Label aggregation method and device based on artificial intelligence, and storage medium

Abstract

The present disclosure relates to a tag aggregation method and device based on artificial intelligence, and a storage medium. In the scheme of the present disclosure, a neural network for obtaining the probability of tags of content sources is used to obtain the probability of one or more tags corresponding to each of multiple first content sources; obtain the first tag and the second tag with the highest probability among the probabilities of one or more tags corresponding to each of the first content sources; obtain the aggregation degree of the first tag and the second tag; and if the aggregation degree of the first tag and the second tag is greater than a first set threshold, aggregate the first tag and the second tag to obtain a third tag. Using the scheme of the present disclosure, by aggregating two tags corresponding to multiple content sources whose aggregation degree is greater than a set threshold, tag aggregation is achieved, which can make the tag system of the content source more complete and accurate.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based tag aggregation method and device, and a storage medium. Background Art

[0002] Tags are very important for the first content source, such as video tags, article tags, picture tags, music tags, etc. Tags can be automatically mined through the associated first content source, or they can be manually annotated.

[0003] In many scenarios, due to technical and human limitations, the first content source has only one label. For example, a video can be either "funny" or "food"; a news article can be either "political" or "domestic". However, in reality, since labeling is relatively complex, these labels are not independent of each other. For example, a video can be both "funny" and "food", and news can be both "domestic" and "military". Therefore, how to aggregate these labels to make the labels corresponding to the first content source more accurate is an important issue.

[0004] For example, if we find that "funny" and "food" can be combined into a new tag "funny#food" to represent certain videos, then we can add the tag "funny#food" to our new tag system. The tags of videos that are only funny or only about food remain unchanged, while videos that are both funny and related to food can be tagged with "funny#food".

[0005] But there is no solution for tag aggregation yet. Summary of the invention

[0006] The present disclosure provides a solution for label aggregation based on artificial intelligence.

[0007] In a first aspect, a tag aggregation method based on artificial intelligence is provided, comprising:

[0008] Using a neural network for obtaining the probability of labels of content sources, obtaining the probability of one or more labels corresponding to each first content source in the plurality of first content sources;

[0009] Obtaining a first tag and a second tag with the highest probability among the probabilities of the one or more tags corresponding to each first content source;

[0010] Obtaining a degree of aggregation of the first tag and the second tag;

[0011] If the aggregation degree of the first tag and the second tag is greater than a first set threshold, the first tag and the second tag are aggregated to obtain a third tag.

[0012] In one implementation, obtaining the degree of aggregation of the first tag and the second tag includes:

[0013] Acquire a first number of first content sources corresponding to the first tag and the second tag among the plurality of first content sources, a second number of first content sources corresponding to the first tag among the plurality of first content sources, and a third number of first content sources corresponding to the second tag among the plurality of first content sources;

[0014] The degree of aggregation of the first tag and the second tag is determined according to the first number, the second number, and the third number.

[0015] In yet another implementation, the method further includes:

[0016] The third tag is updated to the tag set of the first content source.

[0017] In yet another implementation, the method further includes:

[0018] The plurality of first content sources are labeled using the third tag.

[0019] In yet another implementation, the method further includes:

[0020] Using a neural network for obtaining the probability of labels of content sources, obtaining the probability of one or more labels corresponding to each second content source in the plurality of second content sources;

[0021] Obtaining a third label and a fourth label with the highest probability among the probabilities of the one or more labels corresponding to each second content source;

[0022] Obtaining a degree of aggregation of the third tag and the fourth tag;

[0023] If the degree of aggregation of the third tag and the fourth tag is greater than the first set threshold, the third tag and the fourth tag are aggregated to obtain a fifth tag;

[0024] If the aggregation degree of the third tag and the fourth tag is less than or equal to the first set threshold, the third tag is not aggregated.

[0025] In yet another implementation, the method further includes:

[0026] Receive a query request, wherein the query request includes a tag corresponding to a content source requested for query;

[0027] Finding a content source corresponding to the tag;

[0028] The found content source is returned.

[0029] In yet another implementation, the method further includes:

[0030] Acquire user preference information, where the preference information corresponds to one or more tags;

[0031] determining whether the label of the item includes one or more labels corresponding to the preference information;

[0032] If the tag of the item includes any tag corresponding to the preference information, the item is recommended to the user.

[0033] In a second aspect, a label aggregation device based on artificial intelligence is provided, comprising:

[0034] A first acquisition unit, configured to acquire a probability of one or more labels corresponding to each of the plurality of first content sources by using a neural network for acquiring probabilities of labels of content sources;

[0035] A second acquisition unit, configured to acquire a first label and a second label with the highest probability among the probabilities of the one or more labels corresponding to each first content source;

[0036] A third acquisition unit, configured to acquire a degree of aggregation of the first tag and the second tag;

[0037] An aggregation unit is used to aggregate the first tag and the second tag to obtain a third tag if the aggregation degree of the first tag and the second tag is greater than a first set threshold.

[0038] In one implementation, the third acquisition unit includes:

[0039] a fourth acquiring unit, configured to acquire a first number of first content sources corresponding to the first tag and the second tag from among the plurality of first content sources, a second number of first content sources corresponding to the first tag from among the plurality of first content sources, and a third number of first content sources corresponding to the second tag from among the plurality of first content sources;

[0040] The first determining unit is configured to determine a degree of aggregation of the first tag and the second tag according to the first number, the second number, and the third number.

[0041] In yet another implementation, the apparatus further includes:

[0042] An updating unit is configured to update the third tag to a tag set of the first content source.

[0043] In yet another implementation, the apparatus further includes:

[0044] A labeling unit is used to label the multiple first content sources using the third label.

[0045] In yet another implementation, the first acquisition unit is further configured to use a neural network for acquiring probabilities of labels of content sources to acquire probabilities of one or more labels corresponding to each of the plurality of second content sources;

[0046] The second acquisition unit is further used to acquire a third label and a fourth label with the highest probability among the probabilities of the one or more labels corresponding to each second content source;

[0047] The third acquisition unit is further used to acquire the degree of aggregation of the third tag and the fourth tag;

[0048] The aggregation unit is further configured to aggregate the third tag and the fourth tag to obtain a fifth tag if the aggregation degree of the third tag and the fourth tag is greater than the first set threshold;

[0049] The aggregation unit is further configured to not aggregate the third tag if the aggregation degree of the third tag and the fourth tag is less than or equal to the first set threshold.

[0050] In yet another implementation, the apparatus further includes:

[0051] A receiving unit, configured to receive a query request, wherein the query request includes a tag corresponding to a content source requested for query;

[0052] A search unit, configured to search for a content source corresponding to the tag;

[0053] The returning unit is used to return the found content source.

[0054] In yet another implementation, the apparatus further includes:

[0055] a fifth acquiring unit, configured to acquire user preference information, wherein the preference information corresponds to one or more tags;

[0056] a second determining unit, configured to determine whether the label of the item includes one or more labels corresponding to the preference information;

[0057] A recommendation unit is used to recommend the item to the user if the tag of the item includes any tag corresponding to the preference information.

[0058] In a third aspect, an artificial intelligence-based tag aggregation device is provided, comprising: a processor and a memory, wherein program instructions are stored in the memory, and the processor executes the program instructions to implement the method described in the first aspect or any one of the implementations of the first aspect.

[0059] According to a fourth aspect, a computer storage medium is provided, wherein the computer storage medium stores one or more instructions, wherein the one or more instructions are suitable for being loaded by a processor and executing the method described in the first aspect or any implementation of the first aspect.

[0060] The solution disclosed in this disclosure has the following beneficial effects:

[0061] By aggregating two tags corresponding to multiple content sources whose aggregation degree is greater than a set threshold, tag aggregation is achieved, which can make the tag system of the content source more complete and accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0063] Figure 1 is a schematic diagram of a system architecture of tag aggregation based on artificial intelligence provided by an embodiment of the present disclosure;

[0064] Figure 2 It is a flowchart of a tag aggregation method based on artificial intelligence provided by an embodiment of the present disclosure;

[0065] Figure 3 It is a flowchart of another tag aggregation method based on artificial intelligence provided by an embodiment of the present disclosure;

[0066] Figure 4 is a flow chart of using aggregated tags for content source search according to an embodiment of the present disclosure;

[0067] Figure 5 is a schematic diagram of a process of using aggregated tags for item recommendation according to an embodiment of the present disclosure;

[0068] Figure 6 It is a structural schematic diagram of an artificial intelligence-based tag aggregation device provided in an embodiment of the present disclosure;

[0069] Figure 7 It is a structural diagram of another artificial intelligence-based label aggregation device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0070] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0071] See also Figure 1 , is a schematic diagram of a system architecture of label aggregation based on artificial intelligence provided by an embodiment of the present disclosure. The embodiment of the present disclosure uses a trained multi-classification model to extract labels corresponding to each content source from each of the multiple content sources. The multi-classification model is obtained by training a plurality of content source samples whose labels are known to determine the labels that can be aggregated. Therefore, the multi-classification model can be used to extract labels corresponding to the content source to be predicted. Figure 1 As shown in FIG. 1 , the multi-classification model extracts multiple tags, selects the two tags with the highest probability, and determines the aggregation degree of the two selected tags. When the aggregation degree is greater than the set threshold, the two tags can be aggregated to update the tag system. The above process can be repeated until the tags of the content source cannot be aggregated. Thus, the aggregation of tags is achieved, which can make the tag system of the content source more complete and accurate.

[0072] See also Figure 2 , is a flow chart of a tag aggregation method based on artificial intelligence provided by an embodiment of the present disclosure, and the method may include the following steps:

[0073] S101. Use a neural network for obtaining the probability of labels of content sources to obtain the probability of one or more labels corresponding to each first content source in a plurality of first content sources.

[0074] Specifically, the training process of the multi-classification model is as follows:

[0075] First, there is a batch of content sources, each with only one label, and then use this batch of content sources to train a multi-classification model. For example, suppose the current label system is {food, funny, military, politics, domestic, games, religion}. After a round of iterations, it is found that "food" and "funny" can be used as labels for some content sources, so "food" and "funny" are aggregated; similarly, "military" and "politics", "military" and "domestic", "politics" and "domestic" can all be aggregated. After a round of iterations, the new label system becomes: {food, funny, military, politics, domestic, games, religion, food#funny, military#politics, military#domestic, politics#domestic, games#funny}. "Religion" is not aggregated because it is relatively independent. No video is related to both religion and other labels. Note that this round of label aggregation process only works on two labels.

[0076] After the tag system is updated, the tags corresponding to the content source can also be updated. For example, some videos were originally tagged "game", but they are also "funny", so the tag will be updated to "game#funny". Then, continue with the second round of iterations, you may find that "military#politics" and "domestic" can also be aggregated, thus becoming "military#politics#domestic"; it is also possible that "military#politics" and "politics#domestic" can be aggregated, also becoming "military#politics#domestic" (when naming new tags, duplicates will be removed); it may also be found that "game#funny" and "domestic" are also related, aggregated into a new tag "game#funny#domestic". Add these new tags to the tag system and update the tags of the content source corresponding to the new tags. Until no new tags can be found. Note that the method of "tag1#tag2#tag3#tag4..." is used here to name the aggregated tags, and other methods can also be used as long as the new tag can be guaranteed to be represented by the basic tag. Basic tags refer to tags in the original system. The naming of new tags must ensure that there are no repeated basic tags. For example, the result of the aggregation of "military#politics" and "politics#domestic" mentioned above is the same as the result of the aggregation of "military#politics" and "domestic", which is "military#politics#domestic". The former removes the repeated basic tag: politics.

[0077] The above multi-classification model can not only extract the label corresponding to each content source, but also obtain the probability of the label corresponding to each content source. That is, the multi-classification model can be a neural network for obtaining the probability of the label of the content source.

[0078] A content source set includes multiple content sources. Specifically, it includes labeled content sources for training, and also includes unlabeled content sources to be predicted. The multiple first content sources can be part or all of the unlabeled content sources to be predicted in the content source set. For the predicted content source, its real label is also known, but the above-mentioned multi-classification model will also predict a probability distribution corresponding to the probability that the content belongs to different labels. Therefore, each first content source may correspond to one or more labels, and the above-mentioned neural network for obtaining the probability of the label of the content source is used to obtain the probability of one or more labels corresponding to each content source.

[0079] S102: Obtain a first tag and a second tag with the highest probability among the probabilities of one or more tags corresponding to each first content source.

[0080] Generally, a content has only one label, and the maximum probability is used as the prediction value. However, some content can actually be labeled with two labels, so it is necessary to filter out those that can be labeled with two labels. The screening method is: assuming that the two most likely labels of the first content source are A and B, and the corresponding probabilities are Pa and Pb, then for the first content source that can be labeled with two labels, Pa and Pb should be as close as possible, and Pa and Pb should be significantly greater than the probabilities of other categories. Here, "close" and "greater than others" are expressed in mathematical language as follows:

[0081] Pa–Pb <p1

[0082] Pb–Pc>p2

[0083] Wherein, Pc is the probability of other labels corresponding to the first content source except Pa and Pb. The sizes of p1 and p2 are artificially set. Generally, p1 is as small as possible and p2 is as large as possible.

[0084] For example, the probability that a first content source corresponds to the tag "food" is 100%; but the first content source also has the element of "funny", so the probability of obtaining the tag "funny" corresponding to the first content source is 80%; the probability that the first content source corresponds to other tags is less than 50%. The two tags with the highest probability of obtaining are "food" and "funny".

[0085] S103: Obtain the aggregation degree of the first tag and the second tag.

[0086] Through this method, we can get a batch of content sources that can be labeled with two labels (these content sources originally corresponded to only one label). However, it does not mean that as long as a content source can be labeled with two labels, the two labels can be aggregated together. For example, if only one content source among multiple content sources can correspond to the two labels of "religion" and "funny" at the same time, and the other content sources cannot, then the two content sources cannot be aggregated together. Therefore, it is necessary to judge by the degree of aggregation. It is necessary to obtain the degree of aggregation of the above-mentioned first label and the second label. That is, to obtain how many content sources among the above-mentioned multiple content sources can correspond to the first label and the second label at the same time.

[0087] S104: If the aggregation degree of the first tag and the second tag is greater than a first set threshold, aggregate the first tag and the second tag to obtain a third tag.

[0088] If the aggregation degree of the first tag and the second tag is greater than the first set threshold, it means that multiple content sources can correspond to the first tag and the second tag at the same time. Then, the first tag and the second tag can be aggregated to obtain a new third tag, thereby improving the tag system of the content source.

[0089] According to an artificial intelligence-based tag aggregation method provided by an embodiment of the present disclosure, tag aggregation is achieved by aggregating two tags corresponding to multiple content sources whose aggregation degrees are greater than a set threshold, which can make the tag system of the content source more complete and accurate.

[0090] See also Figure 3 , is a flow chart of another method for label aggregation based on artificial intelligence provided by an embodiment of the present disclosure, which method may include the following steps:

[0091] S201. Use a neural network for obtaining the probability of labels of content sources to obtain the probability of one or more labels corresponding to each first content source in a plurality of first content sources.

[0092] The specific implementation of this step can be found in Figure 2 Step S101 of the illustrated embodiment.

[0093] S202: Obtain a first tag and a second tag with the highest probability among the probabilities of one or more tags corresponding to each first content source.

[0094] The specific implementation of this step can be found in Figure 2 Step S102 of the illustrated embodiment.

[0095] The following steps S203 to S204 are for obtaining the degree of aggregation of the first tag and the second tag.

[0096] S203: Obtain a first number of first content sources corresponding to the first tag and the second tag among the multiple first content sources, a second number of first content sources corresponding to the first tag among the multiple first content sources, and a third number of first content sources corresponding to the second tag among the multiple first content sources.

[0097] S204: Determine a degree of aggregation of the first tag and the second tag according to the first number, the second number, and the third number.

[0098] Assuming that the first label and the second label are A and B respectively, the degree of aggregation of A and B is expressed as:

[0099] score(A#B)=cnt(A#B) / cnt(A)+cnt(B)

[0100] Among them, cnt represents the number of content sources with this label. The first number - cnt(A#B) is the number of content sources that can be labeled with both A and B according to the above screening method. The second number - cnt(A) refers to the number of all content sources labeled with A in the original content source. Note that it is "originally", that is, even if it was originally A, it was screened to be labeled with both A and B, but it is still counted in cnt(A). Similarly, the third number - cnt(B) refers to the number of all content sources labeled with B in the original content source. Note that it is "originally", that is, even if it was originally B, it was screened to be labeled with both A and B, but it is still counted in cnt(B).

[0101] S205, determining whether the aggregation degree of the first tag and the second tag is greater than a first set threshold, if so, executing step S206; otherwise, not aggregating the first tag and the second tag, and ending the process.

[0102] Set a threshold s. As long as score(A#B)>s, we can consider that A and B can be aggregated into a new label: A#B. Here s is globally unique and has nothing to do with specific labels.

[0103] S206: Aggregate the first label and the second label to obtain a third label.

[0104] Through this method, a third label can be added after one round of iteration.

[0105] S207: Update the third tag to the tag set of the first content source.

[0106] For the filtered content sources that can be tagged with two tags, if the two tags can be aggregated, the tag set of the first content source is updated, otherwise it is not updated.

[0107] S208: Use the third tag to mark the multiple first content sources.

[0108] In this way, after one round of iteration, a third tag can be added, and then the tag of the content source can be updated.

[0109] The following steps S209 to S214 are to extract the labels of other content sources in the content source set or the content sources newly added to the content source set, obtain the probability of the label corresponding to the second content source, and determine whether the two labels with the highest probability can be aggregated. These two labels may include the third label obtained after aggregation.

[0110] S209: Using a neural network for obtaining the probability of labels of content sources, obtain the probability of one or more labels corresponding to each second content source in the plurality of second content sources.

[0111] S210: Obtain a third tag and a fourth tag with the highest probability among the probabilities of one or more tags corresponding to each second content source.

[0112] S211. Obtain the degree of aggregation of the third tag and the fourth tag.

[0113] S212: Determine whether the degree of aggregation of the third tag and the fourth tag is greater than the first set threshold. If so, execute step S213; otherwise, execute step S214.

[0114] S213: Aggregate the third label and the fourth label to obtain a fifth label.

[0115] Through the above steps, if the third label and the fourth label can be further aggregated, they are aggregated to obtain a fifth label.

[0116] For example, in the above example, the third label "food#funny" is obtained after aggregation. If the probability of obtaining "food#funny" in multiple second content sources is 90%, and the probability of obtaining "Hunan Province" is 80%, and the probabilities of these two labels are much greater than other labels, then "food#funny" and "Hunan Province" can be aggregated to obtain a new label of "food#funny#Hunan Province".

[0117] S214: If the aggregation degree of the third tag and the fourth tag is less than or equal to the first set threshold, the third tag is not aggregated.

[0118] If the aggregation degree between the third tag and any tag of the second content source is less than or equal to the first set threshold, it means that the third tag can no longer be aggregated with other tags, and the third tag is not aggregated.

[0119] According to an artificial intelligence-based tag aggregation method provided by an embodiment of the present disclosure, tag aggregation is achieved by aggregating two tags corresponding to multiple content sources whose aggregation degrees are greater than a set threshold, which can make the tag system of content sources more complete and accurate; different content source combinations of the content source set are respectively detected whether their tags can be aggregated, and the tag system will no longer be updated until there are no tags that can be aggregated.

[0120] See also Figure 4 , is a flow chart of using aggregated tags for content source search in an embodiment of the present disclosure. The method may include the following steps:

[0121] S301: Receive a query request, where the query request includes a tag corresponding to a content source requested for query.

[0122] The tag system is improved by the above tag aggregation, so when some tag-related query requests are input, the search engine can match more accurately. Specifically, a query request is received, and the query request includes a tag corresponding to the content source requested for query. For example, if a user searches for an article tagged "military", the query request carries the tag "military".

[0123] S302: Search for a content source corresponding to the tag.

[0124] After receiving the query request, the content source corresponding to the tag carried in the query request is searched. Specifically, the tags of multiple content sources can be obtained and then matched with the tag carried in the query request to obtain the content source corresponding to the tag carried in the query request.

[0125] For example, before tag aggregation, an article originally only had the tag "Domestic", so it could not be searched. After tag aggregation, it was tagged with "Domestic#Military", so when searching for content sources corresponding to the tag "Military", the article can also be searched. This improves search efficiency and accuracy.

[0126] S303: Return the found content source.

[0127] According to the method for content source search using aggregated tags provided by the embodiment of the present disclosure, search efficiency and accuracy are improved.

[0128] See also Figure 5, is a flow chart of using aggregated tags for item recommendation in an embodiment of the present disclosure. The method may include the following steps:

[0129] S401: Obtain user preference information, where the preference information corresponds to one or more tags.

[0130] When a user buys an item on a shopping website or other network, the user's preference information for the item can be counted. The preference information can be identified by a tag. The preference information corresponds to one or more tags. For example, the user's preference tag for clothing is "casual#solid color".

[0131] The preference information may also be referred to as a user portrait, that is, the user portrait is a description or characterization of the user's preference information.

[0132] S402: Determine whether the label of the item includes one or more labels corresponding to the preference information.

[0133] The tags of each item on the shopping website can also be obtained. Then, according to the tags corresponding to the user's preference information, it is determined whether the above items include one or more tags corresponding to the user's preference information.

[0134] S403: If the label of the item includes any label corresponding to the preference information, recommend the item to the user.

[0135] For example, if the item is clothing and the tag of the item includes “casual#solid color”, the item may be recommended to the user.

[0136] Of course, the user's preference information may correspond to multiple tags of multiple items. If the tags of the multiple items include any tag corresponding to the user's preference information, the item is recommended to the user.

[0137] According to the method for using aggregated tags for item recommendation provided by the embodiment of the present disclosure, the pertinence of item recommendation is improved.

[0138] Based on the same concept of the above-mentioned label aggregation method based on artificial intelligence, the embodiment of the present disclosure also provides a label aggregation device based on artificial intelligence. Figure 6 , which is a schematic diagram of the structure of an artificial intelligence-based tag aggregation device provided in an embodiment of the present disclosure, wherein the device 1000 includes:

[0139] A first acquisition unit 11 is used to acquire the probability of one or more labels corresponding to each of the plurality of first content sources by using a neural network for acquiring the probability of labels of content sources;

[0140] A second acquisition unit 12 is used to acquire a first label and a second label with the highest probability among the probabilities of one or more labels corresponding to each first content source;

[0141] A third acquisition unit 13 is used to acquire the aggregation degree of the first tag and the second tag;

[0142] The aggregation unit 14 is configured to aggregate the first tag and the second tag to obtain a third tag if the aggregation degree of the first tag and the second tag is greater than a first set threshold.

[0143] In one implementation, the third obtaining unit 13 includes:

[0144] a fourth acquiring unit 131, configured to acquire a first number of first content sources corresponding to the first tag and the second tag from among the plurality of first content sources, a second number of first content sources corresponding to the first tag from among the plurality of first content sources, and a third number of first content sources corresponding to the second tag from among the plurality of first content sources;

[0145] The first determining unit 132 is configured to determine a degree of aggregation of the first tag and the second tag according to the first number, the second number, and the third number.

[0146] In yet another implementation, the apparatus further includes:

[0147] The updating unit 15 is configured to update the third tag to the tag set of the first content source.

[0148] In yet another implementation, the apparatus further includes:

[0149] The labeling unit 16 is configured to label the plurality of first content sources using the third label.

[0150] In yet another implementation, the first acquisition unit 11 is further configured to use a neural network for acquiring probabilities of labels of content sources to acquire probabilities of one or more labels corresponding to each of the plurality of second content sources;

[0151] The second acquisition unit 12 is further configured to acquire a third label and a fourth label with the highest probability among the probabilities of the one or more labels corresponding to each second content source;

[0152] The third acquisition unit 13 is further used to acquire the degree of aggregation of the third tag and the fourth tag;

[0153] The aggregation unit 14 is further configured to aggregate the third tag and the fourth tag to obtain a fifth tag if the aggregation degree of the third tag and the fourth tag is greater than the first set threshold;

[0154] The aggregation unit 14 is further configured to not aggregate the third tag if the aggregation degree of the third tag and the fourth tag is less than or equal to the first set threshold.

[0155] In yet another implementation, the apparatus further includes:

[0156] A receiving unit 17, configured to receive a query request, wherein the query request includes a tag corresponding to a content source requested for query;

[0157] A search unit 18, configured to search for a content source corresponding to the tag;

[0158] The returning unit 19 is used to return the found content source.

[0159] In yet another implementation, the apparatus further includes:

[0160] a fifth acquiring unit, configured to acquire user preference information, wherein the preference information corresponds to one or more tags;

[0161] a second determining unit, configured to determine whether the label of the item includes one or more labels corresponding to the preference information;

[0162] A recommendation unit is used to recommend the item to the user if the tag of the item includes any tag corresponding to the preference information.

[0163] For the specific implementation of the above units, please refer to the above Figure 2 to Figure 5 The method embodiments are described in detail herein.

[0164] According to an artificial intelligence-based tag aggregation device provided in an embodiment of the present disclosure, tag aggregation is achieved by aggregating two tags corresponding to multiple content sources whose aggregation degrees are greater than a set threshold, which can make the tag system of content sources more complete and accurate; different content source combinations of the content source set are respectively detected whether their tags can be aggregated, and the tag system will no longer be updated until there are no tags that can be aggregated.

[0165] Figure 7 is a structural diagram of another artificial intelligence-based tag aggregation device provided by an embodiment of the present disclosure. In one embodiment, the artificial intelligence-based tag aggregation device may correspond to the above Figure 2 to Figure 5 In the corresponding embodiment. Figure 7As shown, the artificial intelligence-based label aggregation device may include: a processor, a network interface and a memory. In addition, the above-mentioned artificial intelligence-based label aggregation device may also include: a user interface, and at least one communication bus. Wherein, the communication bus is used to realize the connection and communication between these components. Wherein, the user interface may include a display (display), a keyboard (keyboard), and the optional user interface may also include a standard wired interface and a wireless interface. The network interface may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface). The memory may be a high-speed RAM memory, or it may be a non-volatile memory (non-volatile memory), such as at least one disk storage. The memory may also be optionally at least one storage device located away from the aforementioned processor. As Figure 7 As shown, the memory as a computer-readable storage medium may include an operating system, a network communication module, a user interface module, and a device control application.

[0166] In such Figure 7 In the artificial intelligence-based tag aggregation device shown, the network interface can provide network communication functions; the user interface is mainly used to provide an input interface for the user; and the processor can be used to call the device control application stored in the memory to achieve the above Figure 2 to Figure 5 The description of the tag aggregation method in any corresponding embodiment will not be repeated here. In addition, the description of the beneficial effects of using the same method will not be repeated here.

[0167] It should be understood that the label aggregation device based on artificial intelligence described in the embodiments of the present disclosure can execute the above Figure 2 to Figure 5 The description of the tag aggregation method in any corresponding embodiment will not be repeated here. In addition, the description of the beneficial effects of using the same method will not be repeated here.

[0168] In addition, it should be pointed out here that: the embodiment of the present disclosure also provides a computer-readable storage medium, and the computer-readable storage medium stores a computer program executed by the aforementioned artificial intelligence-based tag aggregation device 1000, and the computer program includes program instructions. When the processor executes the program instructions, it can execute the aforementioned Figure 2 to Figure 5 The description of the tag aggregation method in any corresponding embodiment will not be repeated here. In addition, the description of the beneficial effects of the same method will not be repeated. For technical details not disclosed in the computer-readable storage medium embodiment involved in the present disclosure, please refer to the description of the method embodiment of the present disclosure.

[0169] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0170] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the division of the unit is only a logical function division, and there may be other division methods in actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. The mutual coupling, direct coupling, or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0171] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0172] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can 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 the computer program instructions are loaded and executed on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions can be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (digital subscriber line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media integrations. The available medium may be a read-only memory (ROM), or a random access memory (RAM), or a magnetic medium, such as a floppy disk, a hard disk, a tape, a magnetic disk, or an optical medium, such as a digital versatile disc (DVD), or a semiconductor medium, such as a solid state disk (SSD), etc.

Claims

1. A tag aggregation method based on artificial intelligence, characterized in that: include: Using a neural network for obtaining the probability of labels of content sources, obtaining the probability of multiple labels corresponding to each first content source in multiple first content sources; wherein the content sources include one or more of videos, articles, pictures and music; Obtaining a first label and a second label with the highest probability among the probabilities of the multiple labels corresponding to each first content source; Acquire a degree of aggregation of the first tag and the second tag; wherein the degree of aggregation is a ratio of the number of first content sources corresponding to both the first tag and the second tag to the sum of the numbers, where the sum of the numbers is the sum of the number of first content sources corresponding to the first tag and the number of first content sources corresponding to the second tag; If the aggregation degree of the first tag and the second tag is greater than a first set threshold, the first tag and the second tag are aggregated to obtain a third tag.

2. The method according to claim 1, characterized in that The obtaining the aggregation degree of the first tag and the second tag includes: Acquire a first number of first content sources corresponding to the first tag and the second tag among the plurality of first content sources, a second number of first content sources corresponding to the first tag among the plurality of first content sources, and a third number of first content sources corresponding to the second tag among the plurality of first content sources; The degree of aggregation of the first tag and the second tag is determined according to the first number, the second number, and the third number.

3. The method according to claim 1, characterized in that The method further comprises: The third tag is updated to the tag of the content source set.

4. The method according to any one of claims 1 to 3, characterized in that The method further comprises: The plurality of first content sources are labeled using the third tag.

5. The method according to claim 1, characterized in that The method further comprises: Using a neural network for obtaining the probability of labels of content sources, obtaining the probability of multiple labels corresponding to each second content source in a plurality of second content sources, wherein the plurality of second content sources and the plurality of first content sources belong to the same content source set; Obtaining a third label and a fourth label with the highest probability among the probabilities of the multiple labels corresponding to each second content source; Obtaining a degree of aggregation of the third tag and the fourth tag; If the degree of aggregation of the third tag and the fourth tag is greater than the first set threshold, the third tag and the fourth tag are aggregated to obtain a fifth tag; If the aggregation degree of the third tag and the fourth tag is less than or equal to the first set threshold, the third tag is not aggregated.

6. The method according to any one of claims 1 to 3, characterized in that: The method further comprises: Receive a query request, the query request including a tag corresponding to a content source requested for query; Finding a content source corresponding to the tag; The found content source is returned.

7. The method according to any one of claims 1 to 3, characterized in that The method further comprises: Obtaining user preference information, where the preference information corresponds to one or more aggregated tags; Determining whether the tag of the item includes one or more aggregated tags corresponding to the preference information; If the tag of the item includes any aggregated tag corresponding to the preference information, the item is recommended to the user.

8. A tag aggregation device based on artificial intelligence, characterized in that: include: A first acquisition unit is used to acquire the probability of multiple labels corresponding to each first content source in a plurality of first content sources by using a neural network for acquiring the probability of labels of content sources; wherein the content sources include one or more of videos, articles, pictures and music; A second acquisition unit, configured to acquire a first label and a second label with the highest probability among the probabilities of the multiple labels corresponding to each first content source; a third acquisition unit, configured to acquire a degree of aggregation of the first tag and the second tag; wherein the degree of aggregation is a ratio of the number of first content sources corresponding to both the first tag and the second tag to the sum of the numbers, where the sum of the numbers is the sum of the number of first content sources corresponding to the first tag and the number of first content sources corresponding to the second tag; An aggregation unit is used to aggregate the first tag and the second tag to obtain a third tag if the aggregation degree of the first tag and the second tag is greater than a first set threshold.

9. A tag aggregation device based on artificial intelligence, characterized in that: include: A processor and a memory, wherein the memory stores program instructions, and the processor executes the program instructions to implement the method according to any one of claims 1 to 7.

10. A computer storage medium, characterized in that: The computer storage medium stores one or more instructions, and the one or more instructions are suitable for being loaded by a processor and executing the method according to any one of claims 1 to 7.

Images