Information processing method and device, equipment and medium

Abstract

The embodiment of the invention provides an information processing method and device, equipment and a medium. According to the method, firstly, a plurality of language units are determined based on first input information for a machine learning model, the plurality of language units comprise a first group of language units, the first group of language units comprise one or more language units, and the importance of each language unit in the one or more language units is higher than or equal to an importance threshold value. Next, a correlation metric is determined between each language unit in at least one language unit in the plurality of language units and each language unit in the first set of language units. Second input information for the machine learning model is then determined based on the at least one correlation metric, the second input information comprising at least a portion of the plurality of language units. By means of the method, the information processing efficiency and accuracy can be effectively improved.

Description

Technical Field

[0001] Exemplary embodiments of the present disclosure generally relate to computer technology, and more particularly, to information processing methods, devices, apparatuses, and computer-readable storage media. Background Art

[0002] Nowadays, with the development of economic globalization and the Internet, people need to use more and more information on the Internet in their lives. However, when searching for content on the Internet, it is usually difficult for users to directly obtain the required information due to the massive amount of data. Therefore, how to identify the information or objects required by users from the massive data has become a concern. Summary of the invention

[0003] In a first aspect of the present disclosure, an information processing method is provided. In the method, a plurality of language units are determined based on first input information for a machine learning model, the plurality of language units include a first group of language units, the first group of language units include one or more language units, and the importance of each of the one or more language units is greater than or equal to an importance threshold; a correlation measure between each of at least one of the plurality of language units and each of the first group of language units is determined; and second input information for the machine learning model is determined based on at least one correlation measure, the second input information includes at least a portion of the plurality of language units.

[0004] In a second aspect of the present disclosure, an information processing device is provided. The device includes: a language unit determination module, configured to determine a plurality of language units based on first input information for a machine learning model, the plurality of language units including a first group of language units, the first group of language units including one or more language units, and the importance of each of the one or more language units is greater than or equal to an importance threshold; a correlation measure determination module, configured to determine a correlation measure between each of at least one of the plurality of language units and each of the first group of language units; and an input information determination module, configured to determine second input information for the machine learning model based on at least one correlation measure, the second input information including at least a portion of the plurality of language units.

[0005] In a third aspect of the present disclosure, an electronic device is provided. The device includes at least one processing unit; and at least one memory, the at least one memory is coupled to the at least one processing unit and stores instructions for execution by the at least one processing unit. When the instructions are executed by the at least one processing unit, the device executes the method of the first aspect.

[0006] In a fourth aspect of the present disclosure, a computer-readable storage medium is provided, wherein a computer program is stored on the computer-readable storage medium, and the computer program can be executed by a processor to implement the method of the first aspect.

[0007] It should be understood that the contents described in this content section are not intended to limit the key features or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent in the following detailed description in conjunction with the accompanying drawings. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, wherein:

[0009] Figure 1 A schematic diagram showing a process performed by an information processing device according to some embodiments of the present disclosure;

[0010] Figure 2 A schematic diagram showing a process of determining a correlation measure according to some embodiments of the present disclosure;

[0011] Figure 3A A schematic diagram of an attention matrix according to some embodiments of the present disclosure is shown;

[0012] Figure 3B A schematic diagram showing an example process of an information processing method according to some embodiments of the present disclosure

[0013] Figure 4 A flowchart of an information processing method according to some embodiments of the present disclosure is shown;

[0014] Figure 5 A schematic structural block diagram of an information processing device according to some embodiments of the present disclosure is shown; and

[0015] Figure 6 A block diagram of a device capable of implementing various embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0016] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0017] In the description of the embodiments of the present disclosure, the term "including" and similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The term "some embodiments" should be understood as "at least some embodiments". The following may also include other explicit and implicit definitions. As used herein, the term "model" can represent the association relationship between various data. For example, the above-mentioned association relationship can be obtained based on a variety of technical solutions currently known and / or to be developed in the future.

[0018] It is understandable that the data involved in this technical solution (including but not limited to the data itself, the acquisition or use of the data) shall comply with the requirements of relevant laws, regulations and relevant provisions.

[0019] It is understandable that before using the technical solutions disclosed in the embodiments of the present disclosure, the types, scope of use, usage scenarios, etc. of the personal information involved in the present disclosure should be informed to the user and the user's authorization should be obtained in an appropriate manner in accordance with relevant laws and regulations.

[0020] For example, in response to receiving the user's input, a prompt message is sent to the user to clearly prompt the user that the operation performed in response to the user's input will require obtaining and using the user's personal information. Thus, the user can independently choose whether to provide personal information to the electronic device, application, server, storage medium or other software or hardware that performs the operation of the technical solution of the present disclosure according to the prompt message.

[0021] As an optional but non-limiting embodiment, in response to receiving the user's input, the prompt information is sent to the user in a manner such as a pop-up window, and the pop-up window may present the prompt information in text form. In addition, the pop-up window may also carry a selection control for the user to select "agree" or "disagree" to provide personal information to the electronic device.

[0022] It is understandable that the above notification and the process of obtaining user authorization are merely illustrative and do not constitute a limitation to the embodiments of the present disclosure. Other methods that meet relevant laws and regulations may also be applied to the embodiments of the present disclosure.

[0023] The term "in response to" as used herein refers to a state in which a corresponding event occurs or a condition is satisfied. It will be understood that the timing of executing a subsequent action executed in response to the event or condition is not necessarily strongly related to the time when the event occurs or the condition is satisfied. For example, in some cases, the subsequent action may be executed immediately when the event occurs or the condition is satisfied; while in other cases, the subsequent action may be executed some time after the event occurs or the condition is satisfied.

[0024] The term "pruning" used here means improving the efficiency and performance of a machine learning model by removing certain unnecessary parts during the training or reasoning process of the model. The goal of pruning is to reduce the complexity of the model and reduce the computational overhead while minimizing the accuracy of the model. Pruning the input information of a machine learning model involves removing or ignoring some unnecessary language units when processing the input text, thereby reducing the amount of computation required by the model.

[0025] The term "Softmax function" used here refers to an activation function in multi-classification problems, which is used to convert an arbitrary real number vector into a probability distribution. After being processed by the Softmax function, each element in the vector is in the range [0,1], and the sum of all outputs is 1, so these outputs can be regarded as probability values, indicating the predicted probability of each category. By converting the output of the neural network into the probability of each category, the Softmax function enables the machine learning model to make probabilistic predictions.

[0026] The term "token" used here refers to the smallest unit in text processing, which can be a word, subword or character. A token is the basic element for converting raw text into a format that can be processed by a computer. When processing text, a token is a unit that is segmented for further analysis, modeling and generation.

[0027] Intelligent counting plays an important role in the field of information processing. Intelligent counting methods usually involve intelligently retrieving, extracting and analyzing specific digital or numerical information from a large amount of data using machine learning related technologies. Based on intelligent counting methods, users can efficiently find, process and interpret digital data in an intelligent way to support decision analysis, predictive modeling and automated processes.

[0028] In the current field of machine learning model reasoning and application, the requirements for latency, storage occupancy, and energy consumption are relatively high. For intelligent number-checking applications built using machine learning models, due to the complex situations they face, high accuracy requirements are required. Specifically, in the application scenarios of intelligent number-checking, machine learning models face some problems.

[0029] For example, since intelligent number-checking applications require high accuracy of results, users need to input a large amount of input information (e.g., prompt information, guiding language, or trigger words) when interacting with machine learning models. Too much input information will lead to information redundancy and confusion, thereby reducing information processing efficiency. In addition, the input information occupies more language units, which will lead to excessive latency at the application level and excessive storage usage at the reasoning framework level.

[0030] Additionally, in the intelligent number-checking scenario, if the machine learning model cannot solve the user's question in one answer, the user will usually make multiple rounds of queries. During multiple rounds of interaction, the number of language units that the machine learning model needs to process will increase dramatically, which will lead to increasing latency and storage usage.

[0031] In addition, intelligent counting requires high accuracy of the model. As the number of language units increases, the attention of the machine learning model may decrease, causing the accuracy to decrease over time.

[0032] Due to the diversity and complexity of user questions, it is usually difficult to apply systematic pruning to user questions at the application level. Usually, in intelligent number-checking applications, manual simplification methods are used to reduce the input information language units, thereby reducing the amount of model calculations. For example, when the content of the input information is too much, the specific content of the input information is manually analyzed to remove repeated, irrelevant or unnecessary parts, so that the simplified input information is easy to understand and process. In the process of manual simplification, the important information and user intentions in the input information need to be fully retained to ensure the integrity and accuracy of the input information.

[0033] Methods based on manual simplification may cause the loss of some key information due to the subjectivity of manual judgment. For some complex input information, manual simplification may not be able to fully judge the potential meaning. In addition, a large amount of input information will make manual simplification consume too much time and energy, reducing the efficiency of information processing.

[0034] In addition, rewriting input information using machine learning models is also used in intelligent counting scenarios. Machine learning models understand the structure and semantics of language by learning a large amount of text data, so that they can restate the input information. This method may generate overly general or vague statements, resulting in the loss of specific context and details in the original input information. Additionally, the input information rewriting results of the machine learning model may be biased or limited due to the influence of training data. In addition, for some highly professional or specific input information, the machine learning model may not understand the meaning, resulting in poor rewriting results.

[0035] In intelligent number-checking applications, there are still many challenges in simplifying and rewriting input information, and more effective methods require further research.

[0036] In order to at least partially solve the above problems and other potential problems, the present disclosure proposes a scheme for information processing. The scheme of the present disclosure determines multiple language units based on input information for a machine learning model, determines a correlation measure between each of the multiple language units and a language unit whose importance is higher than an importance threshold, and determines another input information for the machine learning model based on the correlation measure. In this way, the input information for the machine learning model is simplified, thereby improving the efficiency of information processing and the accuracy of processing results.

[0037] For the convenience of explanation, the following Figure 1 The information processing process of is discussed as an example. Specifically, Figure 1 A schematic diagram of an information processing process 100 performed by an information processing device 101 according to some embodiments of the present disclosure is shown. The information processing device 101 may be any type of mobile terminal, fixed terminal or portable terminal, including any device that can use Internet services, such as a mobile phone, a desktop computer, a laptop computer, a notebook computer, a netbook computer, a tablet computer, a media computer, a multimedia tablet, a PDA, etc. In addition, the information processing device 101 may also be implemented as an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, and content distribution networks.

[0038] like Figure 1 As shown, the information processing device 101 may first obtain first input information 110 based on a machine learning model. The machine learning model may be a machine learning model based on natural language or other machine learning models that can implement the functions of the embodiments of the present disclosure. In some embodiments, the first input information 110 may include questions described in natural language, such as questions to inquire about the sales of a certain product.

[0039] In an embodiment of the present disclosure, the information processing device 101 first determines language units 120-1, 120-2, ..., 120-N (hereinafter collectively referred to as language units 120) based on the first input information 110 for the machine learning model. In some embodiments, as shown in the figure, the information processing device 101 can determine multiple language units. For example, the first input information 110 can be "Please help me check the GMV and sales volume of the past seven days", and the information processing device 101 can determine multiple language units 120 as "please", "help me", "query", "past", "seven days", "of", "sales volume", "is" and "how much". It should be understood that the above is only an example for determining a language unit, which is exemplary and not restrictive, and the scope of the present disclosure is not limited in this respect.

[0040] Among the plurality of language units 120, some language units 120 have importance higher than the importance threshold, and these language units 120 are included in the first group of language units. In other words, the plurality of language units 120 include the first group of language units, and the importance of each language unit in the first group of language units is higher than or equal to the importance threshold.

[0041] In some embodiments, the information processing device 101 may determine the importance of the language unit 120 based on entity recognition. For example, the information processing device 101 may determine the entity in the first input information 110 based on named entity recognition (NER). Next, the information processing device 101 may determine entities with more important categories from these entities based on predetermined rules and strategies, that is, the importance is higher than the importance threshold. In some embodiments, these entities may include one or more language units 120. It should be understood that the above is merely an example for determining the importance of a language unit, which is exemplary rather than restrictive, and the scope of the present disclosure is not limited in this respect.

[0042] For example, for the multiple language units mentioned above, “please”, “help me”, “query”, “past”, “seven days”, “of”, “GMV”, “and”, “sales”, “is” and “how much”, the information processing device 101 can determine that “past seven days” is an important screening condition based on entity recognition, so “past” and “seven days” can be determined as having an importance higher than the importance threshold, that is, “past” and “seven days” can be language units in the first group of language units. Additionally, the information processing device 101 can determine that “GMV” and “sales” are important indicators (or dimensions, dimension values) based on entity recognition, so “GMV” and “sales” can be language units in the first group of language units. It should be understood that the above is merely an example for determining the importance of a language unit, which is exemplary rather than restrictive, and the scope of the present disclosure is not limited to this.

[0043] Additionally, the plurality of language units 120 may further include a second group of language units. The importance of each language unit in the second group of language units is lower than the importance threshold. For example, the information processing device 101 may determine the important language units (the first group of language units) in the plurality of language units 120 based on the entity recognition method, and the other language units in the plurality of language units 120 may be determined as the second group of language units.

[0044] Next, for some or all of the language units in the plurality of language units 120, the information processing device 101 may determine a correlation measure between each of the language units and each of the language units in the first group of language units. In some embodiments, the information processing device 101 may determine a correlation measure between each of the language units in the second group of language units and each of the language units in the first group of language units.

[0045] Specifically, the information processing device 101 may determine feature representations of the plurality of language units 120. Then, the information processing device 101 may determine a correlation measure based on the feature representations of some or all of the language units in the plurality of language units 120 and the feature representations of each language unit in the first group of language units. For example, the information processing device 101 may convert the plurality of language units 120 into a plurality of vectors by word embedding. Then, the information processing device 101 may obtain a dot product by performing a dot product operation on the vector corresponding to the language unit 120 and the vector corresponding to another language unit, and the dot product is a correlation measure between the two language units.

[0046] In some embodiments, the correlation measure is included in the attention matrix, and each row of the attention matrix includes the correlation measure between each language unit in at least one of the multiple language units and each language unit in the first group of language units. For example, the information processing device 101 can perform a dot product operation on the vectors corresponding to some or all of the language units in the multiple language units 120 and the vectors corresponding to each language unit in the first group of language units to obtain multiple correlation measures. Next, the information processing device 101 can determine the attention matrix based on the multiple correlation measures, and the elements of the attention matrix are the values ​​of the correlation measures.

[0047] Next, the information processing device 101 determines second input information 130 for the machine learning model based on at least one correlation metric, where the second input information 130 includes at least a portion of the plurality of language units 120. In some embodiments, the information processing device 101 may determine the second input information 130 and input the second input information 130 into the machine learning model as simplified input information, thereby reducing the amount of data processing and improving the efficiency of information processing.

[0048] In some implementations, the information processing device 101 may determine the weights corresponding to the plurality of language units 120 based on the correlation measures of the plurality of language units 120. Then, the information processing device 101 may determine one or more target language units from the plurality of language units 120 based on the weights. Next, the information processing device 101 may determine the second input information 130 based on the one or more target language units and at least one of the first group of language units. Specifically, the information processing device 101 may determine the weight by summing a portion of the correlation measures of each language unit in the at least one language unit 120. In some embodiments, at least a portion of the plurality of language units includes one or more target language units, and the weight of each of the one or more target language units is greater than a first threshold.

[0049] For example, among the multiple language units "please", "help me", "query", "past", "seven days", "of", "GMV", "and", "sales", "is" and "how much" mentioned above, the information processing device 101 can sum a part of the correlation measures of "please" with other language units (for example, the correlation measures of "please" relative to the language units whose importance is higher than the importance threshold in other language units) to determine the weight corresponding to "please", and similarly determine the weights of the various language units 120. In some embodiments, the information processing device 101 can determine one or more language units with larger weights among the multiple language units 120 as target language units based on a first threshold. For example, when the weight of a language unit 120 is greater than or equal to the first threshold, the language unit 120 is determined as a target language unit. It should be understood that the above is merely an example for determining a target language unit, which is exemplary rather than restrictive, and the scope of the present disclosure is not limited in this respect.

[0050] Next, the information processing device 101 may determine the second input information 130 based on the target language unit. For example, the information processing device 101 may combine one or more target language units with the language units in the first group of language units to determine the second input information 130. In some embodiments, the information processing device 101 may use a natural language-based machine learning model to determine the second input information 130 based on the one or more target language units and the language units in the first group of language units.

[0051] In some implementations, the first threshold is determined based on a first number of language units in the first group of language units and a second number of language units in at least one of the plurality of language units. For example, the information processing device 101 may determine the first threshold based on the first number of language units in the first group of language units and the second number of language units in the second group of language units.

[0052] In some embodiments, the information processing device 101 may repeatedly perform the above process multiple times, for example, repeatedly processing the determined second input information 130 as the first input information 110 until a predetermined number of processing times is reached or a related indicator meets a predetermined threshold.

[0053] In this way, the information processing device 101 can determine the second input information 130 based on the first input information 110 for the machine learning model, effectively reducing the amount of information to be processed by the machine learning model and retaining language units with higher importance or higher relevance, thereby effectively improving the efficiency of the information processing process and the accuracy of the information processing results.

[0054] Combine the following Figure 2 , Figure 3A and Figure 3B The process executed by the information processing device 101 according to the embodiment of the present application is described.

[0055] Figure 2 A schematic diagram of a process 200 for determining a correlation measure according to some embodiments of the present disclosure is shown. In some embodiments, the process 200 may be performed by the information processing device 101. It should be understood that although the process 200 is described below using the information processing device 101 as an example, this is merely exemplary and is not intended to limit the embodiments of the present disclosure in any way.

[0056] exist Figure 2 In the embodiment, the first input information 110 is still discussed as "Please help me check the GMV and sales volume in the past seven days?" In box 210, the information processing device 101 can determine the vectors corresponding to the language units "past", "seven days", "GMV" and "how much" of the first input information 110. It should be understood that the above is only an example of determining a part of the language units of the first input information 110, which is exemplary and not restrictive, and the scope of the present disclosure is not limited to this. Another part of the language units in the first input information 110 are in Figure 2 Not shown, process 200 may include these language units that are not shown.

[0057] For example, vector 220 corresponding to "past", vector 222 corresponding to "seven days", vector 224 corresponding to "GMV", and vector 226 corresponding to "how much". These vectors can be represented as a 1×d matrix, where d is the dimension of the vector, which is determined during the word embedding process. In some embodiments, the information processing device 101 can determine the vector corresponding to the language unit based on methods such as Word2Vec or Glove.

[0058] Next, the information processing device 101 can determine the correlation measure between the two language units. The above determination process of the correlation measure can be performed through an attention function. For example, the information processing device 101 can perform a dot product operation on two of the above vectors 220, 222, 224 and 226 to obtain a correlation measure. Specifically, the information processing device 101 can perform a dot product operation on vector 220 and vector 222 to obtain a dot product as the correlation measure between the language units "past" and "seven days". In some embodiments, the correlation measure can also be obtained by other means, such as scaled dot product attention or additive attention.

[0059] As shown in the figure, the information processing device 101 can determine the correlation metric 230 corresponding to the language unit "past" by the dot product of vector 220 with vectors 220, 222, vector 224 and vector 226. Similarly, the information processing device 101 can determine the correlation metric 232 corresponding to the language unit "seven days", the correlation metric 234 corresponding to the language unit "GMV" and the correlation metric 236 corresponding to the language unit "how much". These correlation metrics can be represented as an n×1 matrix, where n is equal to the number of language units. Additionally, the sum of the elements in the matrix representing the correlation metric can be equal to 1. The elements in these matrices are the values ​​of the correlation metrics of the language unit corresponding to the matrix relative to a certain language unit.

[0060] Next, in block 240, the information processing device 101 may normalize the correlation metrics 230, 232, 234, and 236. For example, the information processing device 101 may use a Softmax function to normalize the correlation metrics 230, 232, 234, and 236, respectively, to obtain normalized correlation metrics 250, 252, 254, and 256, respectively. In some embodiments, the normalization process may be performed in a neural network.

[0061] Then, the information processing device 101 may determine the attention matrix based on the normalized correlation metrics 250, 252, 254, 256. For example, the information processing device 101 may use the n×1 matrix corresponding to the normalized correlation metrics 250, 252, 254, 256 as the row of the attention matrix. It should be understood that the above is only an example for determining the correlation metric, which is exemplary rather than restrictive, and the scope of the present disclosure is not limited in this respect.

[0062] Furthermore, in some exemplary implementations, the information processing device 101 may utilize the above-described process of determining attention weights to determine correlation metrics and weights between important entities and unimportant entities.

[0063] In this way, the information processing device 101 can determine the correlation measure between the language units 120, and remove some language units based on the correlation measure and determine the second input information 130, thereby reducing the amount of information to be processed by the machine learning model and improving information processing efficiency and accuracy.

[0064] In some embodiments, the correlation measure between language units obtained by process 200 can be included in an attention matrix. Figure 3A FIG. 3 shows a schematic diagram of an attention matrix 300A according to some embodiments of the present disclosure. Figure 3AAs shown, the elements in the attention matrix 300A can be represented as the items in the table in the figure. Specifically, each element in the table corresponds to a value of a correlation measure. The correlation measure is the value of the correlation measure between the language unit represented by the row label of the row where the element is located relative to the language unit represented by the column label of the column where the element is located. For example, the element "0.02" in the dotted box 301 in the figure is the correlation measure between the language unit "please" corresponding to the row label of the row and the language unit "help me" corresponding to the column label of the column.

[0065] exist Figure 3A In the embodiment of the present invention, the values ​​of these correlation measures are normalized, so the sum of all elements in the same row of the table is 1. It should be understood that in some embodiments, the correlation measures do not need to be normalized, which is also feasible.

[0066] Additionally, in Figure 3A In the embodiment of the figure, the language units "past", "seven days", "GMV" and "sales" corresponding to the column labels of the columns where the elements in the dashed boxes 302, 304, 306 and 308 are located are determined by the information processing device 101 as language units of important entities, that is, language units in the first group of language units. In some embodiments, the correlation measure of other language units (that is, language units in the second group of language units) relative to the language units in the first group of language units is used to determine the weight of the other language units.

[0067] For example, for the language unit "how much" corresponding to the row label of the row where the elements in the dotted boxes 302, 304, 306, and 308 are located, the sum of the correlation measure of the language unit relative to "past" in the first group of language units (i.e., the element "0.09" in box 302), the correlation measure of the language unit relative to "seven days" in the first group of language units (i.e., the element "0.09" in box 303), the correlation measure of the language unit relative to "GMV" in the first group of language units (i.e., the element "0.11" in box 304), and the correlation measure of the language unit relative to "sales" in the first group of language units (i.e., the element "0.13" in box 305) is determined as the attention weight of the language unit "how much". In some embodiments, the weight can also be determined as the average or weighted sum of the correlation measures of the language units in the second group of language units relative to the language units in the first group of language units. It should be understood that the above is only an example for determining the weight, which is exemplary and not restrictive, and the scope of the present disclosure is not limited in this respect.

[0068] In this way, the information processing device 101 can determine an attention matrix based on the relevance measure of the language units, and thereby determine the weights of each language unit based on the attention matrix for determining the language units used to determine the second input information 130, thereby reducing the amount of information to be processed by the machine learning model and improving information processing efficiency and accuracy.

[0069] Reference below Figure 3B A schematic diagram describing an example process 300B of an information processing method according to some embodiments of the present disclosure. It should be understood that the process 300B may be executed by the information processing device 101 or by other appropriate devices, and the embodiments of the present disclosure are not limited thereto.

[0070] Box 310 is the first input information 110 for the machine learning model, for example, "Please help me check the GMV and sales in the past seven days?" In box 312, the information processing device 101 can perform entity decomposition on the first input information 110 based on NER entity recognition to obtain the decomposed first input information 110, as shown in box 314. "Past seven days", "GMV" and "sales" in the first input information 110 are determined by the information processing device 101 as important entities, that is, language units in the first language unit. In addition, the information processing device 101 can determine multiple language units 120 based on the first input information 110.

[0071] Next, in block 320, the information processing device 101 may perform an attention matrix calculation process based on the plurality of language units 120. The attention matrix calculation process may include Figure 2 and Figure 3A The process in the embodiment of the present invention.

[0072] Then, in block 330, the information processing device 101 may perform a weighted average of the relevance measures of the various language units according to the elements corresponding to the important entities in the attention matrix. For example, for the language unit "please", the information processing device 101 may perform a weighted average of the relevance measures of "past seven days", "GMV", and "sales" relative to "please" in the attention matrix to obtain the weight of "please".

[0073] In block 340, the information processing device 101 may determine a language unit whose weight is less than the first threshold based on the weight obtained in block 330, and remove the language unit, and not use the language unit whose weight is less than the first threshold when determining the second input information 130. Figure 3B In the embodiment of the present invention, the theoretical average score of all language units can be calculated by the following formula:

[0074]

[0075] Wherein n is the number of language units of important entities (ie, the first number of language units in the first group of language units), and m is the number of language units of unimportant entities (ie, the second number).

[0076] In some implementations, if the weight of a language unit is greater than or equal to the theoretical average score determined based on formula (1), the language unit can be determined as a language unit with a strong correlation with other language units. The information processing device 101 can determine the target language unit based on the theoretical average score. Specifically, the information processing device 101 can determine a first threshold based on the theoretical average score, and remove language units with a weight less than the first threshold.

[0077] The first threshold can be calculated by the following formula:

[0078]

[0079] Wherein h is the first threshold, α is a predetermined coefficient (α>1), n ​​is the number of language units of important entities (i.e., the first number of language units in the first group of language units), and m is the number of language units of unimportant entities (i.e., the second number). In some embodiments, α may be predefined or precalculated by the information processing device 101.

[0080] For language units with weights less than the first threshold, the information processing device 101 may remove them. In other words, the information processing device 101 may not use language units with weights less than the first threshold when determining the second input information 130. Additionally, the information processing device 101 may retain language units with weights greater than or equal to the first threshold. The information processing device 101 may use language units with weights greater than or equal to the first threshold when determining the second input information 130. In some embodiments, language units of important entities may be retained. In other words, language units in the first group of language units are retained for determining the second input information 130.

[0081] The information processing device 101 may obtain the retained language units 120 , which include “query,” “past seven days,” “GMV,” and “sales,” as shown in block 342 .

[0082] In block 350, the information processing device 101 may determine whether the stopping criterion is met. In some embodiments, if the weights of all language units in block 342 are greater than or equal to the first threshold, the information processing device 101 may determine that the stopping criterion is met, otherwise the information processing device 101 may determine that the stopping criterion is not met.

[0083] If the information processing device 101 determines that the stopping criterion is not met, the information processing device 101 returns to box 320 and performs an attention matrix calculation process based on multiple language units 120 after removing language units whose weights are less than the first threshold, until the weights of all language units are greater than the first threshold, that is, the stopping criterion is met.

[0084] If the information processing device 101 determines that the stop criterion is reached, the information processing device 101 may determine the second input information 130 based on the language unit in the box 342 in box 360. In some embodiments, the information processing device 101 may determine the second input information 130 based on the language unit in the box 342 by using a natural language-based machine learning model, and the second input information 130 is output as natural language.

[0085] In addition, the second input information 130 can be input into a machine learning model for intelligent counting, or other related processing.

[0086] In this way, the information processing device 101 can remove language units with lower relevance indicators and importance in the first input information 110, reduce the amount of information that the machine learning model needs to process, and thus improve the efficiency and accuracy of information processing.

[0087] Figure 4 FIG. 4 is a flowchart of a method 400 for information processing according to some embodiments of the present disclosure. It should be understood that the method 400 may be, for example, Figure 1 The information processing device 101 or other appropriate device is used for execution.

[0088] At 410, the information processing device 101 determines a plurality of language units based on first input information for a machine learning model, the plurality of language units including a first group of language units, the first group of language units including one or more language units, the importance of each of the one or more language units being greater than or equal to an importance threshold. At 720, the information processing device 101 determines a correlation measure between each of at least one of the plurality of language units and each of the language units in the first group of language units. At 730, the information processing device 101 determines second input information for the machine learning model based on at least one correlation measure, the second input information including at least a portion of the plurality of language units.

[0089] In some exemplary embodiments, the information processing device 101 may determine the correlation measure by, for example, determining feature representations of a plurality of language units, and determining the correlation measure based on the feature representation of each language unit in at least one language unit and the feature representations of each language unit in the first group of language units.

[0090] In some exemplary embodiments, the information processing device 101 may determine the second input information by, for example, determining a weight corresponding to each language unit in at least one language unit based on a correlation measure of each language unit in at least one language unit, determining one or more target language units from at least one language unit based on the weight, and determining the second input information based on the one or more target language units and at least one item in the first group of language units.

[0091] In some exemplary embodiments, the information processing device 101 may determine the weight in the following manner, for example, by summing the correlation measure of each language unit in the at least one language unit to determine the weight.

[0092] In some exemplary embodiments, at least one of the plurality of language units may include a second group of language units, and the importance of each language unit in the second group of language units is lower than an importance threshold.

[0093] In some exemplary embodiments, at least a portion of the plurality of language units may include one or more target language units, and a weight corresponding to each of the one or more target language units is greater than a first threshold.

[0094] In some exemplary embodiments, the first threshold may be determined based on a first number of language units in the first group of language units and a second number of language units in at least one language unit of the plurality of language units.

[0095] In some exemplary embodiments, the correlation measure may be included in an attention matrix, and each row of the attention matrix may include a correlation measure between each language unit of at least one language unit of the plurality of language units and each language unit of the first group of language units.

[0096] Figure 5 1 is a block diagram of an apparatus 500 for information processing according to some embodiments of the present disclosure. The apparatus 500 may be implemented as or included in Figure 1 The information processing device 101 or other appropriate devices. Each module / component in the apparatus 500 may be implemented by hardware, software, firmware or any combination thereof.

[0097] As shown in the figure, the device 500 includes a language unit determination module 510, which is configured to determine multiple language units based on first input information for a machine learning model, the multiple language units include a first group of language units, the first group of language units includes one or more language units, and the importance of each of the one or more language units is higher than or equal to an importance threshold; a correlation measure determination module 520, which is configured to determine a correlation measure between each language unit of at least one language unit in the multiple language units and each language unit in the first group of language units; and an input information determination module 530, which is configured to determine second input information for the machine learning model based on at least one correlation measure, and the second input information includes at least a portion of the multiple language units.

[0098] In some exemplary embodiments, the correlation metric determination module 520 may be further configured to determine feature representations of a plurality of language units; and determine a correlation metric based on the feature representation of each language unit in at least one language unit and the feature representations of each language unit in the first group of language units.

[0099] In some exemplary embodiments, the input information determination module 530 can be further configured to determine a weight corresponding to each language unit in the at least one language unit based on a relevance measure of each language unit in the at least one language unit; determine one or more target language units from the at least one language unit based on the weight; and determine the second input information based on the one or more target language units and at least one item in the first group of language units.

[0100] In some exemplary embodiments, the input information determination module 530 may be further configured to determine the weight by summing the relevance measure of each language unit in the at least one language unit.

[0101] In some exemplary embodiments, at least one of the plurality of language units may include a second group of language units, and the importance of each language unit in the second group of language units is lower than an importance threshold.

[0102] In some exemplary embodiments, at least a portion of the plurality of language units may include one or more target language units, and a weight corresponding to each of the one or more target language units is greater than a first threshold.

[0103] In some exemplary embodiments, the first threshold may be determined based on a first number of language units in the first group of language units and a second number of language units in at least one language unit of the plurality of language units.

[0104] In some exemplary embodiments, the correlation measure may be included in an attention matrix, and each row of the attention matrix may include a correlation measure between each language unit of at least one language unit of the plurality of language units and each language unit of the first group of language units.

[0105] Figure 6 FIG. 6 is a block diagram of an electronic device 600 capable of implementing an embodiment of the present disclosure. It should be understood that Figure 6 The electronic device 600 shown is merely exemplary and should not constitute any limitation on the functionality and scope of the embodiments described herein. Figure 6 The electronic device 600 shown may be used to implement the method described above. Figure 6 The electronic device 600 shown can be used to implement Figure 1 The information processing device 101 or Figure 5 information processing device 500.

[0106] like Figure 6 As shown, the electronic device 600 is in the form of a general-purpose computing device. The components of the electronic device 600 may include, but are not limited to, one or more processors or processing units 610, a memory 620, a storage device 630, one or more communication units 640, one or more input devices 650, and one or more output devices 660. The processing unit 610 may be an actual or virtual processor and is capable of performing various processes according to a program stored in the memory 620. In a multi-processor system, multiple processing units execute computer executable instructions in parallel to improve the parallel processing capabilities of the electronic device 600.

[0107] The electronic device 600 typically includes a plurality of computer storage media. Such media may be any available media accessible to the electronic device 600, including but not limited to volatile and non-volatile media, removable and non-removable media. The memory 620 may be a volatile memory (e.g., a register, a cache, a random access memory (RAM)), a non-volatile memory (e.g., a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), flash memory), or some combination thereof. The storage device 630 may be a removable or non-removable medium, and may include a machine-readable medium, such as a flash drive, a disk, or any other medium, which may be capable of being used to store information and / or data (e.g., training data for training) and may be accessed within the electronic device 600.

[0108] The electronic device 600 may further include additional removable / non-removable, volatile / non-volatile storage media. Figure 6As shown in , a disk drive for reading or writing from a removable, non-volatile disk (e.g., a "floppy disk") and an optical drive for reading or writing from a removable, non-volatile optical disk may be provided. In these cases, each drive may be connected to the bus (not shown) by one or more data media interfaces. The memory 620 may include a computer program product 625 having one or more program modules that are configured to perform various methods or actions of various embodiments of the present disclosure.

[0109] The communication unit 640 implements communication with other computing devices through a communication medium. Additionally, the functions of the components of the electronic device 600 can be implemented in a single computing cluster or multiple computing machines that can communicate through a communication connection. Therefore, the electronic device 600 can operate in a networked environment using a logical connection with one or more other servers, a network personal computer (PC), or another network node.

[0110] The input device 650 may be one or more input devices, such as a mouse, a keyboard, a tracking ball, etc. The output device 660 may be one or more output devices, such as a display, a speaker, a printer, etc. The electronic device 600 may also communicate with one or more external devices (not shown) through the communication unit 640 as needed, such as a storage device, a display device, etc., communicate with one or more devices that allow a user to interact with the electronic device 600, or communicate with any device that allows the electronic device 600 to communicate with one or more other computing devices (e.g., a network card, a modem, etc.). Such communication may be performed via an input / output (I / O) interface (not shown).

[0111] According to an exemplary embodiment of the present disclosure, a computer-readable storage medium is provided, on which computer-executable instructions are stored, wherein the computer-executable instructions are executed by a processor to implement the method described above. According to an exemplary embodiment of the present disclosure, a computer program product is also provided, the computer program product is tangibly stored on a non-transitory computer-readable medium and includes computer-executable instructions, and the computer-executable instructions are executed by a processor to implement the method described above. According to an exemplary embodiment of the present disclosure, a computer program product is provided, on which a computer program is stored, and when the program is executed by a processor, the method described above is implemented.

[0112] Various aspects of the present disclosure are described herein with reference to the flowcharts and / or block diagrams of the methods, devices, equipment, and computer program products implemented according to the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer-readable program instructions.

[0113] These computer-readable program instructions can be provided to a processing unit of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processing unit of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0114] Computer-readable program instructions can be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operational steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, so that the instructions executed on the computer, other programmable data processing apparatus, or other device implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0115] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple implementations of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of a module, program segment or instruction includes one or more executable instructions for realizing the logical function of the specification. In some implementations as replacements, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be realized by a special hardware-based system that performs the function or action of the specification, or can be realized by a combination of special hardware and computer instructions.

[0116] The above descriptions of various implementations of the present disclosure are exemplary, non-exhaustive, and not limited to the disclosed implementations. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described implementations. The selection of terms used herein is intended to best explain the principles of the implementations, practical applications, or improvements to the technology in the marketplace, or to enable other persons of ordinary skill in the art to understand the various embodiments disclosed herein.

Claims

1. An information processing method, comprising: Determine a plurality of language units based on first input information to the machine learning model, the plurality of language units comprising a first group of language units, the first group of language units comprising one or more language units, the importance of each of the one or more language units being greater than or equal to an importance threshold; determining a correlation measure between each language unit of at least one language unit of the plurality of language units and each language unit of the first set of language units; as well as Based on the at least one relevance metric, second input information for the machine learning model is determined, the second input information including at least a portion of the plurality of language units.

2. The method of claim 1 , wherein determining the correlation metric comprises: determining feature representations of the plurality of language units; as well as The correlation measure is determined based on the feature representation of each language unit in the at least one language unit and the feature representation of each language unit in the first group of language units.

3. The method according to claim 1 or 2, wherein determining the second input information comprises: determining a weight corresponding to each of the at least one language unit based on the relevance metric of each of the at least one language unit; determining one or more target language units from the at least one language unit based on the weight; as well as The second input information is determined based on at least one of the one or more target language units and the first group of language units.

4. The method of claim 3, wherein determining the weight comprises: The weight is determined by summing a portion of the relevance measure for each of the at least one language unit. 5 . The method of claim 1 , wherein the at least one language unit of the plurality of language units comprises a second group of language units, and the importance of each language unit in the second group of language units is below an importance threshold.

6. The method according to claim 3, wherein the at least part of the plurality of language units includes one or more target language units, and a weight corresponding to each of the one or more target language units is greater than a first threshold. 7 . The method of claim 6 , wherein the first threshold is determined based on a first number of language units in the first set of language units and a second number of language units in at least one of the plurality of language units.

8. A method according to claim 1, wherein the relevance measure is included in an attention matrix, each row of the attention matrix includes a relevance measure between each language unit of at least one language unit of the multiple language units and each language unit of the first group of language units.

9. An information processing device, comprising: A language unit determination module is configured to determine a plurality of language units based on first input information for a machine learning model, wherein the plurality of language units include a first group of language units, the first group of language units includes one or more language units, and importance of each of the one or more language units is greater than or equal to an importance threshold; a correlation measure determination module configured to determine a correlation measure between each language unit of at least one language unit of the plurality of language units and each language unit of the first group of language units; as well as An input information determination module is configured to determine second input information for the machine learning model based on the at least one correlation metric, wherein the second input information includes at least a portion of the multiple language units.

10. An electronic device, comprising: at least one processing unit; as well as At least one memory, the at least one memory being coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit, the instructions causing the device to perform the method according to any one of claims 1 to 8 when executed by the at least one processing unit.

11. A computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the method according to any one of claims 1 to 8 when executed by a processor.

Images