Method, device, computer equipment and storage medium for entity alias mining

Abstract

The present invention relates to an entity alias mining method, device, computer equipment and storage medium. In this embodiment, the method includes: obtaining user behavior data and user sound playback data, wherein the user behavior data includes retrieval click behavior data and playback behavior data; extracting session data based on the user behavior data; extracting entities from the retrieval click behavior data and user sound playback data; constructing samples based on the session data, entities and user sound playback data, and inputting them into a word vector model for training to obtain entity vectors; identifying entity aliases based on the cosine similarity, edit distance and pinyin edit distance between entity vectors. By introducing user behavior data to construct the representation model sample data, the embedding effect of the entity is improved, the problem of multiple words and one meaning of the entity is solved, and the entity aliases with higher accuracy can be mined without requiring a lot of manpower and with low maintenance costs.

Description

Technical Field

[0001] The present invention relates to the field of natural language processing technology, and in particular to an entity alias mining method, device, computer equipment and storage medium. Background Art

[0002] Natural Language Processing (NLP) is a key area of ​​research in computer science and artificial intelligence. It studies the theories and methods that enable effective communication between humans and computers using natural language. NLP is primarily used in machine translation, public opinion monitoring, automatic summarization, opinion extraction, text classification, question answering, text semantic comparison, speech recognition, and Chinese optical character recognition (OCR).

[0003] Currently, there are two main methods for entity-based alias mining:

[0004] 1. Alias ​​mining method for entities based on knowledge base

[0005] This method obtains aliases with the help of existing knowledge bases (such as the "HIT Information Retrieval Laboratory Synonym Dictionary Extended Edition" and HowNet, etc.).

[0006] shortcoming:

[0007] The knowledge base has limited coverage and is not available in every niche area;

[0008] It lacks the ability to automatically discover aliases from text, resulting in a low recall rate for words that do not exist in the knowledge base;

[0009] The semantics of the words are not considered;

[0010] Knowledge bases require a lot of manpower and material resources, have high maintenance costs, and require a lot of expert knowledge.

[0011] 2. Alias ​​mining method for entities based on text context relevance

[0012] The basic idea behind this method is that the more similar the contexts of two words are, the greater the probability that one word is an alias of the other. Using word embeddings to mine aliases is a common approach. For example, word2vec is used to train word embeddings, and then cosine similarity is calculated. The top k most similar words are then used to obtain their k aliases.

[0013] shortcoming:

[0014] Failing to consider user behavior;

[0015] Many of the mined aliases are not real aliases and require manual screening;

[0016] The effect is poor for words with low word frequency. Summary of the Invention

[0017] In view of the above problems, the present invention is proposed to provide an entity alias mining method, apparatus, computer equipment and storage medium that overcome the above problems or at least partially solve the above problems, and can mine entity aliases with higher accuracy.

[0018] In a first aspect, an embodiment of the present invention provides a method for mining entity aliases, the method comprising the following steps:

[0019] Obtaining user behavior data and user sound playback data, wherein the user behavior data includes search click behavior data and playback behavior data;

[0020] Extract session data based on user behavior data;

[0021] Extract entities from the data of click behavior and user-played sounds;

[0022] Construct samples based on conversation data, entities, and user-played sound data, and input them into the word vector model for training to obtain entity vectors;

[0023] Identify entity aliases based on cosine similarity, edit distance, and phonetic edit distance between entity vectors.

[0024] Optionally, extracting session data based on user behavior data includes:

[0025] Extract click session data based on retrieval click behavior data;

[0026] Extract playback session data based on playback behavior data.

[0027] Optionally, the retrieved click behavior data includes user information and retrieval behavior data; and extracting click session data based on the retrieved click behavior data includes:

[0028] Extract click session data based on the user's search behavior data at a preset operation time interval.

[0029] Optionally, the playback behavior data includes user information, the time when the user clicks to play, and the playback duration of the sound.

[0030] Optionally, extracting playback session data based on playback behavior data includes:

[0031] Extracting basic playback session data based on playback behavior data;

[0032] Generate a graph based on basic play session data and obtain frequent subgraphs through frequent subgraph mining;

[0033] Generate augmented play session data based on frequent subgraphs.

[0034] Optionally, the search click behavior data includes the user's search term, and the user's sound playback data includes a sound unique identifier, title, description, tag, and host unique identifier; extracting entities from the search click behavior data and the user's sound playback data includes:

[0035] Extract entities from search terms, titles, and descriptions.

[0036] Optionally, the entity alias mining method further includes:

[0037] The entity alias is integrated with the pre-built alias knowledge base to complete the entity alias.

[0038] In a second aspect, an embodiment of the present invention provides an entity alias mining device, the device comprising:

[0039] A data acquisition module is used to acquire user behavior data and user sound playback data, wherein the user behavior data includes search click behavior data and playback behavior data;

[0040] Session data extraction module, used to extract session data based on user behavior data;

[0041] The entity extraction module is used to extract entities from the data of click behavior and user-played sounds;

[0042] The entity vector representation module is used to construct samples based on conversation data, entities, and user-played sound data, and input them into the word vector model for training to obtain entity vectors;

[0043] The alias identification module is used to identify entity aliases based on the cosine similarity, edit distance, and pinyin edit distance between entity vectors.

[0044] Optionally, the session data extraction module includes:

[0045] A click session data extraction submodule, for extracting click session data based on the retrieved click behavior data;

[0046] The playback session data extraction submodule is used to extract playback session data based on the playback behavior data.

[0047] Optionally, the retrieval click behavior data includes user information and retrieval behavior data; the click session data extraction submodule is further used to extract click session data based on the user's retrieval behavior data at a preset operation time interval.

[0048] Optionally, the playback behavior data includes user information, the time when the user clicks to play, and the playback duration of the sound.

[0049] Optionally, the playback session data extraction submodule is further configured to:

[0050] Extracting basic playback session data based on playback behavior data;

[0051] Generate a graph based on basic play session data and obtain frequent subgraphs through frequent subgraph mining;

[0052] Generate augmented play session data based on frequent subgraphs.

[0053] Optionally, the search click behavior data includes the user's search terms, and the data of the user playing sound includes the sound unique identifier, title, description, tag, and anchor unique identifier; the entity extraction module is also used to extract entities from the search terms, title, and description.

[0054] Optionally, the device further comprises:

[0055] The alias fusion module is used to fuse the entity's alias with the pre-built alias knowledge base to complete the entity's alias.

[0056] In a third aspect, an embodiment of the present invention provides a computer device, comprising:

[0057] one or more processors;

[0058] a memory for storing one or more programs;

[0059] When the one or more programs are executed by the one or more processors, the one or more processors implement the entity alias mining method as described in any one of the first aspects.

[0060] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium.

[0061] The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the entity alias mining method as described in any one of the first aspects is implemented.

[0062] In this embodiment, user behavior data and user audio playback data are obtained, including search and click behavior data and audio playback data. Session data is extracted based on the user behavior data. Entities are extracted from the search and click behavior data and audio playback data. Samples are constructed based on the session data, entity data, and audio playback data, and then input into a word vector model for training to obtain entity vectors. Entity aliases are identified based on cosine similarity, edit distance, and phonetic edit distance between entity vectors. Incorporating user behavior data into the representation model sample data improves entity embedding, resolves the problem of entities with multiple words and one meaning, and can mine more accurate entity aliases without requiring extensive manpower and resulting in low maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0064] Figure 1 A flowchart of a method for mining entity aliases provided in the first embodiment of the present invention;

[0065] Figure 1A A schematic diagram of a click session data session combination provided in the first embodiment of the present invention;

[0066] Figure 1B A schematic diagram of a basic playback session data session combination provided in the first embodiment of the present invention;

[0067] Figure 2 A schematic diagram of the structure of an entity alias mining device provided in the second embodiment of the present invention;

[0068] Figure 3 A schematic diagram of the structure of a computer device provided in Example 3 of the present invention. DETAILED DESCRIPTION

[0069] The present invention will be further described in detail below with reference to the accompanying drawings and examples. It will be understood that the specific embodiments described herein are intended only to illustrate the present invention and are not intended to limit the present invention. It should also be noted that, for ease of description, the accompanying drawings only illustrate portions relevant to the present invention, not all structures.

[0070] Natural Language Processing (NLP) is a key area of ​​research in computer science and artificial intelligence. It studies the theories and methods that enable effective communication between humans and computers using natural language. NLP is primarily used in machine translation, public opinion monitoring, automatic summarization, opinion extraction, text classification, question answering, text semantic comparison, speech recognition, and Chinese optical character recognition (OCR).

[0071] Currently, there are two main methods for entity-based alias mining:

[0072] 1. Alias ​​mining method for entities based on knowledge base

[0073] This method obtains aliases with the help of existing knowledge bases (such as the "HIT Information Retrieval Laboratory Synonym Dictionary Extended Edition" and HowNet, etc.).

[0074] shortcoming:

[0075] The knowledge base has limited coverage and is not available in every niche area;

[0076] It lacks the ability to automatically discover aliases from text, resulting in a low recall rate for words that do not exist in the knowledge base;

[0077] The semantics of the words are not considered;

[0078] Knowledge bases require a lot of manpower and material resources, have high maintenance costs, and require a lot of expert knowledge.

[0079] 2. Alias ​​mining method for entities based on text context relevance

[0080] The basic idea behind this method is that the more similar the contexts of two words are, the greater the probability that one word is an alias of the other. Using word embeddings to mine aliases is a common approach. For example, word2vec is used to train word embeddings, and then cosine similarity is calculated. The top k most similar words are then used to obtain their k aliases.

[0081] shortcoming:

[0082] Failing to consider user behavior;

[0083] Many of the mined aliases are not real aliases and require manual screening;

[0084] The effect is poor for words with low word frequency.

[0085] To overcome the above problems or at least partially solve the above problems, embodiments of the present application provide an entity alias mining method that can mine entity aliases with higher accuracy. Detailed description will be given below through embodiments.

[0086] Example 1

[0087] Figure 1 This is a flowchart of a method for mining entity aliases provided in a first embodiment of the present invention. This method can be performed by an entity alias mining device. The entity alias mining device can be implemented by software and / or hardware and can be configured in a computer device, such as a server, a personal computer, etc. The entity alias mining method specifically includes the following steps:

[0088] Step 101: Obtain user behavior data and user sound playback data.

[0089] In the sound data retrieval scenario, when the user enters the search content in the search bar and clicks the search button, the background extracts keywords based on the search content (search terms) entered by the user, matches them in the background sound database based on the extracted search keywords, extracts the data of the matching search result sound, and displays it to the user. The user clicks and plays one or more sounds based on his understanding of the displayed sound data, or searches again.

[0090] The user behavior data may include retrieval click behavior data and playback behavior data, wherein the retrieval click behavior data may include user information, retrieval behavior data, and user search terms; the playback behavior data may include user information, user click playback time, and sound playback duration.

[0091] The sound data can be stored in the database with data such as the sound unique identifier, title, description, tags (primary tags, secondary tags, etc.), the host unique identifier, and the sound itself.

[0092] Of course, in actual applications, the above-mentioned user behavior data and sound data may also contain other information or be replaced by other information, etc. The specific information or data included may vary for different usage scenarios, and the embodiments of the present invention do not impose specific restrictions on this.

[0093] Step 102: Extract session data based on user behavior data.

[0094] In the context of audio data retrieval, each user behavior cannot be isolated; rather, one behavior has more or less relationships with its predecessors and successors. Introducing user behavior data can improve the accuracy of entity mining.

[0095] In one embodiment, step 102 includes:

[0096] Sub-step 1021: extract click session data based on the retrieved click behavior data.

[0097] Specifically, the retrieval click behavior data includes user information and retrieval behavior data. In one embodiment, sub-step 1021 includes extracting click session data based on the user's retrieval behavior data at a preset operation time interval.

[0098] The user enters a search term, and the background displays the matching sound data based on the search term. The user's behavior of clicking on the matching sound data belongs to the same click session data session.

[0099] If the time interval between two adjacent search behaviors of a user (search behaviors include user input of search terms and click on sound data) does not exceed the preset operation time interval, then the click behaviors of the two adjacent search behaviors also belong to the same click session data session.

[0100] For example, Figure 1A ,The sounds that user U1 clicks to play when searching query1 are I1, I3, and I8, and the sounds that user U1 clicks to play when searching query2 are I2, I6, and I5. Since the time interval between the two search behaviors exceeds the preset operation time interval (the preset operation time interval is 20 seconds), the click session data session1 is user U1’s click to play I1, I3, and I8, and the click session data session2 is user U1’s click to play I2, I6, and I5.

[0101] The sounds that user U2 clicks to play when searching query3 are I5, I4, I9, and I2. The sounds that user U2 clicks to play when searching query4 are I7 and I6. Since the time interval between the two search behaviors does not exceed the preset operation time interval (the preset operation time interval is 20 seconds), the click session data session3 is user U2's clicks to play I5, I4, I9, I2, I7, and I6.

[0102] Sub-step 1022: extracting playback session data based on the playback behavior data.

[0103] In one embodiment, sub-step 1022 includes:

[0104] Extracting basic playback session data based on playback behavior data;

[0105] Generate a graph based on basic play session data and obtain frequent subgraphs through frequent subgraph mining;

[0106] Generate augmented play session data based on frequent subgraphs.

[0107] Playback behavior data includes user information, the time when the user clicks to play, and the duration of the sound playback. Playback session data includes basic playback session data and extended playback session data.

[0108] A play sequence table is generated based on the user's click behavior on the sound data. If the time interval between the user click times of two adjacent sounds in the play sequence table does not exceed the preset click time interval, and the play duration of each sound is not less than the preset play duration, then they belong to the same basic play session data session.

[0109] For example, Figure 1B The sounds clicked and played by user U1 are arranged as follows: I1, I3, I8, I2, I6, and I5. The playback time of I1 is 32 seconds, the playback time of I3 is 26 seconds, the playback time of I8 is 46 seconds, the playback time of I2 is 198 seconds, the playback time of I6 is 32 seconds, and the playback time of I5 is 59 seconds. The playback time of each sound exceeds the preset playback time (the preset playback time is 20 seconds), and only the time interval between the user clicks of the two adjacent sounds I8 and I2 exceeds the preset click time interval (the preset click time interval is 20 minutes). Then the basic playback session data session4 is the click-to-play of I1, I3, and I8 by user U1, and the basic playback session data session5 is the click-to-play of I2, I6, and I5 by user U1.

[0110] The sounds clicked and played by user U2 are arranged as follows: I1, I3, I8, I2, I7, and I6, among which the playback time of I1 is 32 seconds, the playback time of I3 is 26 seconds, the playback time of I8 is 46 seconds, the playback time of I2 is 198 seconds, the playback time of I7 is 32 seconds, and the playback time of I6 is 59 seconds. The playback time of each sound exceeds the preset playback time (the preset playback time is 20 seconds), and the time interval between the user clicks of two adjacent sounds does not exceed the preset click time interval (the preset click time interval is 20 minutes). The basic playback session data session6 is the click-to-play of I1, I3, I8, I2, I7, and I6 by user U2.

[0111] For basic playback session data extracted based on playback behavior data, in fact, each basic playback session data can generate a graph, and frequent subgraphs can be obtained through frequent subgraph mining.

[0112] For example, taking the minimum support threshold as 2, three graphs can be generated from the three basic playback session data session4, session5, and session6 obtained above. Based on these three graphs, it can be found that I1, I3, and I8 are a frequent subgraph. Then, the order of I1, I3, and I8 is randomly shuffled to generate two new subgraphs, such as I1, I8, I3 and I8, I3, I1.

[0113] Step 103: Extract entities from the retrieved click behavior data and the user played sound data.

[0114] In one embodiment, step 103 includes extracting entities from search terms, titles, and descriptions.

[0115] Entities can be extracted using traditional machine learning methods such as the Hidden Markov Model (HMM), Maximum Entropy Markov Model (MEMM), Support Vector Machine (SVM), and Conditional Random Field (CRF). Deep learning methods such as the BILSTM-CRF model and the LSTM-CRF model can also be used. This embodiment does not limit the specific method for entity extraction.

[0116] Step 104: construct a sample based on the conversation data, entity, and user sound playback data, and input it into the word vector model for training to obtain the entity vector.

[0117] In this embodiment, the method of constructing samples based on click session data is different from the method of constructing samples based on play session data. For example, the sample constructed based on click session data session1 is as follows:

[0118] Query1 entity I1 entity I1 sound unique identifier I1 anchor unique identifier I1 first-level tag I1 second-level tag Query1 entity I3 entity I3 sound unique identifier I3 anchor unique identifier I3 first-level tag I3 second-level tag Query1 entity I8 entity I8 sound unique identifier I8 anchor unique identifier I8 first-level tag I8 second-level tag.

[0119] The sample constructed based on the playback session data session4 is as follows:

[0120] I1 entity I1 sound unique identifier I1 anchor unique identifier I1 first-level label I1 second-level label I3 entity I3 sound unique identifier I3 anchor unique identifier I3 first-level label I3 second-level label I8 entity I8 sound unique identifier I8 anchor unique identifier I8 first-level label I8 second-level label.

[0121] The constructed samples are input into a word vector model, such as the word2vec model, for training to obtain entity vectors, vector representations of the unique sound identifier, vector representations of the unique host identifier, vector representations of the primary labels, and vector representations of the secondary labels.

[0122] Step 105: Identify entity aliases based on the cosine similarity, edit distance, and pinyin edit distance between entity vectors.

[0123] According to the vector representation VEC of entity W W, calculate all other entities W i Cosine similarity with Edit distance sim dit , the edit distance of pinyin sim pinyin Then make a judgment, the judgment logic can be as follows:

[0124] 1) If sim cos >0.8, then W i It is an alias for W;

[0125] 2) If sim cos >0.5, and sim pinyin >0.9 or sim edit >0.7, then W i It is an alias of W.

[0126] In one embodiment, the entity alias mining method further includes:

[0127] Step 106: Fuse the entity's alias with the pre-built alias knowledge base to complete the entity's alias.

[0128] The pre-built alias knowledge base can be a knowledge base that has been built by others, such as the extended version of the Synonymous Dictionary of Harbin Institute of Technology Information Retrieval Laboratory, HowNet, etc., or it can be an alias knowledge base after deduplication by integrating two or more knowledge bases. This embodiment does not limit the specific source and construction method of the alias knowledge base.

[0129] By integrating the entity alias with the pre-built alias knowledge base, the alias of each entity can be obtained to complete the entity alias.

[0130] In this embodiment, user behavior data and user audio playback data are obtained, including search and click behavior data and audio playback data. Session data is extracted based on the user behavior data. Entities are extracted from the search and click behavior data and audio playback data. Samples are constructed based on the session data, entity data, and audio playback data, and then input into a word vector model for training to obtain entity vectors. Entity aliases are identified based on cosine similarity, edit distance, and phonetic edit distance between entity vectors. Incorporating user behavior data into the representation model sample data improves entity embedding, resolves the problem of entities with multiple words and one meaning, and can mine more accurate entity aliases without requiring extensive manpower and resulting in low maintenance costs.

[0131] It should be noted that for the sake of simplicity, the method embodiments are described as a series of actions. However, those skilled in the art should be aware that the embodiments of the present invention are not limited by the order of the actions described, because according to the embodiments of the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present invention.

[0132] Example 2

[0133] Figure 2 This is a schematic diagram of the structure of an entity alias mining device provided in the second embodiment of the present invention. The entity alias mining device may specifically include the following modules:

[0134] The data acquisition module 201 is used to acquire user behavior data and user sound playback data, wherein the user behavior data includes search click behavior data and playback behavior data;

[0135] A session data extraction module 202 is used to extract session data based on user behavior data;

[0136] Entity extraction module 203, used to extract entities from the search click behavior data and the user play sound data;

[0137] The entity vector representation module 204 is used to construct samples based on the conversation data, entities, and user-played sound data, and input them into the word vector model for training to obtain entity vectors;

[0138] The alias identification module 205 is used to identify the alias of an entity based on the cosine similarity, edit distance, and pinyin edit distance between entity vectors.

[0139] In one embodiment, the session data extraction module 202 includes:

[0140] A click session data extraction submodule, for extracting click session data based on the retrieved click behavior data;

[0141] The playback session data extraction submodule is used to extract playback session data based on the playback behavior data.

[0142] In one embodiment, the retrieval click behavior data includes user information and retrieval behavior data; the click session data extraction submodule is further used to extract click session data based on the user's retrieval behavior data at a preset operation time interval.

[0143] In one embodiment, the playback behavior data includes user information, the time when the user clicks to play, and the playback duration of the sound.

[0144] In one embodiment, the playback session data extraction submodule is further configured to:

[0145] Extracting basic playback session data based on playback behavior data;

[0146] Generate a graph based on basic play session data and obtain frequent subgraphs through frequent subgraph mining;

[0147] Generate augmented play session data based on frequent subgraphs.

[0148] In one embodiment, the retrieval click behavior data includes the user's search terms, and the data of the user playing sound includes the sound unique identifier, title, description, tag, and anchor unique identifier; the entity extraction module 203 is also used to extract entities from the search terms, title, and description.

[0149] In one embodiment, the entity alias mining device further includes:

[0150] The alias fusion module is used to fuse the entity's alias with the pre-built alias knowledge base to complete the entity's alias.

[0151] The entity alias mining device provided by the embodiment of the present invention can execute the entity alias mining method provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0152] Example 3

[0153] Figure 3 A schematic diagram of the structure of a computer device provided in Example 3 of the present invention. Figure 3 A block diagram of an exemplary computer device 12 suitable for use in implementing embodiments of the present invention is shown. Figure 3 The computer device 12 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0154] like Figure 3 As shown, computer device 12 is implemented as a general-purpose computing device. Components of computer device 12 may include, but are not limited to, one or more processors or processing units 16, system memory 28, and a bus 18 that connects various system components (including system memory 28 and processing unit 16).

[0155] Bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures. Examples of these architectures include, but are not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MAC) bus, an Enhanced ISA bus, a Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus.

[0156] The computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0157] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. Computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be configured to read and write non-removable, non-volatile magnetic media ( Figure 3 Not shown, often called a "hard drive"). Although Figure 3 Not shown, a magnetic disk drive for reading and writing to a removable non-volatile magnetic disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 via one or more data media interfaces. Memory 28 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of various embodiments of the present invention.

[0158] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data, each of which, or some combination thereof, may include an implementation of a network environment. Program modules 42 generally implement the functions and / or methods of the embodiments described herein.

[0159] The computer device 12 can also communicate with one or more external devices 14 (e.g., a keyboard, pointing device, display 24, etc.), one or more devices that enable a user to interact with the computer device 12, and / or any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., a network card, a modem, etc.). Such communication can occur via an input / output (I / O) interface 22. Furthermore, the computer device 12 can communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network such as the Internet) via a network adapter 20. As shown, the network adapter 20 communicates with the other modules of the computer device 12 via a bus 18. It should be understood that, although not shown, other hardware and / or software modules can be used in conjunction with the computer device 12, including but not limited to microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0160] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the entity alias mining method provided by the embodiment of the present invention.

[0161] Example 4

[0162] Embodiment 4 of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the various processes of the above-mentioned entity alias mining method are implemented and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0163] Among them, computer-readable storage media can include, for example, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or components, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device or device.

[0164] Note that the above are only preferred embodiments of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments, and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments and may include many other equivalent embodiments without departing from the concept of the present invention. The scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for mining entity aliases, characterized in that: include: Obtaining user behavior data and user sound playback data, wherein the user behavior data includes search click behavior data and playback behavior data; Extract session data based on user behavior data; Extract entities from the data of click behavior and user-played sounds; Construct samples based on conversation data, entities, and user-played sound data, and input them into the word vector model for training to obtain entity vectors; Identify entity aliases based on cosine similarity, edit distance, and phonetic edit distance between entity vectors; The extracting of session data based on user behavior data includes: Extract click session data based on retrieval click behavior data; Extracting playback session data based on playback behavior data; The retrieval click behavior data includes user information and retrieval behavior data; and the extraction of click session data based on the retrieval click behavior data includes: Extracting click session data based on the user's search behavior data at a preset operation time interval; The search click behavior data includes the user's search term, and the user's sound playback data includes the sound unique identifier, title, description, tag, and anchor unique identifier; extracting entities from the search click behavior data and the user's sound playback data includes: Extract entities from search terms, titles, and descriptions; The identifying entity aliases based on cosine similarity, edit distance, and phonetic edit distance between entity vectors includes: If the cosine similarity between the two entity vectors is greater than a first preset threshold, the two entities are aliases of each other; If the cosine similarity between two entity vectors is greater than a second preset threshold, the second preset threshold is less than the first preset threshold, and the edit distance between the two entity vectors is greater than a third preset threshold, or the pinyin edit distance between the two entity vectors is greater than a fourth preset threshold, then the two entities are aliases of each other.

2. The method according to claim 1, characterized in that The playback behavior data includes user information, the time when the user clicks to play, and the playback duration of the sound.

3. The method according to claim 2, characterized in that The extracting of playback session data based on playback behavior data includes: Extracting basic playback session data based on playback behavior data; Generate a graph based on basic play session data and obtain frequent subgraphs through frequent subgraph mining; Generate augmented play session data based on frequent subgraphs.

4. The method according to claim 1, characterized in that Also includes: The entity alias is integrated with the pre-built alias knowledge base to complete the entity alias.

5. An entity alias mining device, characterized in that: include: A data acquisition module is used to acquire user behavior data and user sound playback data, wherein the user behavior data includes search click behavior data and playback behavior data; Session data extraction module, used to extract session data based on user behavior data; The entity extraction module is used to extract entities from the data of click behavior and user-played sounds; The entity vector representation module is used to construct samples based on conversation data, entities, and user-played sound data, and input them into the word vector model for training to obtain entity vectors; Alias ​​identification module, used to identify entity aliases based on cosine similarity, edit distance, and phonetic edit distance between entity vectors; The session data extraction module includes: A click session data extraction submodule, for extracting click session data based on the retrieved click behavior data; A playback session data extraction submodule is used to extract playback session data based on playback behavior data; The retrieval click behavior data includes user information and retrieval behavior data; the click session data extraction submodule is further used to extract click session data based on the user's retrieval behavior data at a preset operation time interval; The search click behavior data includes the user's search terms, and the user's sound playback data includes the sound unique identifier, title, description, tag, and anchor unique identifier; the entity extraction module is further used to extract entities from the search terms, title, and description; The alias identification module is further configured to determine that if the cosine similarity between two entity vectors is greater than a first preset threshold, the two entities are aliases of each other; if the cosine similarity between the two entity vectors is greater than a second preset threshold, the second preset threshold is less than the first preset threshold, and the edit distance between the two entity vectors is greater than a third preset threshold, or the pinyin edit distance between the two entity vectors is greater than a fourth preset threshold, the two entities are aliases of each other.

6. A computer device, characterized in that: The computer device comprises: one or more processors; a memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the entity alias mining method according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the entity alias mining method according to any one of claims 1 to 4 is implemented.

Images