recalled sub-label vector to obtain a to-be-recalled label vector;
Summing the target tag vectors to obtain target tag set vectors, and summing the tag vectors to be recalled to obtain tag set vectors to be recalled;
calculating the similarity between the target tag set vector and the tag set vector to be recalled;
and recalling videos from the videos to be recalled corresponding to the label set vectors to be recalled, wherein the similarity of the videos to be recalled meets a preset condition.
2. The method of claim 1, wherein the obtaining a target tag vector corresponding to a target tag in the target tag set and obtaining a recall tag vector corresponding to a recall tag in the recall tag set comprise:
acquiring a vector set; the vector set comprises label vectors corresponding to the video labels;
obtaining a target tag vector corresponding to a target tag in the target tag set from the vector set, and obtaining a to-be-recalled tag vector corresponding to a to-be-recalled tag in the to-be-recalled tag set from the vector set.
3. The method of claim 2, wherein prior to the obtaining the set of vectors, the method further comprises:
Acquiring video tags with co-occurrence relations, wherein the co-occurrence relations are used for representing the probability of the simultaneous occurrence of the video tags;
when the synonyms exist in the video tags, replacing the synonyms in the video tags with specified synonyms corresponding to the synonyms;
and inputting a preset word vector model to obtain a corresponding tag vector by adopting the video tag after replacing the synonym, and storing the tag vector into a vector set.
4. The method of claim 3, wherein when there is a synonym in the video tag, replacing the synonym in the video tag with a specified synonym corresponding to the synonym comprises:
and when the matched synonyms exist in the preset synonym table in the video tag, replacing the synonyms in the video tag with specified synonyms corresponding to the synonyms.
5. The method according to claim 2, wherein recalling video from the to-be-recalled videos corresponding to the to-be-recalled tag set vector whose similarity satisfies a preset condition, comprises:
acquiring a target to-be-recalled tag set which is larger than a preset similarity threshold and corresponds to the to-be-recalled tag set vector of the previous N in sequence, wherein N is a positive integer; the target to-be-recalled tag set comprises a target tag set of the target video;
Creating a similar video index list according to the target to-be-recalled tag set;
acquiring the associated video to be recalled as a target video to be recalled according to the similar video index list;
determining similar videos from the target video to be recalled based on a preset posterior index; the posterior indicator includes click rate.
6. The method of claim 5, wherein the target video and the video to recall have channel identifications, the obtaining video to recall associated with the target set of tags to recall as target video to recall comprising:
and acquiring the video to be recalled, which is associated with the target video to be recalled tag set and has the same channel identification as the channel identification of the target video, as a target video to be recalled.
7. A video recall device, the device comprising:
the tag set acquisition module is used for acquiring a target tag set of a target video and a to-be-recalled tag set of a to-be-recalled video;
the tag vector acquisition module is used for acquiring a target tag vector corresponding to a target tag in the target tag set and acquiring a to-be-recalled tag vector corresponding to a to-be-recalled tag in the to-be-recalled tag set; the method comprises the steps that a tag vector is obtained from a vector set, and the vector set is generated according to a video tag which has a co-occurrence relationship and replaces a synonym with a specified synonym corresponding to the synonym; when a target label vector corresponding to a target label in the target label set does not exist in the vector set, word segmentation is carried out on the target label to obtain a sub-word, a target sub-label vector corresponding to the sub-word of the target label is obtained from the vector set, summation is carried out on the target sub-label vector to obtain a target label vector, and when a to-be-recalled label vector corresponding to the to-be-recalled label in the to-be-recalled label set does not exist in the vector set, word segmentation is carried out on the to-be-recalled label to obtain a sub-word, a to-be-recalled sub-label vector corresponding to the sub-word of the to-be-recalled label is obtained from the vector set, and summation is carried out on the to-be-recalled sub-label vector to obtain a to-be-recalled label vector;
The vector summation module is used for summing the target tag vectors to obtain target tag set vectors, and summing the tag vectors to be recalled to obtain tag set vectors to be recalled;
the similarity calculation module is used for calculating the similarity between the target tag set vector and the to-be-recalled tag set vector;
and the video recall module is used for recalling videos from the videos to be recalled corresponding to the label set vectors to be recalled, wherein the similarity of the videos to be recalled meets a preset condition.
8. The electronic equipment is characterized by comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory are communicated with each other through the communication bus;
a memory for storing a computer program;
a processor for carrying out the method steps of any one of claims 1-6 when executing a program stored on a memory.
9. A computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the method according to any of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110156410.5A CN112989118B (en) | 2021-02-04 | 2021-02-04 | Video recall method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110156410.5A CN112989118B (en) | 2021-02-04 | 2021-02-04 | Video recall method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112989118A CN112989118A (en) | 2021-06-18 |
| CN112989118B true CN112989118B (en) | 2023-08-18 |
Family
ID=76347134
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110156410.5A Active CN112989118B (en) | 2021-02-04 | 2021-02-04 | Video recall method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112989118B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113392266B (en) * | 2021-08-17 | 2021-12-14 | 北京达佳互联信息技术有限公司 | Training and sorting method and device of sorting model, electronic equipment and storage medium |
| CN116915925B (en) * | 2023-06-20 | 2024-02-23 | 天翼爱音乐文化科技有限公司 | Video generation method, system, electronic equipment and medium based on video template |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2006139484A (en) * | 2004-11-11 | 2006-06-01 | Nippon Telegr & Teleph Corp <Ntt> | Information retrieval method, system therefor and computer program |
| CN103279513A (en) * | 2013-05-22 | 2013-09-04 | 百度在线网络技术（北京）有限公司 | Method for generating content label and method and device for providing multi-media content information |
| CN103488787A (en) * | 2013-09-30 | 2014-01-01 | 北京奇虎科技有限公司 | Method and device for pushing online playing entry objects based on video retrieval |
| CN107122413A (en) * | 2017-03-31 | 2017-09-01 | 北京奇艺世纪科技有限公司 | A kind of keyword extracting method and device based on graph model |
| CN108228665A (en) * | 2016-12-22 | 2018-06-29 | 阿里巴巴集团控股有限公司 | Determine object tag, the method and device for establishing tab indexes, object search |
| CN109587554A (en) * | 2018-10-29 | 2019-04-05 | 百度在线网络技术（北京）有限公司 | Processing method, device and the readable storage medium storing program for executing of video data |
| CN109635157A (en) * | 2018-10-30 | 2019-04-16 | 北京奇艺世纪科技有限公司 | Model generating method, video searching method, device, terminal and storage medium |
| CN110457581A (en) * | 2019-08-02 | 2019-11-15 | 达而观信息科技（上海）有限公司 | A kind of information recommended method, device, electronic equipment and storage medium |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8943070B2 (en) * | 2010-07-16 | 2015-01-27 | International Business Machines Corporation | Adaptive and personalized tag recommendation |
| US20120323968A1 (en) * | 2011-06-14 | 2012-12-20 | Microsoft Corporation | Learning Discriminative Projections for Text Similarity Measures |
| US20130036061A1 (en) * | 2011-08-02 | 2013-02-07 | International Business Machines Corporation | Product recall information management |
| US9998772B2 (en) * | 2015-07-28 | 2018-06-12 | Google Llc | Methods, systems, and media for presenting media content items belonging to a media content group |
| US9589237B1 (en) * | 2015-11-17 | 2017-03-07 | Spotify Ab | Systems, methods and computer products for recommending media suitable for a designated activity |
-
2021
- 2021-02-04 CN CN202110156410.5A patent/CN112989118B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2006139484A (en) * | 2004-11-11 | 2006-06-01 | Nippon Telegr & Teleph Corp <Ntt> | Information retrieval method, system therefor and computer program |
| CN103279513A (en) * | 2013-05-22 | 2013-09-04 | 百度在线网络技术（北京）有限公司 | Method for generating content label and method and device for providing multi-media content information |
| CN103488787A (en) * | 2013-09-30 | 2014-01-01 | 北京奇虎科技有限公司 | Method and device for pushing online playing entry objects based on video retrieval |
| CN108228665A (en) * | 2016-12-22 | 2018-06-29 | 阿里巴巴集团控股有限公司 | Determine object tag, the method and device for establishing tab indexes, object search |
| CN107122413A (en) * | 2017-03-31 | 2017-09-01 | 北京奇艺世纪科技有限公司 | A kind of keyword extracting method and device based on graph model |
| CN109587554A (en) * | 2018-10-29 | 2019-04-05 | 百度在线网络技术（北京）有限公司 | Processing method, device and the readable storage medium storing program for executing of video data |
| CN109635157A (en) * | 2018-10-30 | 2019-04-16 | 北京奇艺世纪科技有限公司 | Model generating method, video searching method, device, terminal and storage medium |
| CN110457581A (en) * | 2019-08-02 | 2019-11-15 | 达而观信息科技（上海）有限公司 | A kind of information recommended method, device, electronic equipment and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 基于向量化标签的视频推荐算法研究与实现;许良武;《无线互联科技》;第132-134、165页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112989118A (en) | 2021-06-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105224699B (en) | News recommendation method and device | |
| CN112347778B (en) | Keyword extraction method, keyword extraction device, terminal equipment and storage medium | |
| CN109189990B (en) | Search word generation method and device and electronic equipment | |
| CN112434151A (en) | Patent recommendation method and device, computer equipment and storage medium | |
| CN112989118B (en) | Video recall method and device | |
| Lin et al. | Mitigating sentiment bias for recommender systems | |
| US11574123B2 (en) | Content analysis utilizing general knowledge base | |
| CN112464100B (en) | Information recommendation model training method, information recommendation method, device and equipment | |
| Setty et al. | Truth be told: Fake news detection using user reactions on reddit | |
| CN111639696B (en) | User classification method and device | |
| Kalloubi et al. | Harnessing semantic features for large-scale content-based hashtag recommendations on microblogging platforms | |
| Guy et al. | Identifying informational vs. conversational questions on community question answering archives | |
| US20220164546A1 (en) | Machine Learning Systems and Methods for Many-Hop Fact Extraction and Claim Verification | |
| Nasir et al. | Semantic enhanced Markov model for sequential E-commerce product recommendation | |
| Liu et al. | Question popularity analysis and prediction in community question answering services | |
| CN112836126A (en) | Recommendation method and device based on knowledge graph, electronic equipment and storage medium | |
| Loizou | How to recommend music to film buffs: enabling the provision of recommendations from multiple domains | |
| BR et al. | Corpus-based Topic Derivation and Timestamp-based Popular Hashtag Prediction in Twitter. | |
| CN111984867B (en) | Network resource determining method and device | |
| Kalloubi | Learning to suggest hashtags: Leveraging semantic features for time-sensitive hashtag recommendation on the Twitter network | |
| Martina et al. | A Virtual Assistant for the Movie Domain Exploiting Natural Language Preference Elicitation Strategies | |
| Tibély et al. | Comparing the hierarchy of keywords in on-line news portals | |
| Chaudhary et al. | Fake News Detection During 2016 US Elections Using Bootstrapped Metadata-Based Naïve Bayesian Classifier | |
| Sun et al. | Leveraging user profiling in click-through rate prediction based on Zhihu data | |
| Santosa et al. | S3PaR: Section-Based Sequential Scientific Paper Recommendation for Paper Writing Assistance |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |