CN114528842A

CN114528842A - Word vector construction method, device and equipment and computer readable storage medium

Info

Publication number: CN114528842A
Application number: CN202011322164.8A
Authority: CN
Inventors: 兰国兴; 许娟婷; 张少永
Original assignee: Huawei Technologies Co Ltd
Current assignee: Huawei Technologies Co Ltd
Priority date: 2020-11-23
Filing date: 2020-11-23
Publication date: 2022-05-24

Abstract

The application provides a word vector construction method, a device, equipment and a computer readable storage medium, wherein the method is applied to the field of terminal artificial intelligence and the natural language processing of corresponding sub-fields, and comprises the following steps: acquiring new words, wherein the new words are unknown words or new-meaning words of old words and have first semantics; determining a word vector of a second semantic contained in each similar meaning word based on the word vector of at least one similar meaning word corresponding to the new word, wherein the second semantic corresponds to the first semantic; constructing a word vector of a new word based on the word vector of the second semantic and the number of the similar words contained in each similar word; the word vector for the new word is sent to the end-side device. By adopting the method and the device, the word vector of the new word is constructed according to the word vector of the near-meaning word after the redundant information is removed, so that the word vector of the near-meaning word is more fit with the semantics of the new word of the word vector to be constructed, and the accuracy of the constructed word vector can be improved.

Description

Word vector construction method, device and equipment and computer readable storage medium

Technical Field

The present application relates to natural language processing technology in the field of artificial intelligence, and in particular, to a word vector construction method, apparatus, device, and computer-readable storage medium.

Background

With the rapid development of Artificial Intelligence (AI), end-side devices (such as mobile phones, intelligent robots, bluetooth headsets, etc.) can also perform Natural Language Processing (NLP) related tasks. The end-side equipment executes the NLP task by calling a word vector library and an NLP task model which are downloaded from a server side in advance, in other words, the end-side equipment identifies the semantics of the linguistic data to be processed by depending on the word vector library, and then calls the NLP task model to execute the NLP task according to the semantics. It can be seen that, when the word vector library does not include any vocabulary in the corpus to be processed, the end-side device cannot identify the corpus to be processed, and thus cannot execute NLP related tasks. For example, the word vector library has a word of "blue thin", and when the intelligent question-answering robot receives the corpus to be processed as "blue thin" today, the intelligent question-answering robot cannot recognize the corpus to be processed.

The specific operations of the server to construct a word vector of an unknown word are: and determining a similar word which is most similar to the unknown word in the semantic meaning from the existing word vector library, and taking the word vector of the similar word as the word vector of the unknown word. For example, if the word to be constructed is "orlistat", and the word with the most similar semantic meaning is "oil filling", the server may use the word vector of "oil filling" as the word vector of "orlistat". However, if the similar meaning of the unknown word has a word ambiguity (e.g. the word vector of "refuel" also has the meaning of "add gasoline"), it is not accurate to use the word vector of the similar word as the word vector of the unknown word.

Disclosure of Invention

The application provides a word vector construction method, a word vector construction device, word vector construction equipment and a computer readable storage medium, wherein the word vector of a new word corresponding to a near-sense word can be constructed according to the word vector of the near-sense word without redundant information, so that the word vector of the new word is more accurate.

In a first aspect, the present application provides a word vector construction method, which may be applied to a server, and includes:

acquiring new words, wherein the new words are unknown words or new-meaning vocabularies of old words, the new words have first semantics, the unknown words are vocabularies without word vectors in a server, the new-meaning vocabularies of the old words are vocabularies with word vectors in the server, and the semantics indicated by the word vectors are different from the first semantics; determining a word vector of a second semantic contained in each similar meaning word based on the word vector of at least one similar meaning word corresponding to the new word, wherein the second semantic corresponds to the first semantic; constructing a word vector of the new word based on the number of the near-meaning words corresponding to the new word and the word vector of the second semantic meaning contained in each near-meaning word; the word vector for the new word is sent to the end-side device.

Based on the method described in the first aspect, the server may obtain the similar meaning words of the vocabulary of the new meaning of the unknown word or the old word, and remove the redundant information in the word vectors corresponding to the similar meaning words, so that the word vectors of the similar meaning words more conform to the semantics of the vocabulary of the new meaning of the unknown word or the old word.

In one possible implementation, the server determines, for each of at least one near word, whether the near word contains a third semantic meaning other than the second semantic meaning; if the similar meaning word contains a third semantic, a word vector corresponding to the third semantic is obtained; and determining a word vector of a second semantic contained in the similar meaning word based on the word vector of the similar meaning word and the word vector corresponding to the third semantic. By realizing the possibility, the server can more accurately remove redundant information in the word vector of each similar meaning word, so that the word vector of the similar meaning word is more consistent with the semantics of the vocabulary of the unknown word or the new meaning of the old word.

In one possible implementation, if the near word does not include at least one third semantic, the server determines that the word vector of the near word is a word vector of the second semantic. By realizing the possibility, when the similar meaning words do not contain redundant information, the server can directly construct word vectors of unknown words or new-sense words of old words according to the word vectors of the similar meaning words, and further, the computing resources can be saved.

In one possible implementation, the server obtains a weight value corresponding to each similar meaning word, and constructs a word vector of the new word based on the weight value corresponding to each similar meaning word, the number of the similar meaning words corresponding to the new word, and a word vector of a second semantic included in each similar meaning word. By realizing the possibility, when the vocabulary of the new meaning of the unknown word or the old word has a plurality of similar words, the server can acquire the weighted value of each similar word according to the attaching degree between each similar word and the vocabulary of the new meaning of the unknown word or the old word, and further construct a more accurate word vector of the vocabulary of the new meaning of the unknown word or the old word according to the weighted value of each similar word, the number of the similar words and the word vector of the similar word with redundant information removed.

In a second aspect, the present application provides a semantic recognition method, which may be applied to an end-side device, the method including:

receiving a word vector of a new word sent by a server, wherein the new word has a first semantic meaning, the new word is a vocabulary of an unknown word or a new meaning of an old word, the unknown word is a vocabulary without the word vector in the server, the vocabulary of the new meaning of the old word is a vocabulary with the word vector in the server, and the semantic meaning indicated by the word vector is different from the first semantic meaning; storing the word vectors of the new words in a second word vector library; segmenting the corpus to be processed to obtain at least one participle; if the first target vocabulary has a first word vector in the first word vector library and the first target vocabulary has a second word vector in the second word vector library, determining the language probability corresponding to the corpus to be processed based on the first word vector or the second word vector, wherein the first target vocabulary is any vocabulary in at least one participle; and determining the first word vector or the second word vector as a corresponding word vector of the first target vocabulary in the linguistic data to be processed based on the language probability.

Based on the method described in the second aspect, the end-side device obtains and stores a word vector of a new word (an unknown word or a new word of an old word) from the server, and further, the end-side device can perform semantic recognition on the corpus to be processed containing the new word based on the word vector of the new word (the unknown word or the new word of the old word), thereby improving the semantic recognition capability of the end-side device.

In one possible implementation, if the first target vocabulary does not have word vectors in the first word vector library and the first target vocabulary has word vectors in the second word vector library, it is determined that the word vectors of the first target vocabulary in the second word vector library are the corresponding word vectors of the first target vocabulary in the corpus to be processed. By realizing such a possibility, the end-side device can find the above-mentioned unknown word in the corpus to be processed, and determine the word vector of the unknown word to perform semantic recognition on the unknown word.

In one possible implementation, the end-side device calls a language model based on word vectors of a first word vector and a second target vocabulary in a first word vector library, and determines a language probability corresponding to a corpus to be processed, wherein the second target vocabulary is a vocabulary of at least one participle except the first target vocabulary; if the language probability is smaller than the probability threshold, determining a second word vector as a corresponding word vector of the first target vocabulary in the linguistic data to be processed; and if the language probability is greater than or equal to the probability threshold, determining the first word vector as a corresponding word vector of the first target vocabulary in the linguistic data to be processed. By implementing the possibility, the end-side device can distinguish the new-meaning word from the linguistic data to be processed, and can accurately distinguish the semantic meaning of the new-meaning word in the current application scene, so that the accuracy of the semantic recognition capability of the end-side device is improved.

In one possible implementation, a language model is called based on a second word vector and a word vector of a second target word in a first word vector library, and the language probability corresponding to the corpus to be processed is determined, wherein the second target word is a word except for the first target word in at least one word segmentation; if the language probability is greater than or equal to the probability threshold, determining a second word vector as a corresponding word vector of the first target vocabulary in the linguistic data to be processed; and if the language probability is smaller than the probability threshold, determining the first word vector as a corresponding word vector of the first target vocabulary in the linguistic data to be processed. By implementing the possibility, the end-side device can distinguish the new-meaning word from the to-be-processed corpus, and can accurately distinguish the semantic meaning expressed by the new-meaning word in the to-be-processed corpus, so that the accuracy of the semantic recognition capability of the end-side device is improved.

In a third aspect, the present application provides a word vector construction apparatus, which may be an apparatus in a server or an apparatus capable of being used in cooperation with the server, and the word vector construction apparatus may include: the device comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring new words, the new words are words of unknown words or new meanings of old words, the new words have first semantics, the unknown words are words of which word vectors do not exist in a server, the words of the new meanings of the old words are words of which word vectors exist in the server, and the semantics indicated by the word vectors are different from the first semantics; the determining unit is used for determining a word vector of a second semantic contained in each similar meaning word based on the word vector of at least one similar meaning word corresponding to the new word, wherein the second semantic corresponds to the first semantic; the building unit is used for building a word vector of the new word based on the number of the similar meaning words corresponding to the new word and the word vector of the second semantic contained in each similar meaning word; a sending unit for sending the word vector of the new word to the end-side device.

In a possible implementation, the determining unit is specifically configured to: determining, for each of the at least one near word, whether the near word contains a third semantic meaning other than the second semantic meaning; if the similar meaning word contains a third semantic, a word vector corresponding to the third semantic is obtained; and determining a word vector of a second semantic contained in the similar meaning word based on the word vector of the similar meaning word and the word vector corresponding to the third semantic.

In one possible implementation, the determining unit is further configured to: and if the similar meaning word does not contain at least one third semantic, determining the word vector of the similar meaning word as the word vector of the second semantic.

In a possible implementation, the building unit is specifically configured to: acquiring a weight value corresponding to each similar meaning word; and constructing a word vector of the new word based on the weight value corresponding to each similar meaning word, the number of the similar meaning words corresponding to the new word and the word vector of the second semantic contained in each similar meaning word.

The function of the word vector constructing device can be realized by hardware, and can also be realized by hardware executing corresponding software. The hardware or software includes one or more units corresponding to the above functions. The unit may be software and/or hardware. The operations and advantageous effects performed by the word vector construction apparatus may refer to the method and advantageous effects described in the first aspect, and repeated details are not repeated.

In a fourth aspect, the present application provides a semantic recognition apparatus, which may be an apparatus in a peer-to-peer device or an apparatus capable of being used in matching with the peer-to-peer device, and the semantic recognition apparatus may include: the receiving unit is used for receiving a word vector of a new word sent by the server, wherein the new word has a first semantic meaning, the new word is a vocabulary of an unknown word or a new meaning of an old word, the unknown word is a vocabulary of which the word vector does not exist in the server, the vocabulary of the new meaning of the old word is a vocabulary of which the word vector exists in the server, and the semantic meaning indicated by the word vector is different from the first semantic meaning; the storage unit is used for storing the word vectors of the new words in a second word vector library; the segmentation unit is used for segmenting the linguistic data to be processed to obtain at least one participle; the determining unit is used for determining the language probability corresponding to the corpus to be processed based on the first word vector or the second word vector if the first target vocabulary has the first word vector in the first word vector library and the first target vocabulary has the second word vector in the second word vector library, wherein the first target vocabulary is any vocabulary in at least one participle; the determining unit is further configured to determine, based on the language probability, a first word vector or a second word vector as a word vector corresponding to the first target vocabulary in the corpus to be processed.

In one possible implementation, the determining unit is further configured to: and if the first target vocabulary does not have word vectors in the first word vector library and the first target vocabulary has word vectors in the second word vector library, determining that the word vectors of the first target vocabulary in the second word vector library are the corresponding word vectors of the first target vocabulary in the corpus to be processed.

In a possible implementation, the determining unit is specifically configured to: calling a language model based on word vectors of a first word vector and a second target word in a first word vector library, and determining the language probability corresponding to the linguistic data to be processed, wherein the second target word is a word except the first target word in at least one participle; if the language probability is smaller than the probability threshold, determining a second word vector as a corresponding word vector of the first target vocabulary in the linguistic data to be processed; and if the language probability is greater than or equal to the probability threshold, determining the first word vector as a corresponding word vector of the first target vocabulary in the linguistic data to be processed.

In a possible implementation, the determining unit is specifically configured to: calling a language model based on a second word vector and a word vector of a second target word in the first word vector library, and determining the language probability corresponding to the linguistic data to be processed, wherein the second target word is a word except the first target word in at least one participle; if the language probability is greater than or equal to the probability threshold, determining a second word vector as a corresponding word vector of the first target vocabulary in the corpus to be processed; and if the language probability is smaller than the probability threshold, determining the first word vector as a corresponding word vector of the first target vocabulary in the linguistic data to be processed.

The function of the semantic recognition device can be realized by hardware, and can also be realized by executing corresponding software by hardware. The hardware or software includes one or more units corresponding to the above functions. The unit may be software and/or hardware. The operations and advantageous effects executed by the semantic recognition device can be referred to the method and advantageous effects described in the second aspect, and repeated details are not repeated.

In a fifth aspect, embodiments of the present application further provide a server, which may include a memory and a processor, where the memory is used for storing a computer program that supports a device to execute the above method, and the computer program includes program instructions, and the processor is configured to call the program instructions to execute the word vector construction method according to any one of the above first aspects.

In a sixth aspect, the present application further provides an end-side device, which may include a memory and a processor, where the memory is used for storing a computer program that supports a device to execute the above method, and the computer program includes program instructions, and the processor is configured to call the program instructions to execute the semantic recognition method according to any one of the above second aspects.

In a seventh aspect, this application embodiment further provides a computer-readable storage medium, where a computer program is stored, the computer program includes program instructions, and when executed by a processor, the program instructions cause the processor to execute any one of the word vector construction methods described in any one of the above first aspects or any one of the semantic recognition methods described in any one of the above second aspects.

In an eighth aspect, the present application further provides a computer program, where the computer program includes computer software instructions, and the computer software instructions, when executed by a computer, cause the computer to perform any one of the word vector construction methods according to any one of the first aspects or any one of the semantic recognition methods according to any one of the second aspects.

In a ninth aspect, the present application further provides a chip, where the chip is configured to implement any one of the word vector construction methods described in any one of the first aspects or any one of the semantic recognition methods described in any one of the second aspects.

Drawings

FIG. 1a is a schematic diagram of a language model according to an embodiment of the present application;

FIG. 1b is a schematic diagram of an architecture of another language model provided in an embodiment of the present application;

FIG. 2 is a schematic structural diagram of an end-side device according to an embodiment of the present disclosure;

fig. 3 is a block diagram of a software structure of an end-side device according to an embodiment of the present application;

fig. 4 is a schematic structural diagram of a server according to an embodiment of the present application;

FIG. 5 is a schematic diagram of a language model architecture according to an embodiment of the present application;

fig. 6 is a schematic flowchart of a word vector construction method according to an embodiment of the present application;

FIG. 7 is a diagram illustrating a word vector decomposition according to an embodiment of the present application;

fig. 8 is a schematic diagram of a method for constructing a word vector of an unknown word according to an embodiment of the present application;

FIG. 9 is a diagram illustrating a method for constructing a word vector of a new-sense vocabulary according to an embodiment of the present application;

fig. 10 is a schematic flowchart of a semantic recognition method according to an embodiment of the present application;

fig. 11 is a schematic diagram of a method for determining a probability threshold according to an embodiment of the present disclosure;

fig. 12 is a schematic diagram of another method for determining a probability threshold according to an embodiment of the present disclosure;

FIG. 13 is a flow chart illustrating another semantic identification method according to an embodiment of the present disclosure;

fig. 14 is a schematic structural diagram of a word vector constructing apparatus according to an embodiment of the present application;

fig. 15 is a schematic structural diagram of a semantic recognition device according to an embodiment of the present application.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more clear, the present application will be further described in detail with reference to the accompanying drawings.

The terms "first" and "second," and the like in the description, claims, and drawings of the present application are used for distinguishing between different objects and not for describing a particular order. Furthermore, the terms "include" and "have," as well as any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, system, article, or apparatus that comprises a list of operations or elements is not limited to those listed but may alternatively include other operations or elements not listed or inherent to such process, method, article, or apparatus.

Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is explicitly and implicitly understood by one skilled in the art that the embodiments described herein can be combined with other embodiments.

In the present application, "at least one" means one or more, "a plurality" means two or more, "at least two" means two or three and three or more, "and/or" for describing the correspondence of the corresponding objects, indicating that three relationships may exist, for example, "a and/or B" may indicate: only A, only B and both A and B are present, wherein A and B may be singular or plural. The character "/" generally indicates that the preceding and following counterpart pairs are in an "or" relationship. "at least one of the following" or similar expressions refer to any combination of these items, including any combination of single item(s) or plural items. For example, at least one (one) of a, b, or c, may represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", wherein a, b, c may be single or plural.

In order to better understand the word vector construction method and the semantic recognition method provided in the embodiments of the present application, a system architecture applied in the embodiments of the present application is introduced below.

Referring to fig. 1a, fig. 1a is a schematic diagram of a system architecture provided in an embodiment of the present application, wherein the system architecture 10 includes a peer device 100 and a server 200. It should be understood that the number of the end-side devices 100 and the number of the servers 200 shown in fig. 1a are only illustrative, and the present application is not limited thereto.

Referring to fig. 1b, fig. 1b is another system architecture diagram, in which fig. 1b reflects a data interaction process between the end-side device 100 and the server 200, and the server 200 is mainly used for acquiring a new word and constructing a word vector of the new word; training a language model according to the text corpus set in the storage space of the server 200; and calculating a probability threshold according to the text corpus set and the language model. Further, the server 200 transmits the new word vector, the language model and the probability threshold to the end-side device 100, so that the end-side device 100 can perform NLP task processing, which may include smart question answering, text classification, natural language generation, emotion classification, dialog management, and the like, according to the new word vector, the language model and the probability threshold transmitted by the server 200.

The end-side device 100 referred to in the embodiments of the present application, which may also be referred to as a terminal device, is an entity on the user side for running a natural language processing model (e.g., segmenting the corpus to be processed and running a semantic recognition method). For example, the end-side device 100 may be a handheld device, a vehicle-mounted device, or the like having a wireless connection function. The end-side device may also be other processing devices connected to the wireless modem. The end-side device may also communicate with a Radio Access Network (RAN), which is a wireless terminal device, a subscriber unit (subscriber unit), a subscriber station (subscriber station), a mobile station (mobile), a remote station (remote station), an access point (access point), a remote terminal device (remote), an access terminal device (access terminal), a user terminal device (user terminal), a user agent (user agent), a user device (user device), or a User Equipment (UE), etc. The end-side devices may also be mobile terminal devices such as mobile telephones (or so-called "cellular" telephones) and computers with mobile terminal devices, e.g. mobile devices which may be portable, pocket, hand-held, computer-included or vehicle-mounted, which exchange language and/or data with the radio access network. The end-side device may also be a Personal Communication Service (PCS) phone, a cordless phone, a Session Initiation Protocol (SIP) phone, a Wireless Local Loop (WLL) station, a Personal Digital Assistant (PDA), or the like. A common end-side device 100 may include: the mobile internet device comprises an automobile, an unmanned aerial vehicle, a mechanical arm, a mobile phone, a tablet computer, a notebook computer, a handheld computer, a Mobile Internet Device (MID), and a wearable device, such as a smart watch, a smart bracelet, a pedometer, and the like, but the embodiment of the present application is not limited thereto.

The structure of the end-side device 100 will be described below. Referring to fig. 2, fig. 2 is a schematic structural diagram of an end-side device 100 according to an embodiment of the present application.

The end-side device 100 may include a processor 110, an external memory interface 120, an internal memory 121, a Universal Serial Bus (USB) interface 130, a charging management module 140, a power management module 141, a battery 142, an antenna 1, an antenna 2, a mobile communication module 150, a wireless communication module 160, an audio module 170, a speaker 170A, a receiver 170B, a microphone 170C, an earphone interface 170D, a sensor module 180, a button 190, a motor 191, an indicator 192, a camera 193, a display screen 194, a Subscriber Identity Module (SIM) card interface 195, and the like. The sensor module 180 may include a pressure sensor 180A, a gyroscope sensor 180B, an air pressure sensor 180C, a magnetic sensor 180D, an acceleration sensor 180E, a distance sensor 180F, a proximity light sensor 180G, a fingerprint sensor 180H, a temperature sensor 180J, a touch sensor 180K, an ambient light sensor 180L, a bone conduction sensor 180M, and the like.

It is to be understood that the illustrated structure of the embodiment of the present invention does not constitute a specific limitation to the end-side device 100. In other embodiments of the present application, the end-side device 100 may include more or fewer components than shown, or combine certain components, or split certain components, or a different arrangement of components. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

Processor 110 may include one or more processing units, such as: the processor 110 may include an Application Processor (AP), a modem processor, a Graphics Processing Unit (GPU), an Image Signal Processor (ISP), a controller, a memory, a video codec, a Digital Signal Processor (DSP), a baseband processor, and/or a neural-Network Processing Unit (NPU), etc. The different processing units may be separate devices or may be integrated into one or more processors.

Wherein the controller can be a neural center and a command center of the end-to-end device 100. The controller can generate an operation control signal according to the instruction operation code and the timing signal to complete the control of instruction fetching and instruction execution.

A memory may also be provided in processor 110 for storing instructions and data. In some embodiments, the memory in the processor 110 is a cache memory. The memory may hold instructions or data that have just been used or recycled by the processor 110. If the processor 110 needs to reuse the instruction or data, it can be called directly from the memory. Avoiding repeated accesses reduces the latency of the processor 110, thereby increasing the efficiency of the system.

In some embodiments, processor 110 may include one or more interfaces. The interface may include an integrated circuit (I2C) interface, an integrated circuit built-in audio (I2S) interface, a Pulse Code Modulation (PCM) interface, a universal asynchronous receiver/transmitter (UART) interface, a Mobile Industry Processor Interface (MIPI), a general-purpose input/output (GPIO) interface, a Subscriber Identity Module (SIM) interface, and/or a Universal Serial Bus (USB) interface, etc.

The I2C interface is a bi-directional synchronous serial bus that includes a serial data line (SDA) and a Serial Clock Line (SCL). In some embodiments, processor 110 may include multiple sets of I2C buses. The processor 110 may be coupled to the touch sensor 180K, the charger, the flash, the camera 193, etc. through different I2C bus interfaces, respectively. For example: the processor 110 may be coupled to the touch sensor 180K via an I2C interface, such that the processor 110 and the touch sensor 180K communicate via an I2C bus interface to implement the touch functionality of the end-side device 100.

The I2S interface may be used for audio communication. In some embodiments, processor 110 may include multiple sets of I2S buses. The processor 110 may be coupled to the audio module 170 via an I2S bus to enable communication between the processor 110 and the audio module 170. In some embodiments, the audio module 170 may communicate audio signals to the wireless communication module 160 via the I2S interface, enabling answering of calls via a bluetooth headset.

The PCM interface may also be used for audio communication, sampling, quantizing and encoding analog signals. In some embodiments, the audio module 170 and the wireless communication module 160 may communicate through a PCM interface. In some embodiments, the audio module 170 may also transmit audio signals to the wireless communication module 160 through the PCM interface, so as to implement a function of answering a call through a bluetooth headset. Both the I2S interface and the PCM interface may be used for audio communication.

The UART interface is a universal serial data bus used for asynchronous communications. The bus may be a bidirectional communication bus. It converts the data to be transmitted between serial communication and parallel communication. In some embodiments, a UART interface is generally used to connect the processor 110 with the wireless communication module 160. For example: the processor 110 communicates with a bluetooth module in the wireless communication module 160 through a UART interface to implement a bluetooth function. In some embodiments, the audio module 170 may transmit the audio signal to the wireless communication module 160 through a UART interface, so as to realize the function of playing music through a bluetooth headset.

MIPI interfaces may be used to connect processor 110 with peripheral devices such as display screen 194, camera 193, and the like. The MIPI interface includes a Camera Serial Interface (CSI), a Display Serial Interface (DSI), and the like. In some embodiments, the processor 110 and the camera 193 communicate through a CSI interface, enabling the capture functionality of the end-side device 100. The processor 110 and the display screen 194 communicate through the DSI interface to implement the display function of the end-side device 100.

The GPIO interface may be configured by software. The GPIO interface may be configured as a control signal and may also be configured as a data signal. In some embodiments, a GPIO interface may be used to connect the processor 110 with the camera 193, the display 194, the wireless communication module 160, the audio module 170, the sensor module 180, and the like. The GPIO interface may also be configured as an I2C interface, an I2S interface, a UART interface, a MIPI interface, and the like.

The USB interface 130 is an interface conforming to the USB standard specification, and may specifically be a Mini USB interface, a Micro USB interface, a USB Type C interface, or the like. The USB interface 130 may be used to connect a charger to charge the end-side device 100, and may also be used to transmit data between the end-side device 100 and a peripheral device. And the earphone can also be used for connecting an earphone and playing audio through the earphone. The interface may also be used to connect other electronic devices, such as AR devices and the like.

It should be understood that the interface connection relationship between the modules in the embodiment of the present invention is only illustrated schematically, and does not form a structural limitation on the end-side device 100. In other embodiments of the present application, the end-side device 100 may also adopt different interface connection manners or a combination of a plurality of interface connection manners in the above embodiments.

The charging management module 140 is configured to receive charging input from a charger. The charger may be a wireless charger or a wired charger.

The power management module 141 is used to connect the battery 142, the charging management module 140 and the processor 110. The power management module 141 receives the input of the battery 142 and/or the charging management module 140 and supplies power to the processor 110, the internal memory 121, the external memory, the display 194, the camera 193, the wireless communication module 160, and the like. in other embodiments, the power management module 141 may also be disposed in the processor 110. In other embodiments, the power management module 141 and the charging management module 140 may be disposed in the same device.

The wireless communication function of the end-side device 100 may be implemented by the antenna 1, the antenna 2, the mobile communication module 150, the wireless communication module 160, a modem processor, a baseband processor, and the like.

The antennas 1 and 2 are used for transmitting and receiving electromagnetic wave signals. Each antenna in the end-side device 100 may be used to cover a single or multiple communication bands. Different antennas can also be multiplexed to improve the utilization of the antennas. For example: the antenna 1 may be multiplexed as a diversity antenna of a wireless local area network. In other embodiments, the antenna may be used in conjunction with a tuning switch.

The mobile communication module 150 may provide a solution including wireless communication of 2G/3G/4G/5G, etc. applied on the end-side device 100. The mobile communication module 150 may include at least one filter, a switch, a power amplifier, a Low Noise Amplifier (LNA), and the like. The mobile communication module 150 may receive the electromagnetic wave from the antenna 1, filter, amplify, etc. the received electromagnetic wave, and transmit the electromagnetic wave to the modem processor for demodulation. The mobile communication module 150 may also amplify the signal modulated by the modem processor, and convert the signal into electromagnetic wave through the antenna 1 to radiate the electromagnetic wave. In some embodiments, at least some of the functional modules of the mobile communication module 150 may be disposed in the processor 110. In some embodiments, at least some of the functional modules of the mobile communication module 150 may be provided in the same device as at least some of the modules of the processor 110.

The modem processor may include a modulator and a demodulator. The modulator is used for modulating a low-frequency baseband signal to be transmitted into a medium-high frequency signal. The demodulator is used for demodulating the received electromagnetic wave signal into a low-frequency baseband signal. The demodulator then passes the demodulated low frequency baseband signal to a baseband processor for processing. The low frequency baseband signal is processed by the baseband processor and then transferred to the application processor. The application processor outputs a sound signal through an audio device (not limited to the speaker 170A, the receiver 170B, etc.) or displays an image or video through the display screen 194. In some embodiments, the modem processor may be a stand-alone device. In other embodiments, the modem processor may be provided in the same device as the mobile communication module 150 or other functional modules, independent of the processor 110.

The wireless communication module 160 may provide a solution for wireless communication applied on the end-side device 100, including Wireless Local Area Networks (WLANs), such as Wi-Fi networks, Bluetooth (BT), BLE broadcasting, Global Navigation Satellite System (GNSS), Frequency Modulation (FM), Near Field Communication (NFC), infrared (infrared, IR), and the like. The wireless communication module 160 may be one or more devices integrating at least one communication processing module. The wireless communication module 160 receives electromagnetic waves via the antenna 2, performs frequency modulation and filtering processing on electromagnetic wave signals, and transmits the processed signals to the processor 110. The wireless communication module 160 may also receive a signal to be transmitted from the processor 110, perform frequency modulation and amplification on the signal, and convert the signal into electromagnetic waves through the antenna 2 to radiate the electromagnetic waves.

In some embodiments, antenna 1 of the end-side device 100 is coupled to the mobile communication module 150 and antenna 2 is coupled to the wireless communication module 160 such that the end-side device 100 can communicate with networks and other devices through wireless communication techniques. The wireless communication technology may include global system for mobile communications (GSM), General Packet Radio Service (GPRS), code division multiple access (code division multiple access, CDMA), Wideband Code Division Multiple Access (WCDMA), time-division code division multiple access (time-division code division multiple access, TD-SCDMA), Long Term Evolution (LTE), LTE, BT, GNSS, WLAN, NFC, FM, and/or IR technologies, etc. The GNSS may include a Global Positioning System (GPS), a global navigation satellite system (GLONASS), a beidou navigation satellite system (BDS), a quasi-zenith satellite system (QZSS), and/or a Satellite Based Augmentation System (SBAS).

The end-side device 100 implements display functions through the GPU, the display screen 194, and the application processor, etc. The GPU is a microprocessor for image processing, and is connected to the display screen 194 and an application processor. The GPU is used to perform mathematical and geometric calculations for graphics rendering. The processor 110 may include one or more GPUs that execute program instructions to generate or alter display information.

The display screen 194 is used to display images, video, and the like. The display screen 194 includes a display panel. The display panel may adopt a Liquid Crystal Display (LCD), an organic light-emitting diode (OLED), an active-matrix organic light-emitting diode (active-matrix organic light-emitting diode, AMOLED), a flexible light-emitting diode (FLED), a miniature, a Micro-oeld, a quantum dot light-emitting diode (QLED), and the like. In some embodiments, the end-side device 100 may include 1 or N display screens 194, N being a positive integer greater than 1.

The end-side device 100 may implement a camera function via the ISP, camera 193, video codec, GPU, display screen 194, application processor, etc.

The ISP is used to process the data fed back by the camera 193. For example, when a photo is taken, the shutter is opened, light is transmitted to the camera photosensitive element through the lens, the optical signal is converted into an electrical signal, and the camera photosensitive element transmits the electrical signal to the ISP for processing and converting into an image visible to naked eyes. The ISP can also carry out algorithm optimization on noise, brightness and skin color of the image. The ISP can also optimize parameters such as exposure, color temperature and the like of a shooting scene. In some embodiments, the ISP may be provided in camera 193.

The camera 193 is used to capture still images or video. The object generates an optical image through the lens and projects the optical image to the photosensitive element. The photosensitive element may be a Charge Coupled Device (CCD) or a complementary metal-oxide-semiconductor (CMOS) phototransistor. The light sensing element converts the optical signal into an electrical signal, which is then passed to the ISP where it is converted into a digital image signal. And the ISP outputs the digital image signal to the DSP for processing. The DSP converts the digital image signal into image signal in standard RGB, YUV and other formats. In some embodiments, the end-side device 100 may include 1 or N cameras 193, N being a positive integer greater than 1.

The digital signal processor is used for processing digital signals, and can process digital image signals and other digital signals. For example, when the end-side device 100 is in frequency bin selection, the digital signal processor is used to perform fourier transform or the like on the frequency bin energy.

Video codecs are used to compress or decompress digital video. The end-side device 100 may support one or more video codecs. In this way, the end-side device 100 may play or record video in a variety of encoding formats, such as: moving Picture Experts Group (MPEG) 1, MPEG2, MPEG3, MPEG4, and the like.

The NPU is a neural-network (NN) computing processor that processes input information quickly by using a biological neural network structure, for example, by using a transfer mode between neurons of a human brain, and can also learn by itself continuously. Applications such as intelligent awareness of the peer-to-peer device 100 can be implemented by the NPU, for example: image recognition, face recognition, speech recognition, text understanding, and the like.

The external memory interface 120 may be used to connect an external memory card, such as a Micro SD card, to implement the storage capability of the expansion end-side device 100. The external memory card communicates with the processor 110 through the external memory interface 120 to implement a data storage function. For example, files such as music, video, etc. are saved in an external memory card.

The internal memory 121 may be used to store computer-executable program code, which includes instructions. The processor 110 executes various functional applications of the end-side device 100 and data processing by executing instructions stored in the internal memory 121. The internal memory 121 may include a program storage area and a data storage area. The storage program area may store an operating system, an application program (such as a sound playing function, an image playing function, and the like) required by at least one function, and the like. The storage data area may store data (such as audio data, a phonebook, etc.) created during use of the end-side device 100, and the like. In addition, the internal memory 121 may include a high speed random access memory, and may also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, a Universal Flash Storage (UFS), and the like.

The end-side device 100 may implement audio functions via an audio module 170, a speaker 170A, a receiver 170B, a microphone 170C, an earphone interface 170D, and an application processor, among others. Such as music playing, recording, etc.

The audio module 170 is used to convert digital audio information into an analog audio signal output and also to convert an analog audio input into a digital audio signal. The audio module 170 may also be used to encode and decode audio signals. In some embodiments, the audio module 170 may be disposed in the processor 110, or some functional modules of the audio module 170 may be disposed in the processor 110.

The speaker 170A, also called a "horn", is used to convert the audio electrical signal into a sound signal. The end-side device 100 can listen to music through the speaker 170A or listen to a hands-free conversation.

The receiver 170B, also called "earpiece", is used to convert the electrical audio signal into an acoustic signal. When the end-side device 100 answers a call or voice information, the voice can be answered by placing the receiver 170B close to the human ear.

The microphone 170C, also referred to as a "microphone," is used to convert sound signals into electrical signals. When making a call or transmitting voice information, the user can input a voice signal to the microphone 170C by speaking the user's mouth near the microphone 170C. The end-side device 100 may be provided with at least one microphone 170C. In other embodiments, the end-side device 100 may be provided with two microphones 170C, which may implement a noise reduction function in addition to collecting sound signals. In other embodiments, the end-side device 100 may further include three, four or more microphones 170C for collecting sound signals, reducing noise, identifying sound sources, performing directional recording, and the like.

The headphone interface 170D is used to connect a wired headphone. The headset interface 170D may be the USB interface 130, or may be a 3.5mm open mobile electronic device platform (OMTP) standard interface, a cellular telecommunications industry association (cellular telecommunications industry association of the USA, CTIA) standard interface.

The pressure sensor 180A is used for sensing a pressure signal, and converting the pressure signal into an electrical signal. In some embodiments, the pressure sensor 180A may be disposed on the display screen 194.

The gyro sensor 180B may be used to determine the motion posture of the end-side device 100. In some embodiments, the angular velocity of the end-side device 100 about three axes (i.e., x, y, and z axes) may be determined by the gyroscope sensor 180B. The gyro sensor 180B may be used for photographing anti-shake. The gyroscope sensor 180B may also be used for navigation, somatosensory gaming scenes.

The air pressure sensor 180C is used to measure air pressure. In some embodiments, the end-side device 100 calculates altitude, aiding in positioning and navigation, from barometric pressure values measured by the barometric pressure sensor 180C.

The magnetic sensor 180D includes a hall sensor. The end-side device 100 may detect the opening and closing of the flip holster using the magnetic sensor 180D.

The acceleration sensor 180E can detect the magnitude of acceleration of the end-side device 100 in various directions (generally, three axes). The magnitude and direction of gravity can be detected when the end-side device 100 is stationary. The method can also be used for recognizing the posture of the equipment at the end side, and is applied to horizontal and vertical screen switching, pedometers and other applications.

A distance sensor 180F for measuring a distance. The end-side device 100 may measure distance by infrared or laser. In some embodiments, taking a scene, the end-side device 100 may utilize the range sensor 180F to range to achieve fast focus.

The proximity light sensor 180G may include, for example, a Light Emitting Diode (LED) and a light detector, such as a photodiode. The light emitting diode may be an infrared light emitting diode. The end-side device 100 emits infrared light outward through the light emitting diode. The end-side device 100 detects infrared reflected light from a nearby object using a photodiode to automatically extinguish the screen for power saving purposes. The proximity light sensor 180G may also be used in a holster mode, a pocket mode automatically unlocks and locks the screen.

The ambient light sensor 180L is used to sense the ambient light level. The front-side device 100 may adaptively adjust the brightness of the display screen 194 according to the perceived ambient light level. The ambient light sensor 180L may also be used to automatically adjust the white balance when taking a picture. The ambient light sensor 180L may also cooperate with the proximity light sensor 180G to detect whether the end-side device 100 is in a pocket to prevent accidental touches.

The fingerprint sensor 180H is used to collect a fingerprint. The end-side device 100 may utilize the collected fingerprint characteristics to implement fingerprint unlocking, access an application lock, fingerprint photographing, fingerprint incoming call answering, and the like.

The temperature sensor 180J is used to detect temperature. In some embodiments, the end-side device 100 executes a temperature processing strategy using the temperature detected by the temperature sensor 180J.

The touch sensor 180K is also referred to as a "touch panel". The touch sensor 180K may be disposed on the display screen 194, and the touch sensor 180K and the display screen 194 form a touch screen, which is also called a "touch screen". The touch sensor 180K is used to detect a touch operation applied thereto or nearby. The touch sensor can communicate the detected touch operation to the application processor to determine the touch event type. Visual output associated with the touch operation may be provided through the display screen 194. In other embodiments, the touch sensor 180K can also be disposed on a surface of the end-side device 100 at a different location than the display screen 194.

The bone conduction sensor 180M may acquire a vibration signal. In some embodiments, the bone conduction sensor 180M may acquire a vibration signal of the human vocal part vibrating the bone mass.

The keys 190 include a power-on key, a volume key, and the like. The keys 190 may be mechanical keys. Or may be touch keys. The end-side device 100 may receive key inputs, generating key signal inputs related to user settings and function controls of the end-side device 100.

The motor 191 may generate a vibration cue. The motor 191 may be used for incoming call vibration cues, as well as for touch vibration feedback. For example, touch operations applied to different applications (e.g., photographing, audio playing, etc.) may correspond to different vibration feedback effects. The motor 191 may also respond to different vibration feedback effects for touch operations applied to different areas of the display screen 194. Different application scenes (such as time reminding, receiving information, alarm clock, game and the like) can also correspond to different vibration feedback effects. The touch vibration feedback effect may also support customization.

Indicator 192 may be an indicator light that may be used to indicate a state of charge, a change in charge, or a message, missed call, notification, etc.

The SIM card interface 195 is used to connect a SIM card. The SIM card can be brought into and out of contact with the end-side device 100 by being inserted into the SIM card interface 195 or being pulled out of the SIM card interface 195. The end-side device 100 may support 1 or N SIM card interfaces, N being a positive integer greater than 1. The SIM card interface 195 may support a Nano SIM card, a Micro SIM card, a SIM card, etc. Multiple cards can be inserted into the same SIM card interface 195 at the same time. The types of the plurality of cards may be the same or different. The SIM card interface 195 may also be compatible with different types of SIM cards. The SIM card interface 195 may also be compatible with external memory cards. The end-side device 100 interacts with the network through the SIM card to implement functions such as a call and data communication. In some embodiments, the end-side device 100 employs esims, namely: an embedded SIM card. The eSIM card may be embedded in the end-side device 100 and may not be separable from the end-side device 100.

The software system of the peer device 100 may employ a layered architecture, an event-driven architecture, a micro-core architecture, a micro-service architecture, or a cloud architecture. The embodiment of the present invention takes an Android system with a layered architecture as an example, and exemplarily illustrates a software structure of the end-side device 100. Fig. 3 is a block diagram of a software configuration of the end-side device 100 according to an embodiment of the present application. The layered architecture divides the software into several layers, each layer having a clear role and division of labor. The layers communicate with each other through a software interface. In some embodiments, the Android system is divided into four layers, an application layer, an application framework layer, an Android runtime (Android runtime) and system library, and a kernel layer from top to bottom.

The application layer may include a series of application packages. As shown in fig. 3, the application package may include applications such as camera, gallery, calendar, phone call, map, navigation, WLAN, bluetooth, music, video, short message, etc.

The application framework layer provides an Application Programming Interface (API) and a programming framework for the application program of the application layer. The application framework layer includes a number of predefined functions. As shown in FIG. 3, the application framework layers may include a window manager, content provider, view system, phone manager, resource manager, notification manager, and the like.

The window manager is used for managing window programs. The window manager can obtain the size of the display screen, judge whether a status bar exists, lock the screen, intercept the screen and the like.

The content provider is used to store and retrieve data and make it accessible to applications. The data may include video, images, audio, calls made and received, browsing history and bookmarks, phone books, etc.

The view system includes visual controls such as controls to display text, controls to display pictures, and the like. The view system may be used to build applications. The display interface may be composed of one or more views. For example, the display interface including the short message notification icon may include a view for displaying text and a view for displaying pictures.

The phone manager is used to provide the communication function of the end-side device 100. Such as management of call status (including on, off, etc.).

The resource manager provides various resources for the application, such as localized strings, icons, pictures, layout files, video files, and the like.

The notification manager enables the application to display notification information in the status bar, can be used to convey notification-type messages, can disappear automatically after a short dwell, and does not require user interaction. Such as a notification manager used to inform download completion, message alerts, etc. The notification manager may also be a notification that appears in the form of a chart or scroll bar text at the top status bar of the system, such as a notification of a background running application, or a notification that appears on the screen in the form of a dialog window. For example, prompting text information in the status bar, sounding a prompt tone, vibrating the electronic device, flashing an indicator light, etc.

The Android Runtime comprises a core library and a virtual machine. The Android runtime is responsible for scheduling and managing an Android system.

The core library comprises two parts: one part is a function which needs to be called by java language, and the other part is a core library of android.

The application layer and the application framework layer run in a virtual machine. The virtual machine executes java files of the application layer and the application framework layer as binary files. The virtual machine is used for performing the functions of object life cycle management, stack management, thread management, safety and exception management, garbage collection and the like.

The system library may include a plurality of functional modules. For example: surface managers (surface managers), Media Libraries (Media Libraries), three-dimensional graphics processing Libraries (e.g., OpenGL ES), 2D graphics engines (e.g., SGL), and the like.

The surface manager is used to manage the display subsystem and provide fusion of 2D and 3D layers for multiple applications.

The media library supports a variety of commonly used audio, video format playback and recording, and still image files, among others. The media library may support a variety of audio-video encoding formats, such as MPEG4, h.264, MP3, AAC, AMR, JPG, PNG, and the like.

The three-dimensional graphic processing library is used for realizing three-dimensional graphic drawing, image rendering, synthesis, layer processing and the like.

The 2D graphics engine is a drawing engine for 2D drawing.

The kernel layer is a layer between hardware and software. The inner core layer at least comprises a display driver, a camera driver, an audio driver and a sensor driver.

The structure of the server 200 will be described below. Referring to fig. 4, fig. 4 is a schematic structural diagram of a server 200 according to an embodiment of the present disclosure.

The server 200 includes: the processor 201, the communication interface 202, and the memory 203 are connected to each other via an internal bus 204.

The processor 201 may be formed by one or more general-purpose processors, such as a Central Processing Unit (CPU), or a combination of a CPU and a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a Programmable Logic Device (PLD), or a combination thereof. The PLD may be a Complex Programmable Logic Device (CPLD), a field-programmable gate array (FPGA), a General Array Logic (GAL), or any combination thereof.

The bus 204 may be a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The bus 204 may be divided into an address bus, a data bus, a control bus, and the like. For ease of illustration, only one thick line is shown in FIG. 2, but not only one bus or type of bus.

The memory 203 may include a volatile memory (volatile memory), such as a Random Access Memory (RAM); the memory 203 may also include a non-volatile memory (non-volatile memory), such as a read-only memory (ROM), a flash memory (flash memory), a Hard Disk Drive (HDD), or a solid-state drive (SSD); the memory 203 may also comprise a combination of the above categories.

For a better understanding of the solutions provided in the present application, the following description is provided for the relevant terms to which the embodiments of the present application relate:

the term, also referred to herein as a word or phrase, is the sum of all or a particular range of words and/or phrases in a language (including chinese, english, etc.). In the embodiments of the present application, if not stated, the "word" may mean a word or a phrase. The words include "characters" and "words" in chinese characters, and "words" in languages such as english.

The semantics of a word are a set of characteristic information used to describe the word. Wherein, the characteristic information of the word may include but is not limited to at least one of the following: meaning of words, parts of speech (e.g., noun, adjective, etc.), near-sense words, and anti-sense words, etc. For example, semantic information for "beauty" may include: meaning "nice looking, i.e. close to perfect or ideal boundaries in form, proportion, layout, style, color or sound, making the various senses extremely pleasant"; the part of speech is an adjective; the word "beautiful" in the near sense; the antisense word is "ugly" or the like. The semantic information of the vocabulary may include the characteristic information of the word included in the vocabulary, and the semantic information of the vocabulary in the present application mostly indicates the meaning of the vocabulary.

The word vector, which may also be referred to as a word feature vector, is a vector formed by numbers to which feature information of words is mapped, and is used to represent feature information of words.

Unknown words: i.e. words which are not included in the participle vocabulary (or the word vector library) but have to be separated out, including various proper nouns (names of people, places, names of enterprises, etc.), abbreviations, newly added words, etc.

Vocabulary of the old word new meaning: have been included in word segmentation word lists (or word vector libraries), but have been given new meanings due to changes in application scenarios. For example, generally "mushroom" is a term, which is a food. However, in some situations, people now express "mushroom" as a verb, meaning to cry.

The language model is a probability distribution model constructed for the text corpus, is used for calculating the probability of the language sequence corresponding to the text corpus, and can be used for judging whether a language sequence is a normal sentence, namely judging whether a text corpus is smooth. Fig. 5 is a schematic diagram of a language model, where the language model includes an embedding (embedding) layer, a neural network layer, and an output layer, where the embedding layer is configured to receive vocabularies in a vocabulary and output word vectors corresponding to the vocabularies; the neural network layer is used for processing and analyzing the word vector; the output layer outputs the probabilities of the language sequences. It should be understood that the language model mentioned in this application includes, but is not limited to, the embedding layer being a word2vec model or a Glove model, and the neural network layer being a Recurrent Neural Network (RNN) or a long-term memory neural network (LSTM). Illustratively, when the embedding layer receives the linguistic data to be processed as 'I want to eat mushrooms', the 'I' word vector, the 'want' word vector, the 'eat' word vector and the 'mushroom' word vector are respectively output. The neural network layer processes and analyzes the word vector of 'I', the word vector of 'thought' and the word vector of 'eating' and the word vector of 'mushroom', and adds a start character < s > before the head word of the language sequence 'I' in order to make the conditional probability of the head word 'I' meaningful. The output layer outputs the conditional probability of the occurrence of the word vector of ' I ', ' thinking ' as P (think ' < s >, I), ' eating ' as P (eat ' < s >, I, want), and ' mushroom ' as P (mushroom ' < s >, I, want, eat).

The word vector construction method provided by the embodiment of the present application is further described in detail below:

referring to fig. 6, fig. 6 is a schematic flowchart of a word vector construction method according to an embodiment of the present application. As shown in fig. 6, the word vector construction method includes the following steps 601 to 604. The method execution subject shown in fig. 6 may be a server or a chip in the server. Fig. 6 illustrates an execution subject of the word vector construction method by using a server as an example. Wherein:

601. the method comprises the steps of obtaining a new word, wherein the new word is a vocabulary of an unknown word or a new meaning of an old word, the new word has a first semantic meaning, the unknown word is a vocabulary without a word vector in a server, the vocabulary of the new meaning of the old word is a vocabulary with a word vector in the server, and the semantic meaning indicated by the word vector is different from the first semantic meaning.

The server can receive a word vector construction instruction input by an operator and acquire a new word carried in the word vector construction instruction. For example, in an application scenario, an operator performs internet search on a network hot word in a fixed time period, or performs internet search according to a network hot event, and determines whether a new word appears according to a search result. And if the operator determines that a new word appears, sending a word vector construction instruction aiming at the new word to the server. Illustratively, the operator finds the word "orlistat" by means of internet search, and determines that the word "orlistat" is an unknown word according to the common sense (or knowledge reserve), i.e. the word vector library of the server does not contain the words of the word vectors thereof, and then the operator sends a word vector construction instruction for the word "orlistat" to the server. Further, the server may obtain the new word "olympic from the word vector build instruction.

It should be understood that the number of new words obtained by the server may be obtained by a single word, or obtained by the server in batch, which is not limited to this application.

602. And determining a word vector of a second semantic contained in each similar meaning word based on the word vector of at least one similar meaning word corresponding to the new word, wherein the second semantic corresponds to the first semantic.

After the server acquires the new word, the operation/development personnel lists the similar meaning word of the new word according to self knowledge storage or common knowledge. The server obtains at least one near-meaning word corresponding to the new word listed by the operator/developer, and obtains a word vector of each near-meaning word from the word stock, and further, the server may determine a word vector of a second semantic included in each near-meaning word based on the word vector of the near-meaning word.

It should be noted that the correspondence between the second semantic meaning and the first semantic meaning may include a case where the second semantic meaning is the same as the first semantic meaning, such as "mom" and "mom", and may also include a case where the second semantic meaning is similar to the first semantic meaning, such as "draw" and "grasp" both mean "understand, appreciate and know", but "draw" focuses on emotional experience and appreciation, and "grasp" focuses on rational understanding.

In one possible implementation, the server may determine, for each of the at least one near word, whether the near word contains a third semantic meaning other than the second semantic meaning. If the similar meaning word contains a third semantic, acquiring a word vector corresponding to the third semantic, and determining a word vector of a second semantic contained in the similar meaning word based on the word vector of the similar meaning word and the word vector corresponding to the third semantic.

After the server acquires the word vector of each similar meaning word, it determines whether the similar meaning word is a polysemous word (i.e. the similar meaning word contains a third semantic meaning in addition to the second semantic meaning) for each similar meaning word. In other words, the server determines whether the near word contains two types of semantics (a second semantic and a third semantic). Wherein the second semantic is the same or similar to the first semantic of the new word; the third semantic is a semantic that is different or dissimilar from the first semantic of the new word, and the third semantic may be one or more. If the synonym is a polysemous word and the third semantic meaning is definite (i.e. the word vector of the synonym or the synonym corresponding to the third semantic meaning can be obtained from the word vector library), but the second semantic meaning is not definite (i.e. the word vector of the synonym or the synonym corresponding to the second semantic meaning cannot be obtained from the word vector library), the server obtains the word vector of the synonym or the synonym corresponding to the third semantic meaning from the word vector library according to the third semantic meaning contained in the synonym, and determines the word vector of the third semantic meaning based on the word vector of the synonym or the synonym corresponding to the third semantic meaning. Further, the server obtains the word vector of the similar meaning word from a word vector library, and decomposes the word vector of the similar meaning word based on the word vector corresponding to the third semantic to obtain the word vector of the second semantic contained in the similar meaning word.

It should be understood that the specific manner in which the server determines whether the similar meaning word is an ambiguous word is not specifically limited in this application, and for example, the operator may determine whether the similar meaning word is an ambiguous word (or has multiple semantics) according to common general knowledge.

Illustratively, the operator sends the service a new word of the word vector to be constructed as "orlistat", and lists the synonym of "orlistat" as "refuel". The server outputs the first prompt message of 'fueling' and whether the message is an ambiguous word? And inputting an operation instruction of 'being an ambiguous word' by the operator according to the first prompt message. The server triggers and outputs second prompt information ' please input similar meaning words of other semantics ' based on the operation instruction ', and the operator inputs the similar meaning words of other semantics (namely the third semantics indicated in the foregoing) according to the second prompt information to be ' gasoline adding ' and ' fuel adding '. As shown in fig. 7, fig. 7 is a schematic diagram of word vector decomposition, where the server obtains a word vector of a near word "add gasoline" and a word vector of "add fuel" in a third semantic, calculates an average word vector Y of the word vector of "add gasoline" and the word vector of "add fuel", and determines the average word vector Y as the word vector of the third semantic. Further, the server obtains the word vector X of 'refueling', and removes a third semantic word vector Y in the word vector X of 'refueling' according to a formula (1) to obtain a second semantic word vector Z.

Wherein Z is a word vector Z of the second semantic, X is a word vector X of 'refuel', Y is a word vector Y of the third semantic, X.Y is the inner product of the word vector X and the word vector Y, | | Y | | survival₂Is the two-norm of the word vector Y.

In one possible implementation, if the near word does not contain at least one third semantic, the word vector of the near word is determined to be the word vector of the second semantic.

The server outputs prompt information to prompt an operator to determine whether the similar meaning word is a polysemous word, and the operator inputs an operation instruction of 'not being a polysemous word' according to the prompt information, so that the server determines that the similar meaning word does not contain two types of semantics, namely when the similar meaning word only contains the second semantic, the server can determine the word vector of the similar meaning word as the word vector of the second semantic.

603. And constructing a word vector of the new word based on the number of the similar meaning words corresponding to the new word and the word vector of the second semantic meaning contained in each similar meaning word.

The server may calculate an average word vector of the word vectors of the second semantics included in the similar meaning words according to the number of the similar meaning words corresponding to the new word and the word vectors of the second semantics included in each similar meaning word. For example, the number of the near-meaning words corresponding to the new word is 3, which are respectively near-meaning word 1, near-meaning word2 and near-meaning word 3, and the word vectors of the second semantics corresponding to the near-meaning words are respectively word vector 1, word vector 2 and word vector 3, then the server calculates the average word vector of word vector 1, word vector 2 and word vector 3, and determines the average word vector as the word vector of the new word.

In one example, as shown in fig. 8, a schematic diagram of a method for constructing a word vector for an unknown word is shown. When the word of the word vector to be constructed is the unknown word "blue-thin", the server obtains the synonyms "difficult", "sad" and painful "input by the operator based on the common knowledge, and because the" difficult "," sad "and painful" only contain one semantic (i.e. the aforementioned second semantic), the server can obtain the word vector of the "difficult", "sad" and the word vector of the "painful" from the corresponding word vector library thereof as the word vectors of the second semantic contained in the "difficult", "sad" and "painful", i.e. calculate the average word vector of the "difficult", "sad" and the word vector of the "painful", and determine the average word vector as the blue-thin word vector. And the server stores the constructed word vectors of the blue-thin type in a word vector library corresponding to the server.

In another example, as shown in FIG. 9, a schematic diagram of a method of constructing a word vector for a vocabulary of a new sense of an old word is shown. When the word of the word vector to be constructed is a newly-defined word "shiitake mushroom", the server inputs the word vector "want to cry" and "want to cry without tearing" based on common knowledge according to the operator, and since both the word vector "want to cry" and the word vector "want to cry without tearing" only include one semantic (namely, the second semantic), the server can obtain the word vector "want to cry" and the word vector "want to cry without tearing" from the word vector library corresponding to the server as the word vector of the second semantic included in the word vector "want to cry without tearing" respectively, that is, the average word vector of the word vector "want to cry" and the word vector "want to cry without tearing" is calculated, and the server determines the average word vector as the word vector of the shiitake mushroom. Since the "mushroom" is a new word meaning, that is, the word vector of the "mushroom" exists in the word vector library corresponding to the server (for example, the word vector corresponding to the mushroom 1 in the word vector library of fig. 9), the server determines the average word vector as the word vector of the "mushroom", and stores the word vector of the "mushroom" in the word vector library corresponding to the server by using the identifier of the mushroom 2.

In one possible implementation, the server obtains the weight value corresponding to each similar meaning word, and constructs the word vector of the new word based on the weight value corresponding to each similar meaning word, the number of the similar meaning words corresponding to the new word, and the word vector of the second semantic contained in each similar meaning word. The weight value corresponding to each similar meaning word is set by an operator according to the correlation degree between the semantics of the similar meaning word and the semantics of the new word of the word vector to be constructed, and can be adjusted according to specific conditions without limiting the application.

Illustratively, a specific method for constructing a word vector of a new word based on a weight value corresponding to each near-meaning word, the number of near-meaning words corresponding to the new word, and a word vector of a second semantic meaning included in each near-meaning word may be as shown in formula (2).

Wherein n is the number of similar meaning words corresponding to the new word, V₁For a word vector of the second semantic meaning contained in the first near word corresponding to the new word, V₂Word vectors of a second semantic meaning, V, included for the new word in correspondence with the second synonym_nWord vectors of the second semantic meaning, ω, contained for the new word in correspondence with the nth near word₁The weight, ω, included for the new word corresponding to the first synonym₂Weight, ω, included for the new word in correspondence with the second synonym_nThe new word is associated with the weight contained in the nth synonym,

and the average word vector contained in the corresponding similar meaning word of the new word is also the word vector of the new word constructed by the server.

604. The word vector for the new word is sent to the end-side device.

The server sends the constructed word vector of the new word to the end-side equipment, so that the end-side equipment can perform related tasks of natural language processing according to the word vector of the new word, and the processing capacity of the natural language processing task of the end-side equipment is further improved. It should be understood that the server may send the word vector of at least one new word to the end-side device according to a preset time period, where the preset time period is set by a developer/operator according to a specific application scenario, and is not specifically limited herein. The server may also create a word vector for a new word and send the word vector for the new word to the end-side device.

It can be seen that by implementing the word vector construction method described in fig. 6, the end-side device may decompose the word vector of the near-meaning word corresponding to the new word, and determine, from the decomposed word vector of the near-meaning word, a word vector corresponding to a semantic (a second semantic included in the near-meaning word) that is the same as or similar to the lexical semantic (a first semantic) of the word vector to be constructed. Further, the server can construct a word vector of the new word through a word vector corresponding to the second semantic contained in the near-meaning word, and the accuracy of the constructed word vector of the new word can be improved.

Referring to fig. 10, fig. 10 is a schematic flowchart illustrating a semantic recognition method according to an embodiment of the present disclosure. As shown in fig. 6, the semantic recognition method includes the following steps 1001 to 1005. The method execution subjects shown in fig. 10 may be end-side devices or chips in end-side devices. Fig. 10 illustrates an execution subject of the semantic recognition method as an end-side device.

Wherein:

1001. receiving a word vector of a new word sent by a server, wherein the new word has a first semantic meaning, the new word is a vocabulary of an unknown word or a new meaning of an old word, the vocabulary of the word vector does not exist in the unknown word server, the vocabulary of the new meaning of the old word is the vocabulary of the word vector existing in the server, and the semantic meaning indicated by the word vector is different from the first semantic meaning.

The server sends the word vector of the new word to the end-side device, and the end-side device can output prompt information for prompting the user to input instruction information whether to receive the word vector of the new word. And if the end-side equipment receives instruction information for receiving the word vector of the new word input by the user, receiving the word vector of the new word sent by the server. And if the end-side equipment does not receive instruction information for receiving the word vector of the new word input by the user, ignoring the word vector of the new word sent by the server. It should be understood that, for a specific implementation of the server sending the word vector of the new word to the end-side device, reference may be made to a specific implementation of step 604 in the foregoing embodiment, and redundant description is not repeated here.

1002. And storing the word vector of the new word in a second word vector library.

The end-side device stores the word vector for the new word in a second word vector repository. It should be noted that the second word vector library may be a word vector library that only contains word vectors of new words, in other words, there is no intersection data between the second word vector library and a preset word vector library (i.e., the first word vector library) when the end-side device leaves the factory or the system is installed. In this way, the word vector 1 corresponding to the new word of the old word in the new word sent by the server (i.e. the word vector corresponding to the new semantic) and the word vector 2 corresponding to the new word in the first word vector library (i.e. the word vector corresponding to the old semantic) can be effectively distinguished. It should be understood that after the end-side device stores the word vector of the new word, the application scenarios of natural language processing such as intelligent question answering, text classification, natural language generation, emotion classification, and dialogue management can be performed according to the word vector of the new word.

1003. And segmenting the linguistic data to be processed to obtain at least one participle.

The end-side device receives the corpus to be processed, and segments the corpus to be processed based on a preset word segmentation model to obtain at least one word segmentation (also called as a vocabulary). The preset word segmentation model is obtained by training the existing word segmentation model according to the experimental corpus for developers, and the method is not specifically limited, for example: the existing word segmentation model can be a knot word segmentation model and the like.

Illustratively, when the corpus to be processed is received by the end-side device as "shiitake mushroom today", the corpus to be processed may be segmented according to a preset segmentation model to obtain 3 segmentations: today, the mushroom is very much and shiitake.

1004. If the first target vocabulary has a first word vector in the first word vector library and the first target vocabulary has a second word vector in the second word vector library, determining the language probability corresponding to the corpus to be processed based on the first word vector or the second word vector, wherein the first target vocabulary is any vocabulary in at least one participle.

And the end-side equipment matches each participle in at least one participle obtained by segmenting the linguistic data to be processed with the second word vector library, and if the matching is successful, the vocabulary is determined to be the first target vocabulary. And then, the end-side device matches the first target vocabulary with the first word vector library, and if the matching is successful, determines a vocabulary containing an old word new meaning in the corpus to be processed, and further, the end-side device can determine a language probability corresponding to the corpus to be processed based on the first word vector/the second word vector and the language model, in other words, the end-side device calls the language model and the word vector of each participle in the corpus to be processed in the first word vector library or the word vector in the second word vector library to obtain the language probability of the corpus to be processed. It is necessary to know that the corresponding language probability is larger when the corpus is smooth and smaller when the corpus is not smooth.

Illustratively, the first and second word vector libraries are shown in table 1. Wherein, the first word vector of the lentinus edodes in the first word vector library represents that the semantic is food, and the second word vector of the lentinus edodes in the second word vector library represents that the semantic is crying.

TABLE 1

The method comprises the steps that an end-side device divides a corpus of to-be-processed champignon today to obtain 3 participles of champignon today, champignon very today and champignon very much, the end-side device matches the word today, the word very much and the word champignon very much with a second word vector library respectively, the word champignon matches the word vector library successfully, and the word champignon very much is determined as a first target word (namely a new word, possibly an unknown word or possibly a word with an old word new meaning) by the end-side device. Further, the end-side device matches the mushroom with the first word vector library, and if the matching is successful, the end-side device determines that the mushroom is a new word of the old word in the new word. However, "mushroom" may be a new meaning in the corpus to be processed, corresponding to the second vector in the second word vector library, or "mushroom" may be an old meaning, corresponding to the first word vector in the first word vector library. Further, the end-side device may call the language model to obtain the language probability of the corpus to be processed, namely "shiitake mushroom today", according to the first word vector corresponding to "shiitake mushroom", "word vector 1 corresponding to" today ", and" very "word vector 3 in the first word vector library. Or the end-side device may call the language model to obtain the language probability of the corpus to be processed, namely "shiitake mushroom today", according to the first word vector corresponding to "shiitake mushroom", the word vector 1 corresponding to "today", and the word vector 3 corresponding to "very much".

1005. And determining the first word vector or the second word vector as a corresponding word vector of the first target vocabulary in the linguistic data to be processed based on the language probability.

The end-side device compares the language probability of the corpus to be processed obtained in step 1004 with a probability threshold, and determines a first word vector or a second word vector as a word vector corresponding to the first target vocabulary in the corpus to be processed according to the comparison result. In other words, if the end-side device determines, according to the comparison result, that the first word vector is used as the word vector corresponding to the first target vocabulary in the corpus to be processed, it indicates that the semantic meaning of the first target vocabulary expressed in the corpus to be processed is the old meaning of the vocabulary. And if the end-side equipment determines that the second word vector is used as the word vector corresponding to the first target vocabulary in the corpus to be processed according to the comparison result, the semantic meaning expressed by the first target vocabulary in the corpus to be processed is the new meaning of the vocabulary.

The probability threshold is measured and calculated by the server according to the experimental corpus and the language model and then sent to the end-side equipment, and specific numerical values of the probability threshold are not specifically limited in the application. Fig. 11 is a schematic diagram illustrating a method for determining a probability threshold. The server calls a language model to calculate the language probability of each text corpus in the first text corpus set according to the first text corpus set, so as to obtain a probability density curve shown as 111 in fig. 11. The text corpus in the first text corpus set does not contain new-sense vocabulary of the old word, for example, the text corpus in the first text corpus set may be "cry very often today. "," refuel! "," he is very difficult to suffer. "and the like. The server calls according to the second text corpus setThe language model calculates the language probability of each text corpus in the second text corpus set to obtain a probability density curve as shown by 110 in fig. 11. The text corpora in the second text corpus set include new words, and the text corpora in the second text corpus set correspond to the text corpora in the first text corpus set, that is, the text corpora in the second text corpus set and the text corpora in the first text corpus set represent the same/similar meanings, and if the text corpora in the first text corpus set are "cry very much today", "oil and try" and "difficult for him to accept", the text corpora in the second text corpus set may be "very shiitake today", "easy for giving" and "very thin blue". It should be noted that, according to the statistical principle, the areas of the probability density curve 110 and the probability density curve 111, which are respectively surrounded by the horizontal axis, are both 1. In one application scenario, assuming that for any new-sense word of the old word, the number of the new-sense (i.e., the old word used newly) text corpora in the total text corpora containing the word is the same as the number of the old-sense (i.e., the old word used old) text corpora in the first text corpus set and the second text corpus set, the text corpora containing "shiitake" exemplarily has 1000 pieces in the first text corpus set and the second text corpus set, wherein "shiitake" in 500 text corpora indicates crying, and "shiitake" in the remaining 500 text corpora indicates food shiitake. The server determines the ratio of the number of correct judgments when the new meaning of the old word is expressed as new meaning to the total number of new meanings judged by the new meaning of the old word as precise rate, for example, the total number of times that the "mushroom" is judged as new meaning (i.e. wants to cry) is 700, wherein 500 times are correct, that is, 500 times of judgments in the 700 times of judgments indicate that the new meaning (i.e. wants to cry) is correct, the remaining 200 times of judgments that the "mushroom" indicates that the new meaning (i.e. wants to cry) is wrong, that 200 times of judgments that the "mushroom" is actually expressed as old meaning (i.e. instant mushroom), and the precise rate expression is accurate rate

In other words, the precision (Accuracy) formula can be represented by formula (3), and the Recall (Recall) formulaThe formula is shown in formula (4).

Recall＝1-S₂ (4)

Wherein S is₁Is the area, S, enclosed by the horizontal axis on the left side of the probability threshold k and the probability density curve 111 in FIG. 11₂The horizontal axis on the right side of the probability threshold k in fig. 11 is the area enclosed by the probability density curve 110. As can be seen in FIG. 11, S₁And S₂All can be regarded as dependent variables with probability threshold k as independent variable, namely S₁And S₂Respectively expressed as a functional expression S₁(k) Sum function expression S₂(k) In that respect In the classification problem of machine learning, the harmonic mean of the precision rate and the recall rate is usually used as a measure, and the measure is F₁-fraction (score), of the F₁The expression of-Score is shown in equation (5).

Further, the server may base the F on the solution₁-the maximum value of Score to determine the probability threshold k.

Fig. 12 is a schematic diagram illustrating another method for determining the probability threshold. The server may use the aforementioned first text corpus set to call a language model to calculate a language probability of each text corpus in the first text corpus set, where a probability density curve is shown as 120 in fig. 12, where k is a probability threshold to be obtained. To save computing resources, the server may determine the area S enclosed by the left horizontal axis of the k value and the probability density curve 120₃The area S enclosed by the horizontal axis and the probability density curve 120₄(according to the statistical principle, the area S enclosed by the probability density curve 120 and the horizontal axis₄1) and a preset value, wherein the preset value is set by a developer according to experimental measurement and calculation data, and the application is not particularly limited. In other words, when S₃And S₄When the ratio is a preset value, the ratio can be based on S₃Determines the k value (i.e., the probability threshold). The server can save the computing resources by the probability threshold value determination method.

In one possible implementation, if the first target vocabulary does not have word vectors in the first word vector library and the first target vocabulary has word vectors in the second word vector library, it is determined that the word vectors of the first target vocabulary in the second word vector library are the corresponding word vectors of the first target vocabulary in the corpus to be processed.

And the end-side equipment matches each participle in at least one participle obtained by segmenting the corpus to be processed with the second word vector library, and if the matching is successful, the vocabulary is determined to be the first target vocabulary. And then the end-side equipment matches the first target vocabulary with the first word vector library, and if the matching is unsuccessful, the end-side equipment determines that the word vector of the first target vocabulary in the second word vector library is the corresponding word vector of the first target vocabulary in the corpus to be processed.

Illustratively, the first and second banks of word vectors are as shown in Table 1 above. The method comprises the steps that the end-side device divides a corpus to be processed into 3 participles, namely, the 3 participles are today, the blue and thin, the end-side device matches the today, the very blue and thin with a second word vector library respectively, the blue and thin are successfully matched with the second word vector library, and the end-side device determines the blue and thin as a first target vocabulary (namely a new word, possibly an unregistered word or possibly a word with an old word and a new meaning). Further, the end-side device matches the "blue-thin" with the first word vector library, and if the matching is unsuccessful, the end-side device takes the word vector 4 of the "blue-thin" in the second word vector as the corresponding word vector in the corpus to be processed.

In one possible implementation, the end-side device calls a language model to determine the language probability corresponding to the corpus to be processed based on a first word vector of a first target vocabulary and a word vector of a second target vocabulary in a first word vector library, wherein the second target vocabulary is a vocabulary except the first target vocabulary in at least one word segmentation. Further, if the language probability is smaller than the probability threshold, determining a second word vector as a word vector corresponding to the first target vocabulary in the corpus to be processed. And if the language probability is greater than or equal to the probability threshold, determining a first word vector as a corresponding word vector of the first target vocabulary in the linguistic data to be processed.

Illustratively, the end-side device divides the corpus "today is very shiitake" to be processed to obtain 3 participles as "today", "very" and "shiitake", matches the "today", "very" and "shiitake" with the second word vector library respectively, and successfully matches the "shiitake" with the second word vector library, the end-side device determines the "shiitake" as a first target vocabulary (i.e. a new word, possibly an unregistered word or possibly an old word with a new meaning), and determines the "today" and "very" as a second target vocabulary. Further, the end-side device matches the mushroom with the first word vector library, and if the matching is successful, the end-side device determines that the mushroom is a new word of the old word in the new word. If the end-side equipment calls the language model to obtain the language probability a of the linguistic data to be processed, namely the champignon today, according to the first word vector corresponding to the champignon, the word vector 1 corresponding to the champignon today and the word vector 3 corresponding to the champignon today in the first word vector library. And if the probability threshold is b and a is smaller than b, determining a second word vector corresponding to the lentinus edodes in the second word vector library as a word vector corresponding to the first target vocabulary in the linguistic data to be processed.

In one possible implementation, the end-side device calls a language model based on a second word vector corresponding to the first target vocabulary and a word vector of the second target vocabulary in the first word vector library, and determines the language probability corresponding to the corpus to be processed, wherein the second target vocabulary is a vocabulary of the at least one participle except for the first target vocabulary. Further, if the language probability is greater than or equal to the probability threshold, determining a second word vector as a word vector corresponding to the first target vocabulary in the corpus to be processed. And if the language probability is smaller than the probability threshold, determining the first word vector as a word vector corresponding to the first target vocabulary in the linguistic data to be processed.

Illustratively, if the end-side device calls the language model to obtain the language probability c of the corpus to be processed, namely "shiitake today", according to the second word vector corresponding to "shiitake" in the second word vector library, the word vector 1 corresponding to "today" in the first word vector library, and the word vector 3 corresponding to "very much". And if the probability threshold is b and c is greater than or equal to b, determining a second word vector corresponding to the lentinus edodes in the second word vector library as a word vector corresponding to the first target vocabulary in the linguistic data to be processed.

Exemplarily, as shown in fig. 13, which is a schematic flow diagram of a semantic recognition method, an end-side device receives a corpus to be processed, and segments the corpus to be processed to obtain at least one participle, for each participle, the end-side device matches the participle with a second word vector library, and if the matching is unsuccessful, determines a word vector of the participle according to a first word vector library. And if the matching of the word segmentation and the second word vector library is successful, matching the word segmentation with the first word vector library, and if the matching is unsuccessful, determining the word vector of the word segmentation according to the second word vector library. If the word segmentation is successfully matched with the second word vector library and the word segmentation is successfully matched with the first word vector library, calling a language model, and determining the language probability of the linguistic data to be processed based on the first vector library. The end-side device compares the language probability with a probability threshold, and determines whether the participle is new for the old word (i.e. the vocabulary with the new meaning of the old word represents new meaning in the corpus to be processed) or old for the old word (i.e. the vocabulary with the new meaning of the old word represents old meaning in the corpus to be processed) according to the comparison result of the language probability and the probability threshold. If the language probability of the corpus to be processed is determined to be greater than or equal to the probability threshold value based on the first vector library, determining the first word vector of the participle from the first vector library as the corresponding word vector in the corpus to be processed, that is, the participle is used as an old word (that is, a new-sense vocabulary of the old word represents an old sense in the corpus to be processed). If the language probability of the linguistic data to be processed is determined to be smaller than the probability threshold value based on the first vector library, determining the second word vector of the participle from the second vector library as the corresponding word vector in the linguistic data to be processed, namely, the participle is used newly as an old word (namely, the vocabulary with the new meaning of the old word in the linguistic data to be processed represents that the meaning is the new meaning).

It can be seen that by implementing the semantic recognition method described in fig. 10, the end-side device can correctly recognize whether the vocabulary of the new meaning of the old word represents the new meaning (i.e. the new use case of the old word) or the old meaning (i.e. the old use case of the old word) in the corpus to be processed, and meanwhile, the word vector library of the end-side device is extended by continuously receiving the word vector of the new word from the server, so that the accuracy of semantic recognition is improved.

In the present application, it is noted that, in the specific implementation, some steps in the drawings may be selected for implementation, and the order of the steps in the drawings may also be adjusted for implementation, which is not limited in the present application. It should be understood that the specific implementation of some steps or the order of adjusting the steps in the figures is within the scope of the present application.

Referring to fig. 14, fig. 14 is a schematic structural diagram of a word vector constructing apparatus according to an embodiment of the present application. The word vector construction apparatus shown in fig. 14 may be used to implement part or all of the functions of the server in the embodiment of the method described in fig. 6. The word vector construction apparatus shown in fig. 14 may include an acquisition unit 1401, a determination unit 1402, a construction unit 1403, and a transmission unit 1404. Wherein:

an obtaining unit 1401, configured to obtain a new word, where the new word is an unknown word or a vocabulary of a new meaning of an old word, the new word has a first semantic meaning, the unknown word is a vocabulary without a word vector in the server, the vocabulary of the new meaning of the old word is a vocabulary with a word vector in the server, and a semantic meaning indicated by the word vector is different from the first semantic meaning;

a determining unit 1402, configured to determine, based on a word vector of at least one near-sense word corresponding to the new word, a word vector of a second semantic included in each near-sense word, where the second semantic corresponds to the first semantic;

a constructing unit 1403, configured to construct a word vector of the new word based on the number of near-sense words corresponding to the new word and a word vector of a second semantic included in each near-sense word;

a sending unit 1404, configured to send the word vector of the new word to an end-side device.

In a possible implementation, the determining unit 1402 is specifically configured to: determining, for each of the at least one near word, whether the near word contains a third semantic meaning other than the second semantic meaning; if the similar meaning word contains the third semantic, obtaining a word vector corresponding to the third semantic; determining the word vector of the second semantic contained in the similar meaning word based on the word vector of the similar meaning word and the word vector corresponding to the third semantic.

In one possible implementation, the determining unit 1402 is further configured to: and if the near-meaning word does not contain the at least one third semantic, determining that the word vector of the near-meaning word is the word vector of the second semantic.

In a possible implementation, the building unit 1403 is specifically configured to: acquiring a weight value corresponding to each similar meaning word; and constructing a word vector of the new word based on the weight value corresponding to each similar meaning word, the number of the similar meaning words corresponding to the new word and a word vector of a second semantic contained in each similar meaning word.

It should be noted that, for the functions of each functional unit in the word vector construction apparatus 1400 described in the embodiment of the present application, reference may be made to the related description of steps 601 to 604 in the method embodiment in fig. 6, which is not described herein again.

Referring to fig. 15, fig. 15 is a schematic structural diagram illustrating a semantic recognition device according to an embodiment of the present application. The semantic recognition device shown in fig. 15 can be used to implement part or all of the functions of the end side in the method embodiment described in fig. 10. The semantic recognition apparatus shown in fig. 15 may include a receiving unit 1501, a storage unit 1502, a slicing unit 1503, and a determining unit 1504. Wherein:

the receiving unit 1501 is configured to receive a word vector of a new word sent by a server, where the new word has a first semantic meaning, the new word is a vocabulary of an unknown word or a new meaning of an old word, the unknown word is a vocabulary of the server in which the word vector does not exist, the vocabulary of the new meaning of the old word is a vocabulary of the server in which the word vector exists, and a semantic meaning indicated by the word vector is different from the first semantic meaning;

a storage unit 1502, configured to store the word vector of the new word in a second word vector library;

the segmentation unit 1503 is configured to segment the corpus to be processed to obtain at least one segmented word;

a determining unit 1504, configured to determine, based on the first word vector or the second word vector, a language probability corresponding to the corpus to be processed if a first target vocabulary has the first word vector in a first word vector library and the first target vocabulary has the second word vector in a second word vector library, where the first target vocabulary is any one of the at least one participle;

the determining unit 1504 is further configured to determine, based on the language probability, the first word vector or the second word vector as a word vector corresponding to the first target vocabulary in the corpus to be processed.

In one possible implementation, the determining unit 1504 is further configured to: and if the first target vocabulary does not have word vectors in the first word vector library and the first target vocabulary has word vectors in the second word vector library, determining that the word vectors of the first target vocabulary in the second word vector library are the corresponding word vectors of the first target vocabulary in the corpus to be processed.

In a possible implementation, the determining unit 1504 is specifically configured to: calling a language model based on the word vectors of the first word vector and a second target vocabulary in the first word vector library, and determining the language probability corresponding to the corpus to be processed, wherein the second target vocabulary is a vocabulary of the at least one participle except the first target vocabulary; if the language probability is smaller than a probability threshold, determining the second word vector as a corresponding word vector of a first target vocabulary in the corpus to be processed; and if the language probability is greater than or equal to the probability threshold, determining the first word vector as a word vector corresponding to a first target vocabulary in the corpus to be processed.

In a possible implementation, the determining unit 1504 is specifically configured to: calling a language model based on the second word vector and a word vector of a second target vocabulary in the first word vector library, and determining the language probability corresponding to the corpus to be processed, wherein the second target vocabulary is a vocabulary of the at least one participle except the first target vocabulary; if the language probability is greater than or equal to a probability threshold, determining the second word vector as a corresponding word vector of a first target vocabulary in the corpus to be processed; and if the language probability is smaller than the probability threshold, determining the first word vector as a corresponding word vector of a first target vocabulary in the corpus to be processed.

It should be noted that, for the functions of each functional unit in the semantic recognition device 1500 described in this embodiment, reference may be made to the related description of steps 1001 to 1005 in the method embodiment in fig. 10, which is not described herein again.

The present application further provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, may implement part or all of the steps of any one of the method embodiments described above, and implement the functions of any one of the functional modules described in fig. 6 or fig. 10 above.

Embodiments of the present application also provide a computer program product, which when run on a computer or a processor, causes the computer or the processor to perform one or more steps of any one of the methods of fig. 6 or fig. 10. The respective constituent modules of the above-mentioned apparatus may be stored in the above-mentioned computer-readable storage medium if they are implemented in the form of software functional units and sold or used as independent products.

In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.

It should be understood that the reference herein to first, second, third, fourth, and various numerical designations is merely a convenient division to describe and is not intended to limit the scope of the present application.

It should be understood that the term "and/or" herein is merely one type of association relationship that describes an associated object, meaning that three relationships may exist, e.g., a and/or B may mean: a exists alone, A and B exist simultaneously, and B exists alone. In addition, the character "/" herein generally indicates that the former and latter related objects are in an "or" relationship.

It should also be understood that, in the various embodiments of the present application, the sequence numbers of the above-mentioned processes do not mean the execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.

It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.

In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the above-described division of units is only one type of division of logical functions, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit.

The above functions, if implemented in the form of software functional units and sold or used as a separate product, may be stored in a computer-readable storage medium. Based on such understanding, the technical solution of the present application or portions thereof that substantially contribute to the prior art may be embodied in the form of a software product stored in a storage medium and including instructions for causing a terminal device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a read-only memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.

It should be noted that, for simplicity of description, the above-mentioned method embodiments are described as a series of acts or combination of acts, but those skilled in the art will recognize that the present application is not limited by the order of acts described, as some steps may occur in other orders or concurrently depending on the application. Further, those skilled in the art should also appreciate that the embodiments described in the specification are preferred embodiments and that the acts and modules referred to are not necessarily required in this application.

The descriptions of the embodiments provided in the present application may be referred to each other, and the descriptions of the embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments. For convenience and simplicity of description, for example, the functions of each device, apparatus and steps executed in the embodiments of the present application may refer to the relevant description of the method embodiments of the present application, and may also refer to, be combined with or be cited among the method embodiments and the device embodiments.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A word vector construction method is applied to a server and comprises the following steps:

acquiring a new word, wherein the new word is a vocabulary of an unknown word or a new meaning of an old word, the new word has a first semantic meaning, the unknown word is a vocabulary without a word vector in the server, the vocabulary of the new meaning of the old word is a vocabulary with a word vector in the server, and the semantic meaning indicated by the word vector is different from the first semantic meaning;

determining a word vector of a second semantic contained in each similar meaning word based on a word vector of at least one similar meaning word corresponding to the new word, wherein the second semantic corresponds to the first semantic;

constructing a word vector of the new word based on the number of the similar meaning words corresponding to the new word and a word vector of a second semantic contained in each similar meaning word;

sending a word vector for the new word to an end-side device.

2. The method of claim 1, wherein the determining a word vector of a second semantic meaning included in each of the near-meaning words based on a word vector of at least one near-meaning word corresponding to the new word comprises:

determining, for each of the at least one near word, whether the near word contains a third semantic meaning other than the second semantic meaning;

if the similar meaning word contains the third semantic, obtaining a word vector corresponding to the third semantic;

determining a word vector of the second semantic contained in the similar meaning word based on the word vector of the similar meaning word and the word vector corresponding to the third semantic.

3. The method of claim 2, further comprising:

and if the near-meaning word does not contain the at least one third semantic, determining that the word vector of the near-meaning word is the word vector of the second semantic.

4. The method according to any one of claims 1-3, wherein the constructing the word vector of the new word based on the number of near-meaning words corresponding to the new word and the word vector of the second semantic meaning included in each near-meaning word comprises:

acquiring a weight value corresponding to each similar meaning word;

and constructing a word vector of the new word based on the weight value corresponding to each similar meaning word, the number of the similar meaning words corresponding to the new word and a word vector of a second semantic contained in each similar meaning word.

5. A semantic recognition method applied to a peer-to-peer device, the method comprising:

receiving a word vector of a new word sent by a server, wherein the new word has a first semantic meaning, the new word is a vocabulary of an unknown word or a vocabulary of an old word new sense, the unknown word is a vocabulary of which the word vector does not exist in the server, the vocabulary of the old word new sense is a vocabulary of which the word vector exists in the server, and the semantic meaning indicated by the word vector is different from the first semantic meaning;

storing the word vectors of the new words in a second word vector library;

segmenting the corpus to be processed to obtain at least one participle;

if a first target vocabulary has a first word vector in a first word vector library and the first target vocabulary has a second word vector in a second word vector library, determining the language probability corresponding to the corpus to be processed based on the first word vector or the second word vector, wherein the first target vocabulary is any vocabulary in the at least one participle;

and determining the first word vector or the second word vector as a corresponding word vector of a first target vocabulary in the corpus to be processed based on the language probability.

6. The method of claim 5, further comprising:

and if the first target vocabulary does not have word vectors in the first word vector library and the first target vocabulary has word vectors in the second word vector library, determining that the word vectors of the first target vocabulary in the second word vector library are the corresponding word vectors of the first target vocabulary in the linguistic data to be processed.

7. The method according to claim 5 or 6, wherein the determining the language probability corresponding to the corpus to be processed based on the first word vector or the second word vector comprises:

calling a language model based on the word vectors of the first word vector and a second target vocabulary in the first word vector library, and determining the language probability corresponding to the corpus to be processed, wherein the second target vocabulary is a vocabulary of the at least one participle except the first target vocabulary;

the determining, based on the language probability, the first word vector or the second word vector as a word vector corresponding to a first target vocabulary in the corpus to be processed includes:

if the language probability is smaller than a probability threshold, determining the second word vector as a corresponding word vector of a first target vocabulary in the corpus to be processed;

and if the language probability is greater than or equal to the probability threshold, determining the first word vector as a word vector corresponding to a first target vocabulary in the corpus to be processed.

8. The method according to claim 5 or 6, wherein the determining the language probability corresponding to the corpus to be processed based on the first word vector or the second word vector comprises:

calling a language model based on the second word vector and a word vector of a second target vocabulary in the first word vector library, and determining the language probability corresponding to the corpus to be processed, wherein the second target vocabulary is a vocabulary of the at least one participle except the first target vocabulary;

if the language probability is greater than or equal to a probability threshold, determining the second word vector as a corresponding word vector of a first target vocabulary in the corpus to be processed;

and if the language probability is smaller than the probability threshold, determining the first word vector as a corresponding word vector of a first target vocabulary in the corpus to be processed.

9. A word vector construction apparatus, wherein the apparatus is configured to a server, the apparatus comprising:

the device comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring a new word, the new word is a vocabulary of an unknown word or a new meaning of an old word, the new word has a first semantic meaning, the unknown word is a vocabulary of which a word vector does not exist in the server, the vocabulary of the new meaning of the old word is a vocabulary of which the word vector exists in the server, and the semantic meaning indicated by the word vector is different from the first semantic meaning;

a determining unit, configured to determine, based on a word vector of at least one near-meaning word corresponding to the new word, a word vector of a second semantic included in each near-meaning word, where the second semantic corresponds to the first semantic;

the construction unit is used for constructing a word vector of the new word based on the number of the similar meaning words corresponding to the new word and a word vector of a second semantic contained in each similar meaning word;

a sending unit, configured to send the word vector of the new word to an end-side device.

10. The apparatus according to claim 9, wherein the determining unit is specifically configured to:

11. The apparatus of claim 10, wherein the determining unit is further configured to:

12. The apparatus according to any one of claims 9-11, wherein the construction unit is specifically configured to:

acquiring a weight value corresponding to each similar meaning word;

13. A semantic recognition apparatus, wherein the apparatus is configured to a peer device, and wherein the method comprises:

the receiving unit is used for receiving a word vector of a new word sent by the server, wherein the new word has a first semantic meaning, the new word is a vocabulary of an unknown word or a new meaning of an old word, the unknown word is a vocabulary of which the word vector does not exist in the server, the vocabulary of the new meaning of the old word is a vocabulary of which the word vector exists in the server, and the semantic meaning indicated by the word vector is different from the first semantic meaning;

the storage unit is used for storing the word vectors of the new words in a second word vector library;

the segmentation unit is used for segmenting the linguistic data to be processed to obtain at least one participle;

a determining unit, configured to determine, based on a first word vector or a second word vector if a first target vocabulary has the first word vector in a first word vector library and the first target vocabulary has the second word vector in a second word vector library, a language probability corresponding to the corpus to be processed, where the first target vocabulary is any one of the at least one participle;

the determining unit is further configured to determine, based on the language probability, that the first word vector or the second word vector is used as a word vector corresponding to a first target vocabulary in the corpus to be processed.

14. The apparatus of claim 13, wherein the determining unit is further configured to:

and if the first target vocabulary does not have word vectors in the first word vector library and the first target vocabulary has word vectors in the second word vector library, determining the word vectors of the first target vocabulary in the second word vector library as the corresponding word vectors of the first target vocabulary in the corpus to be processed.

15. The apparatus according to claim 13 or 14, wherein the determining unit is specifically configured to:

16. The apparatus according to claim 13 or 14, wherein the determining unit is specifically configured to:

17. A server, comprising a processor and a memory, the processor and the memory being interconnected, wherein the memory is configured to store a computer program comprising program instructions, the processor being configured to invoke the program instructions to perform the method of any one of claims 1-4.

18. An end-side device comprising a processor and a memory, the processor and the memory being interconnected, wherein the memory is configured to store a computer program comprising program instructions, the processor being configured to invoke the program instructions to perform the method of any one of claims 5-8.

19. A computer-readable storage medium, in which a computer program or instructions are stored, which, when executed by a processor, implement the word vector construction method according to any one of claims 1 to 4 or implement the semantic recognition method according to any one of claims 5 to 8.