Slot information processing method and electronic device

Abstract

The application relates to the technical field of data processing, and discloses a slot information processing method and electronic equipment. The method comprises the following steps: after receiving a text to be processed, determining feature information corresponding to the text to be processed; wherein the feature information comprises slot information and a parallel conjunction; based on the slot information and the parallel conjunction, determining a plurality of slot fields; wherein any slot field contains slot information with the same attribute; allocating the plurality of slot fields to obtain a plurality of target field combinations corresponding to the text to be processed; wherein any target field combination contains a slot field of each attribute. According to the slot information and the parallel conjunction, the plurality of slot fields of the text to be processed are determined, the plurality of slot fields are allocated, the plurality of target field combinations corresponding to the text to be processed are obtained, the plurality of target field combinations reflect the real intention of a user, and intelligent question answering can be accurately carried out based on the plurality of target fields.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a slot information processing method and an electronic device. Background Technology

[0002] With the development of technology, intelligent question answering technology has been widely applied in many fields. Intelligent question answering technology requires matching user questions with preset questions to retrieve corresponding answer information.

[0003] In related technologies, keywords (slot information) are extracted from text (user questions), and then matched with preset questions based on the keywords.

[0004] However, due to the diversity of sentence combinations in text and the differences in user expression, texts with different meanings may extract the same slot information. Therefore, the slot information extracted in the above method is difficult to accurately reflect the user's true intention, affecting the accuracy of intelligent question answering. Summary of the Invention

[0005] This application provides a slot information processing method and an electronic device for determining information that reflects the user's true intent.

[0006] In a first aspect, embodiments of this application provide a slot information processing method, the method comprising:

[0007] Upon receiving the text to be processed, the feature information corresponding to the text to be processed is determined; wherein, the feature information includes slot information and parallel conjunctions;

[0008] Based on the slot information and the parallel conjunctions, multiple slot fields are determined; wherein any slot field contains slot information with the same attributes.

[0009] The multiple slot fields are assigned to obtain multiple target field combinations corresponding to the text to be processed; wherein any target field combination contains a slot field of each preset attribute.

[0010] The above scheme, upon receiving the text to be processed, not only determines the slot information of the text to be processed, but also identifies the coordinating conjunctions (which describe the relationship between the slot information); then, based on the slot information and coordinating conjunctions, it determines multiple slot fields of the text to be processed. By allocating multiple slot fields, it obtains multiple target field combinations corresponding to the text to be processed. These multiple target field combinations reflect the user's true intent, and based on multiple target fields, it can accurately perform intelligent question answering.

[0011] In some optional implementations, multiple slot fields are determined based on the slot information and the coordinating conjunctions, including:

[0012] Based on the attributes of each slot information and the parallel conjunctions, all slot information is divided into slot information combinations corresponding to the attributes;

[0013] For any combination of slot information, based on the preset hierarchical relationship corresponding to the attributes of the combination of slot information, the slot information is combined to obtain at least one slot field.

[0014] In some optional implementations, based on the attributes of each slot information and the parallel conjunctions, all slot information is divided into slot information combinations corresponding to the attributes, including:

[0015] Slot information with the same attributes is grouped into the same initial slot information combination;

[0016] For any initial slot information combination, if the parallel conjunction is associated with slot information in the initial slot information combination, then it is determined whether the slot information associated with the parallel conjunction is at a different level; wherein, the slot information associated with the parallel conjunction is the slot information adjacent to the parallel conjunction.

[0017] If so, the slot information before the parallel conjunction and the slot information after the parallel conjunction in the initial slot information combination are divided into different sub-slot information combinations; otherwise, the initial slot information combination is not divided; wherein, the slot information is sorted according to its position in the text to be processed.

[0018] In some optional implementations, the plurality of slot fields are assigned to obtain a combination of multiple target fields corresponding to the text to be processed, including:

[0019] The slot fields are sorted according to the position of the target slot information in the text to be processed, and adjacent slot fields with the same attributes are bound together; wherein, the target slot information is the slot information corresponding to the lowest level among all slot fields;

[0020] For any slot field, if there is another slot field with the same attribute as the slot field in the previous initial field combination, then the slot field is assigned to the next initial field combination; otherwise, the slot field is assigned to the previous initial field combination.

[0021] For any initial field combination, the target slot field of the missing attribute is added to the initial field combination, and the slot field bound in the initial field combination is combined with the slot fields of other attributes in the initial field combination to obtain the target field combination.

[0022] In some alternative implementations, the target slot field for the missing attribute is determined by the following method:

[0023] For any initial field combination, if the first field combination contains a slot field corresponding to the missing attribute, then the last slot field corresponding to the missing attribute in the first field combination is determined as the target slot field.

[0024] If there is no slot field corresponding to the missing attribute in the first field combination, then the first slot field corresponding to the missing attribute in the second field combination is determined as the target slot field.

[0025] Wherein, the first field combination is another initial field combination preceding the initial field combination; the second field combination is another initial field combination following the initial field combination.

[0026] In some optional implementations, determining the feature information corresponding to the text to be processed includes:

[0027] Based on the word vectors and part-of-speech vectors of the text to be processed, the fusion feature vector of the text to be processed is determined;

[0028] Based on the fused feature vector, the feature information corresponding to the text to be processed is identified.

[0029] In some optional implementations, the feature information further includes intent information; after assigning the plurality of slot fields to obtain a plurality of target field combinations corresponding to the text to be processed, the method further includes:

[0030] For any combination of target fields, the combination of target fields is matched with a preset question corresponding to the intent information, and a target question that is successfully matched is selected from the preset question.

[0031] The response information corresponding to the target question will be notified through a preset notification method.

[0032] Secondly, embodiments of this application provide an electronic device, including a communication unit and a processor;

[0033] The communication unit is used to receive the text to be processed;

[0034] The processor is configured to, upon receiving a text to be processed, determine the feature information corresponding to the text to be processed; wherein the feature information includes slot information and coordinating conjunctions; based on the slot information and the coordinating conjunctions, determine multiple slot fields; wherein any slot field contains slot information with the same attributes; allocate the multiple slot fields to obtain multiple target field combinations corresponding to the text to be processed; wherein any target field combination contains a slot field for each preset attribute.

[0035] Thirdly, embodiments of this application provide a slot information processing device, comprising:

[0036] The feature extraction module is used to determine the feature information corresponding to the text to be processed after receiving the text to be processed; wherein, the feature information includes slot information and parallel conjunctions;

[0037] The field determination module is used to determine multiple slot fields based on the slot information and the parallel conjunctions; wherein any slot field contains slot information with the same attributes;

[0038] The field allocation module is used to allocate the multiple slot fields to obtain multiple target field combinations corresponding to the text to be processed; wherein, any target field combination contains one slot field of each preset attribute.

[0039] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the slot information processing method as described in any of the first aspects.

[0040] Furthermore, the technical effects of any of the implementation methods in the second to fourth aspects can be found in the technical effects of different implementation methods in the first aspect, and will not be repeated here. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a schematic diagram illustrating the first application scenario provided in the embodiments of this application;

[0043] Figure 2 A schematic flowchart illustrating the first slot information processing method provided in this application embodiment;

[0044] Figure 3 A schematic diagram illustrating the feature information determination method provided in the embodiments of this application;

[0045] Figure 4 This is a schematic diagram of vector fusion provided in an embodiment of this application;

[0046] Figure 5 This is a schematic diagram of the self-attention calculation process provided in an embodiment of this application;

[0047] Figure 6 This is a schematic diagram illustrating the feature information determination process provided in the embodiments of this application;

[0048] Figure 7 A schematic flowchart illustrating the second slot information processing method provided in this application embodiment;

[0049] Figure 8 A schematic flowchart illustrating the slot field determination method provided in this application embodiment;

[0050] Figure 9 A schematic flowchart illustrating the slot information partitioning method provided in this application embodiment;

[0051] Figure 10 This is a schematic diagram of the initial slot information combination provided in the embodiments of this application;

[0052] Figure 11 This is a schematic diagram of the sub-slot information combination provided in an embodiment of this application;

[0053] Figure 12 A schematic flowchart illustrating the third slot information processing method provided in this application embodiment;

[0054] Figure 13 A schematic flowchart illustrating the target field combination determination method provided in this application embodiment;

[0055] Figure 14 A schematic flowchart illustrating the method for determining the target slot field provided in an embodiment of this application;

[0056] Figure 15 A schematic flowchart illustrating the fourth slot information processing method provided in this application embodiment;

[0057] Figure 16 A schematic flowchart illustrating the fifth slot information processing method provided in this application embodiment;

[0058] Figure 17 This is a schematic diagram of the structure of the first slot information processing device provided in the embodiments of this application;

[0059] Figure 18 This is a schematic diagram of the structure of the second type of slot information processing device provided in the embodiments of this application;

[0060] Figure 19 A schematic block diagram of an electronic device provided in an embodiment of this application;

[0061] Figure 20 A schematic diagram of the program product provided in the embodiments of this application. Detailed Implementation

[0062] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0063] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0064] In the description of this application, unless otherwise expressly specified and limited, the term "connection" should be interpreted broadly. For example, it can refer to a direct connection, an indirect connection through an intermediate medium, or a connection within two devices. Those skilled in the art can understand the specific meaning of the above term in this application based on the specific circumstances.

[0065] In intelligent question answering technology, it is necessary to match user questions with preset questions in order to retrieve the corresponding answer information.

[0066] In related technologies, keywords (slot information) are extracted from text (user questions), and then matched with preset questions based on the keywords.

[0067] However, due to the diversity of sentence structures in texts and the differences in user expression, texts with different meanings may extract the same slot information.

[0068] For example, Text 1 is "What were the GDP and national happiness index of Qingdao and Jinan in March 2021?" and Text 2 is "What were the GDP and national happiness index of Qingdao and Jinan in March 2021?" These two sentences have different meanings, but the slot information extracted from both Text 1 and Text 2 is the same: {"city":["Qingdao","Jinan"],"index":["GDP","National Happiness Index"],"month":["3"],"year":["2021"]}. Although the meanings of Text 1 and Text 2 are different, the slot information is the same, the pre-defined question is consistent, and the corresponding response information is the same; at least one response is incorrect.

[0069] Therefore, the slot information extracted in the above method is difficult to accurately reflect the user's true intention, affecting the accuracy of intelligent question answering.

[0070] Based on this, embodiments of this application provide a slot information processing method and an electronic device. The method includes: after receiving a text to be processed, determining feature information corresponding to the text to be processed; wherein the feature information includes slot information and parallel conjunctions; determining multiple slot fields based on the slot information and the parallel conjunctions; wherein any slot field contains slot information with the same attributes; allocating the multiple slot fields to obtain multiple target field combinations corresponding to the text to be processed; wherein any target field combination contains a slot field with preset attributes.

[0071] See Figure 1 The illustration shows an application scenario provided by an embodiment of this application. This application scenario includes electronic devices and terminals;

[0072] In some embodiments, the electronic device communicates with the terminal via data. The electronic device may connect to the terminal via a local area network (LAN), wireless local area network (WLAN), or other networks, and the terminal may send user-triggered pending text (user questions) to the electronic device.

[0073] After receiving the text to be processed, the electronic device determines the feature information corresponding to the text to be processed; wherein, the feature information includes slot information and parallel conjunctions; based on the slot information and the parallel conjunctions, multiple slot fields are determined; wherein, any slot field contains slot information with the same attributes; the multiple slot fields are allocated to obtain multiple target field combinations corresponding to the text to be processed; wherein, any target field combination contains a slot field with preset attributes.

[0074] The terminal in this application embodiment can be a user terminal device such as a mobile phone, laptop, or tablet.

[0075] The electronic devices in the embodiments of this application may include one or more types of servers.

[0076] The above scheme, upon receiving the text to be processed, not only determines the slot information of the text to be processed, but also identifies the coordinating conjunctions (which describe the relationship between the slot information); then, based on the slot information and coordinating conjunctions, it determines multiple slot fields of the text to be processed. By allocating multiple slot fields, it obtains multiple target field combinations corresponding to the text to be processed. These multiple target field combinations reflect the user's true intent, and based on multiple target fields, it can accurately perform intelligent question answering.

[0077] The technical solution of this application and how it solves the above-mentioned technical problems will be described in detail below with reference to the accompanying drawings and specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0078] This application provides a first method for processing slot information, which can be applied to the aforementioned electronic devices, such as... Figure 2 As shown, the method may include:

[0079] Step S201: After receiving the text to be processed, determine the feature information corresponding to the text to be processed.

[0080] The feature information includes slot information and parallel conjunctions.

[0081] In practice, for certain complex sentence structures, the slot information does not accurately reflect the user's true intent, resulting in slot information allocation issues. For example, in the question "What were the GDP figures for Qingdao in the first quarter of 2020 and the national happiness index for Jinan in the third quarter of 2021?", the following can be extracted: {"city":["Qingdao", Jinan"], "index":["GDP", "national happiness index"], "season":["1", "3"], "year":["2020", "2021"]}. However, the combinations of "city", "index", and "season" and "year" are not addressed. Even considering only the time-related combinations of "season" and "year", there are instances where the extracted slot information is identical to that of "the first and third quarters of 2020 and the first and third quarters of 2021" and "the first and third quarters of 2020 and the third quarter of 2021", making it impossible to separate these slots.

[0082] Since the coordinating conjunctions in the text describe the relationship between slot information, upon receiving the text to be processed, not only are the slot information of the text to be processed determined, but also the coordinating conjunctions of the text to be processed are determined.

[0083] Step S202: Based on the slot information and the parallel conjunctions, determine multiple slot fields.

[0084] Each slot field contains slot information with the same attributes.

[0085] For example, slot information and coordinating conjunctions are independent information. Slot information with the same attributes may have hierarchical relationships. Based on this, this embodiment needs to determine multiple slot fields of the text to be processed according to the slot information and coordinating conjunctions.

[0086] Step S203: Assign the multiple slot fields to obtain a combination of multiple target fields corresponding to the text to be processed.

[0087] Each target field combination contains a slot field for each of the preset attributes.

[0088] In implementation, each attribute has at least one slot field. These slot fields can be combined into multiple target field combinations. Each target field combination contains one slot field of each attribute. All target field combinations fully reflect the user's true intent.

[0089] In this embodiment, some slots are designed with the same category and corresponding attributes. For example, year, month, and day all belong to time, so the attribute is "time"; province, city, and district all belong to location, so the attribute is "place", and so on.

[0090] The above scheme, upon receiving the text to be processed, not only determines the slot information of the text to be processed, but also identifies the coordinating conjunctions (which describe the relationship between the slot information); then, based on the slot information and coordinating conjunctions, it determines multiple slot fields of the text to be processed. By allocating multiple slot fields, it obtains multiple target field combinations corresponding to the text to be processed. These multiple target field combinations reflect the user's true intent, and based on multiple target fields, it can accurately perform intelligent question answering.

[0091] See Figure 3 As shown, in some optional implementations, the above-mentioned determination of feature information can be achieved in, but is not limited to, the following ways:

[0092] Step S301: Based on the word vectors and part-of-speech vectors of the text to be processed, determine the fusion feature vector of the text to be processed.

[0093] In this embodiment, it is necessary to identify coordinating conjunctions. Word vectors are vector representations of a single field, and it is difficult to accurately identify coordinating conjunctions using only word vectors. The most significant difference between coordinating conjunctions and other fields lies in their part-of-speech tag. Incorporating part-of-speech tag information into the extracted features will significantly improve the extraction of coordinating conjunctions.

[0094] See Figure 4 As shown, by using text of length n, w = {w1, w2, ..., w...} n After processing w using a bidirectional encoder representation from Transformers (BERT), which utilizes an attention mechanism to improve model training speed, we obtain BERT word vectors h = {h1, h2, ..., h} that contain token embeddings, sentence embeddings, and positional embeddings. n Assuming each character or word is represented by an m-dimensional vector, the h-matrix E h ∈R m*V V represents the number of fields;

[0095] After processing w using word2vec (a vector model), the initial part-of-speech vector p = {p1, p2, ..., p} is obtained. n Conjunctions in text generally exist between two slots. By using a self-attention layer, the correlation between the part-of-speech vector of each position and the part-of-speech vector of other positions is extracted, thereby obtaining the degree of dependence of the part-of-speech vector of each position on the part-of-speech vector of other positions.

[0096] See Figure 5 As shown, taking the initial part-of-speech vector p1 as an example, after normalization by softmax (an activation function), the attention score is obtained, which is the degree of attention that p1 gives to all vectors. The attention score is used to obtain the features of different degrees of influence on the part-of-speech vector at the current position. In this way, each part-of-speech vector participates in the calculation of the part-of-speech vector, which pays attention to global information, but the degree of participation of each part-of-speech vector is different, which focuses on key part-of-speech information.

[0097] The word vector h and the part-of-speech vector p′ are fused to obtain the fused feature vector.

[0098] Taking the text to be processed as "What were the GDPs of Qingdao and Jinan in 2021?" as an example, the entire vector processing process can be found in [reference needed]. Figure 6 As shown, the information extracted is as shown in Table 1:

[0099] Table 1

[0100]

[0101] In this context, [CLS] is the start symbol, and [SEP] is the clause separator; in the slot labels, year is the year, city is the city, side is the coordinating conjunction, index is the index, and o is other conjunctions.

[0102] Step S302: Based on the fused feature vector, identify the feature information corresponding to the text to be processed.

[0103] The above scheme incorporates part-of-speech vectors into word vectors, enriches text features by introducing part-of-speech information, and uses part-of-speech rules to introduce auxiliary slot information for recognition, while improving the recognition rate of parallel conjunctions.

[0104] Correspondingly, embodiments of this application provide a second method for processing slot information, such as... Figure 7 As shown, the method may include:

[0105] Step S701: After receiving the text to be processed, determine the fusion feature vector of the text to be processed based on the word vector and part-of-speech vector of the text to be processed.

[0106] Step S702: Based on the fused feature vector, identify the feature information corresponding to the text to be processed.

[0107] Step S703: Based on the slot information and the parallel conjunctions, determine multiple slot fields.

[0108] Step S704: Assign the multiple slot fields to obtain a combination of multiple target fields corresponding to the text to be processed.

[0109] The specific implementation of steps S701 to S704 can be found in the above embodiments, and will not be repeated here.

[0110] See Figure 8 As shown, in some optional implementations, the above slot field can be determined in, but is not limited to, the following ways:

[0111] Step S801: Based on the attributes of each slot information and the parallel conjunctions, divide all slot information into the slot information combination corresponding to the attributes.

[0112] As mentioned above, slot information with the same attributes may have hierarchical relationships. Based on this, this embodiment needs to first classify the slot information into the slot information combination corresponding to the attribute according to the attributes and parallel conjunctions of the slot information.

[0113] Step S802: For any combination of slot information, based on the preset hierarchical relationship corresponding to the attributes of the combination of slot information, combine the slot information to obtain at least one slot field.

[0114] For example, the slot information in each slot information combination can be combined with each other to obtain different slot fields. If a slot information combination is missing the target level (such as the highest level) in the preset hierarchical relationship, and the slot information of the target level needs to be filled, the slot information of the target level in other slot information combinations or the default information of the target level can be added to the slot information combination.

[0115] The above scheme, since slot information with the same attribute may have hierarchical relationships, divides the slot information into slot information combinations corresponding to the attribute based on the attributes and parallel conjunctions of the slot information; then, based on the preset hierarchical relationship corresponding to the attributes of the slot information combination, the slot information is combined to obtain the slot field corresponding to the preset hierarchical relationship of each attribute.

[0116] See Figure 9 As shown, in some optional implementations, the slot information can be divided in, but is not limited to, the following ways:

[0117] Step S901: Group slot information with the same attributes into the same initial slot information combination.

[0118] For example, slot information with different attributes is located in different slot information combinations. Therefore, slot information with the same attribute can be first assigned to the same initial slot information combination.

[0119] Step S902: For any initial slot information combination, if the parallel conjunction is associated with slot information in the initial slot information combination, then determine whether the slot information associated with the parallel conjunction is at a different level; wherein, the slot information associated with the parallel conjunction is the slot information adjacent to the parallel conjunction.

[0120] In this embodiment, since the coordinating conjunction describes the relationship between slot information, the coordinating conjunction modifies the associated preceding and following slot information. If the associated preceding and following slot information is at the same level, then the two slot information are in a parallel same-level relationship and can be combined with each other, that is, they can be classified in the same slot information combination. If the associated preceding and following slot information is at different levels, then the two slot information are in a parallel non-same-level relationship and cannot be combined with each other, that is, they cannot be classified in the same slot information combination.

[0121] Step S903: If yes, then divide the slot information before the parallel conjunction and the slot information after the parallel conjunction in the initial slot information combination into different sub-slot information combinations; otherwise, do not divide the initial slot information combination; wherein, the slot information is sorted according to its position in the text to be processed.

[0122] For example, if the associated slot information is at the same level, then the two slot information are parallel and of the same level, and can be combined with each other, that is, classified in the same slot information combination, without dividing the initial slot information combination, and retaining the initial slot information combination of this attribute.

[0123] If the associated preceding and following slot information are at different levels, then these two slot information are parallel non-same level relationships and cannot be combined with each other, that is, they cannot be classified in the same slot information combination. Therefore, the slot information before the parallel conjunction and the slot information after the parallel conjunction in the initial slot information combination are divided into different sub-slot information combinations.

[0124] See Figure 10 As shown, the slot information is sorted according to its position in the text to be processed to obtain the initial slot information combination 1, which includes slot information 1, slot information 2, slot information 3 and slot information 4; the parallel conjunctions associate slot information 2 and slot information 3. Slot information 2 and slot information 3 are at different levels, so the initial slot information combination 1 needs to be further divided.

[0125] See Figure 11 As shown, slot information 1 and slot information 2 are assigned to sub-slot information combination 1, and slot information 3 and slot information 4 are assigned to sub-slot information combination 2.

[0126] The above Figure 10 as well as Figure 11 This is merely an illustrative example; this embodiment does not specifically limit the combination of initial slot information and sub-slot information.

[0127] The above scheme ensures the uniformity of attributes among slot information in a given combination by first grouping slot information with the same attribute into the same initial slot information combination. Since coordinating conjunctions describe the relationship between slot information, if the slot information before and after the coordinating conjunction is at the same level, then these two slot information are in a parallel sibling relationship. The initial slot information combination with this attribute is retained, allowing slot information with the same attribute before and after the coordinating conjunction to be combined with each other. If the slot information before and after the coordinating conjunction is at different levels, then these two slot information are in a parallel non-sibling relationship. The slot information before and after the coordinating conjunction in the initial slot information combination is divided into different sub-slot information combinations to avoid the combination of these two parts of slot information.

[0128] The following two specific examples illustrate this:

[0129] The text to be processed is "What were the GDP figures for Qingdao and Jinan in 2021 and 2022?"

[0130] The sorted slot keys (KeyList) are: ["year", "city", "side", "city", "year", "index"];

[0131] The corresponding slot values ​​(ValueList) are: ["2021", "Qingdao", "and", "Jinan", "2022", "GDP"];

[0132] The corresponding position information (PositionList) is: [0, 5, 8, 9, 12, 18].

[0133] 1. Group slot information with the same attribute into an initial slot information combination. The initial slot information combination corresponding to the time attribute includes the slot information {year:2021}, {year:2022}; the initial slot information combination corresponding to the place attribute includes the slot information {city:Qingdao}, {city:Jinan}; the initial slot information combination corresponding to the index attribute includes the slot information {index:GDP}.

[0134] 2. The coordinating conjunction "side" connects "Qingdao" and "Jinan". "Qingdao" and "Jinan" are at the same level (both are cities in the place attribute), so the initial slot information combination does not need to be further divided.

[0135] 3. Taking the time attribute as an example: The preset hierarchical relationship corresponding to the time attribute is year-month-day. The initial slot information combination corresponding to the time attribute only contains the slot information at the year level, and no further combination is needed. This results in slot field time1 (containing slot information {year:2021}) and slot field time2 (containing slot information {year:2022}).

[0136] II. The text to be processed is "What are the GDP and national happiness index of Qingdao and Jinan in March and April 2021 respectively?"

[0137] The sorted slot keys (KeyList) are: ["year", "month", "side", "month", "city", "city", "index", "index"];

[0138] The corresponding slot values ​​(ValueList) are: ["2021", "March", "and", "April", "Qingdao", "Jinan", "GDP", "National Happiness Index"];

[0139] The corresponding position information (PositionList) is: [0, 5, 7, 8, 10, 13, 17, 23].

[0140] 1. Group slot information with the same attribute into an initial slot information combination. The initial slot information combination corresponding to the time attribute includes the slot information {year:2021}, {month:March}, and {month:April}; the initial slot information combination corresponding to the place attribute includes the slot information {city:Qingdao} and {city:Jinan}; the initial slot information combination corresponding to the index attribute includes the slot information {index:Gross Domestic Product} and {index:National Happiness Index}.

[0141] 2. The coordinating conjunction "side" connects "March" and "April". "March" and "April" are at the same level (both are months in the time attribute), so the initial slot information combination does not need to be further divided.

[0142] 3. Taking the time attribute as an example, the default hierarchical relationship corresponding to the time attribute is year-month-day. It is necessary to combine "2021" with "March" and "April" respectively to obtain the slot field time1 (containing slot information {year:2021,month:March}) and the slot field time2 (containing slot information {year:2021,month:April}).

[0143] The above examples are merely illustrative of slot fields, and this application does not impose any specific limitations on them.

[0144] Correspondingly, this application provides a third method for processing slot information, such as... Figure 12 As shown, the method may include:

[0145] Step S1201: After receiving the text to be processed, determine the feature information corresponding to the text to be processed.

[0146] Step S1202: Based on the attributes of each slot information and the parallel conjunctions, divide all slot information into slot information combinations corresponding to the attributes.

[0147] Step S1203: For any combination of slot information, based on the preset hierarchical relationship corresponding to the attributes of the combination of slot information, combine the slot information to obtain at least one slot field.

[0148] Step S1204: Assign the multiple slot fields to obtain a combination of multiple target fields corresponding to the text to be processed.

[0149] The specific implementation of steps S1201 to S1204 can be referred to the above embodiments, and will not be repeated here.

[0150] See Figure 13As shown, in some optional implementations, the above-mentioned combination of target fields can be determined in, but is not limited to, the following ways:

[0151] Step S1301: Sort the slot fields according to the position of the target slot information in the text to be processed, and bind adjacent slot fields with the same attributes.

[0152] The target slot information refers to the slot information corresponding to the lowest level in each slot field.

[0153] Step S1302: For any slot field, if there is another slot field with the same attribute as the slot field in the previous initial field combination, then the slot field is assigned to the next initial field combination; otherwise, the slot field is assigned to the previous initial field combination.

[0154] Step S1303: For any initial field combination, add the target slot field of the missing attribute to the initial field combination, and combine the slot field bound in the initial field combination with the slot fields of other attributes in the initial field combination to obtain the target field combination.

[0155] In practice, in order to determine all target field combinations obtained from all slot field combinations, each target field combination contains one slot field for each attribute, and at least one attribute slot field is different in different target field combinations;

[0156] Based on this, this embodiment first sorts the slot fields according to the position of the slot information corresponding to the lowest level in the text, and binds adjacent slot fields with the same attribute (first assigning them to the same initial field combination).

[0157] Then determine whether there are other slot fields in the current initial field combination that have the same attribute as the next slot field. If there are, it means that the current initial field combination is already full of slot fields with the same attribute, and the slot field needs to be assigned to the next initial field combination; if not, it means that the current initial field combination is still missing a slot field with the same attribute, and the slot field needs to be assigned to the current initial field combination.

[0158] After all slot fields are assigned to the initial field combination, there may be initial field combinations that are missing slot fields (i.e., the initial field combination is missing slot fields of some attributes). By adding the target slot field of the missing attribute to the initial field combination, it is ensured that each field combination has a slot field of each attribute.

[0159] As described above, this embodiment binds adjacent slot fields with the same attribute. However, each target field combination only contains one slot field of each attribute. It is also necessary to unbind adjacent slot fields with the same attribute and combine the slot fields bound in the initial field combination with the slot fields of other attributes in the initial field combination.

[0160] See Figure 14 As shown, in some optional implementations, the target slot field can be determined in, but is not limited to, the following ways:

[0161] Step S1401: For any initial field combination, if the first field combination has a slot field corresponding to the missing attribute, then the last slot field corresponding to the missing attribute in the first field combination is determined as the target slot field.

[0162] The first field combination is another initial field combination that precedes the initial field combination.

[0163] Step S1402: If there is no slot field corresponding to the missing attribute in the first field combination, then the first slot field corresponding to the missing attribute in the second field combination is determined as the target slot field.

[0164] The second field combination is another initial field combination following the initial field combination.

[0165] In this embodiment, to avoid generating duplicate target field combinations, the slot fields need to be filled in sequence.

[0166] For example, first look for slot fields corresponding to missing attributes from other initial field combinations (first field combination) before the initial field combination;

[0167] If so, the last slot field corresponding to the missing attribute in the first field combination (the slot field that is closest to the initial field combination mentioned above) is determined as the target slot field;

[0168] If not, find the first slot field corresponding to the missing attribute from the other initial field combinations (second field combinations) after the initial field combination (the slot field closest to the above initial field combination) and use it as the target slot field.

[0169] By using the above method, not only are all target field combinations obtained from all slot field combinations determined, but the generation of duplicate target field combinations is also avoided.

[0170] The following two specific examples illustrate this:

[0171] The text to be processed is "What were the GDP figures for Qingdao and Jinan in 2021 and 2022?"

[0172] 1. After processing, 5 slot fields are obtained: time1 (containing slot information {year:2021}), place1 (containing slot information {city:Qingdao}), place2 (containing slot information {city:Jinan}), time2 (containing slot information {year:2022}), and index1 (containing slot information {index:GDP}).

[0173] 2. Place1 and Place2 are adjacent and have the same attributes. Bind these two slot fields to get [time1,place1,place2] (initial field combination 1) and [time2,index1] (initial field combination 2).

[0174] 3. The initial field combination 1 is missing a slot field with the index attribute. Add the above-mentioned index1 to the initial field combination 1.

[0175] The initial field combination 1 is missing a slot field with the place attribute. The bound place1 and place2 will be added to the initial field combination 2.

[0176] We get [time1,place1,place2,index1] (initial complete field combination 1) and [time2,index1,place1,place2] (initial complete field combination 2).

[0177] 4. Combine place1 and place2 in the initial completion field combination 1 with the slot fields of other attributes in the initial completion field combination 1 to obtain target field combination 1 [time1, place1, index1] and target field combination 2 [time1, place2, index1].

[0178] Combine place1 and place2 in the initial completion field combination 2 with the slot fields of other attributes in the initial completion field combination 2 to obtain target field combination 3 [time2,index1,place1] and target field combination 4 [time2,index1,place2].

[0179] Among them, target field combination 1 is: {year:2021,city:Qingdao,index:GDP};

[0180] The target field combination 2 is: {year:2021,city:Jinan,index:GDP};

[0181] The target field combination 3 is: {year:2022,city:Qingdao,index:GDP};

[0182] The target field combination 4 is: {year:2022,city:Jinan,index:GDP}.

[0183] II. The text to be processed is "What are the GDP and national happiness index of Qingdao and Jinan in March and April 2021 respectively?"

[0184] 1. After processing, 6 slot fields are obtained: time1 (containing slot information {year:2021, month:March}), time2 (containing slot information {year:2021, month:April}), place1 (containing slot information {city:Qingdao}), place2 (containing slot information {city:Jinan}), index1 (containing slot information {index:GDP}), and index2 (containing slot information {index:National Happiness Index}).

[0185] 2. Time1 and Time2 are adjacent and have the same attributes, so bind these two slot fields; Place1 and Place2 are adjacent and have the same attributes, so bind these two slot fields; Index1 and Index2 are adjacent and have the same attributes, so bind these two slot fields.

[0186] The initial field combination is obtained as [time1,time2,place1,place2,index1,index2].

[0187] 3. The initial field combination has no missing attribute slot fields; combine time1, time2 in the initial field combination with the slot fields of other attributes in the initial field combination, combine place1, place2 in the initial field combination with the slot fields of other attributes in the initial field combination, combine index1, index2 in the initial field combination with the slot fields of other attributes in the initial field combination, and combine index1, index2 in the initial field combination with the slot fields of other attributes in the initial field combination to obtain target field combination 1 [time1, place1, index1], target field combination 2 [time2, place1, index1], target field combination 3 [time1, place2, index1], target field combination 4 [time2, place2, index1], target field combination 5 [time1, place1, index2], target field combination 6 [time2, place1, index2], target field combination 7 [time1, place2, index2], and target field combination 8 [time2, place2, index2].

[0188] The target field combination 1 is: {year:2021,month:March,city:Qingdao,index:Gross Domestic Product};

[0189] The target field combination 2 is: {year:2021,month:April,city:Qingdao,index:Gross Domestic Product};

[0190] The target field combination 3 is: {year:2021,month:March,city:Jinan,index:Gross Domestic Product};

[0191] The target field combination 4 is: {year:2021,month:April,city:Jinan,index:Gross Domestic Product};

[0192] The target field combination 5 is: {year:2021,month:March,city:Qingdao,index:National Happiness Index};

[0193] The target field combination 6 is: {year:2021,month:April,city:Qingdao,index:National Happiness Index};

[0194] The target field combination 7 is: {year:2021,month:March,city:Jinan,index:National Happiness Index};

[0195] The target field combination 8 is: {year:2021,month:April,city:Jinan,index:National Happiness Index}.

[0196] The above examples are merely illustrative of the target field combinations, and this application does not impose any specific limitations on them.

[0197] Correspondingly, this application provides a fourth method for processing slot information, such as... Figure 15 As shown, the method may include:

[0198] Step S1501: After receiving the text to be processed, determine the feature information corresponding to the text to be processed.

[0199] Step S1502: Based on the slot information and the parallel conjunctions, determine multiple slot fields.

[0200] Step S1503: Sort the slot fields according to the position of the target slot information in the text to be processed, and bind adjacent slot fields with the same attributes.

[0201] Step S1504: For any slot field, if there is another slot field with the same attribute as the slot field in the previous initial field combination, then the slot field is assigned to the next initial field combination; otherwise, the slot field is assigned to the previous initial field combination.

[0202] Step S1505: For any initial field combination, add the target slot field of the missing attribute to the initial field combination, and combine the slot field bound in the initial field combination with the slot fields of other attributes in the initial field combination to obtain the target field combination.

[0203] The specific implementation of steps S1501 to S1505 can be referred to the above embodiments, and will not be repeated here.

[0204] This application provides a fifth method for processing slot information, such as... Figure 16 As shown, the method may include:

[0205] Step S1601: After receiving the text to be processed, determine the feature information corresponding to the text to be processed.

[0206] Step S1602: Based on the slot information and the parallel conjunctions, determine multiple slot fields.

[0207] Step S1603: Assign the multiple slot fields to obtain a combination of multiple target fields corresponding to the text to be processed.

[0208] The specific implementation of steps S1601 to S1603 can be referred to the above embodiments, and will not be repeated here.

[0209] Step S1604: For any target field combination, match the target field combination with the preset question corresponding to the intent information, and select the target question that is successfully matched from the preset question.

[0210] As mentioned above, the combination of multiple target fields fully reflects the user's true intent, and intelligent question answering can be performed accurately based on multiple target fields.

[0211] The slot design differs depending on the intent, and the pre-set questions matched also differ. For example, the intent of "What was the GDP of Qingdao and Jinan in 2021?" is to ask about the indicators, while the intent of "What was the GDP ranking of Laoshan District and Shibei District in Shandong Province in 2021?" is to ask about the ranking.

[0212] This embodiment does not specifically limit the above matching process. For example, it may determine whether the preset question contains the target field combination. If it does, it may determine that the preset question is successfully matched. Alternatively, it may determine whether the keywords of the preset question are the same as the target field combination. If they are the same, it may determine that the preset question is successfully matched.

[0213] Step S1605: Notify the answer information corresponding to the target question through a preset notification method.

[0214] For example, each preset question has a corresponding answer. The target question is a preset question that successfully matches the target field. Therefore, the answer to the target question is a response to the user's true intent. During the intelligent question answering process, these answer messages need to be notified, such as by sending them to the aforementioned terminal.

[0215] The above solution combines semantic understanding of the user's true intent to reasonably allocate slot information that may cause conflict or misunderstanding. It matches each target field combination with the preset question corresponding to the intent information, and then selects the target question that has successfully matched from the preset question. The target question is obtained by matching the user's true intent. Therefore, by notifying the user of the answer information corresponding to the target question, the accuracy of intelligent question answering is improved.

[0216] like Figure 17 As shown, based on the same inventive concept, this application provides a first slot information processing device 1700, comprising:

[0217] The feature extraction module 1701 is used to determine the feature information corresponding to the text to be processed after receiving the text to be processed; wherein, the feature information includes slot information and parallel conjunctions;

[0218] The field determination module 1702 is used to determine multiple slot fields based on the slot information and the parallel conjunctions; wherein any slot field contains slot information with the same attributes;

[0219] The field allocation module 1703 is used to allocate the plurality of slot fields to obtain a plurality of target field combinations corresponding to the text to be processed; wherein, any target field combination contains a slot field of each preset attribute.

[0220] In some optional implementations, the field determination module 1702 is specifically used for:

[0221] Based on the attributes of each slot information and the parallel conjunctions, all slot information is divided into slot information combinations corresponding to the attributes;

[0222] For any combination of slot information, based on the preset hierarchical relationship corresponding to the attributes of the combination of slot information, the slot information is combined to obtain at least one slot field.

[0223] In some optional implementations, the field determination module 1702 is specifically used for:

[0224] Slot information with the same attributes is grouped into the same initial slot information combination;

[0225] For any initial slot information combination, if the parallel conjunction is associated with slot information in the initial slot information combination, then it is determined whether the slot information associated with the parallel conjunction is at a different level; wherein, the slot information associated with the parallel conjunction is the slot information adjacent to the parallel conjunction.

[0226] If so, the slot information before the parallel conjunction and the slot information after the parallel conjunction in the initial slot information combination are divided into different sub-slot information combinations; otherwise, the initial slot information combination is not divided; wherein, the slot information is sorted according to its position in the text to be processed.

[0227] In some optional implementations, the field allocation module 1703 is specifically used for:

[0228] The slot fields are sorted according to the position of the target slot information in the text to be processed, and adjacent slot fields with the same attributes are bound together; wherein, the target slot information is the slot information corresponding to the lowest level among all slot fields;

[0229] For any slot field, if there is another slot field with the same attribute as the slot field in the previous initial field combination, then the slot field is assigned to the next initial field combination; otherwise, the slot field is assigned to the previous initial field combination.

[0230] For any initial field combination, the target slot field of the missing attribute is added to the initial field combination, and the slot field bound in the initial field combination is combined with the slot fields of other attributes in the initial field combination to obtain the target field combination.

[0231] In some alternative implementations, the target slot field for the missing attribute is determined in the following manner:

[0232] For any initial field combination, if the first field combination contains a slot field corresponding to the missing attribute, then the last slot field corresponding to the missing attribute in the first field combination is determined as the target slot field.

[0233] If there is no slot field corresponding to the missing attribute in the first field combination, then the first slot field corresponding to the missing attribute in the second field combination is determined as the target slot field.

[0234] Wherein, the first field combination is another initial field combination preceding the initial field combination; the second field combination is another initial field combination following the initial field combination.

[0235] In some optional implementations, the feature extraction module 1701 is specifically used for:

[0236] Based on the word vectors and part-of-speech vectors of the text to be processed, the fusion feature vector of the text to be processed is determined;

[0237] Based on the fused feature vector, the feature information corresponding to the text to be processed is identified.

[0238] See Figure 18 As shown, in some optional embodiments, the feature information further includes intent information; this application embodiment provides a second slot information processing device 1800, which, based on the above-mentioned slot information processing device 1700, further includes a notification module 1704, used for:

[0239] After the field allocation module 1703 allocates the multiple slot fields to obtain multiple target field combinations corresponding to the text to be processed, for any target field combination, the target field combination is matched with the preset question corresponding to the intent information, and the target question that is successfully matched is selected from the preset question.

[0240] The response information corresponding to the target question will be notified through a preset notification method.

[0241] Since this device is the same as the device in the method of this application embodiment, and the principle of the device in solving the problem is similar to that of the method, the implementation of the device can be referred to the implementation of the method, and the repeated parts will not be described again.

[0242] like Figure 19 As shown, based on the same inventive concept, this application provides an electronic device 1900, including: a processor 1901 and a memory 1902;

[0243] Memory 1902 may be volatile memory, such as random-access memory (RAM); memory 1902 may also be non-volatile memory, such as read-only memory, flash memory, hard disk drive (HDD), or solid-state drive (SSD); or memory 1902 may be any other medium capable of carrying or storing desired program code having the form of instructions or data structures and accessible by a computer, but is not limited thereto. Memory 1902 may be a combination of the above-described memories.

[0244] The processor 1901 may include one or more central processing units (CPUs), graphics processing units (GPUs), or digital processing units, etc.

[0245] This application embodiment does not limit the specific connection medium between the memory 1902 and the processor 1901. This application embodiment... Figure 19 The memory 1902 and the processor 1901 are connected via a bus 1903, and the bus 1903 is in... Figure 19 The bus 1903, represented by thick lines, can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 19 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0246] The memory 1902 stores program code, which, when executed by the processor 1901, causes the processor 1901 to perform the following processes:

[0247] Upon receiving the text to be processed, the feature information corresponding to the text to be processed is determined; wherein, the feature information includes slot information and parallel conjunctions;

[0248] Based on the slot information and the parallel conjunctions, multiple slot fields are determined; wherein any slot field contains slot information with the same attributes.

[0249] The multiple slot fields are assigned to obtain multiple target field combinations corresponding to the text to be processed; wherein any target field combination contains a slot field of each preset attribute.

[0250] In some alternative implementations, the processor 1901 specifically performs:

[0251] Based on the attributes of each slot information and the parallel conjunctions, all slot information is divided into slot information combinations corresponding to the attributes;

[0252] For any combination of slot information, based on the preset hierarchical relationship corresponding to the attributes of the combination of slot information, the slot information is combined to obtain at least one slot field.

[0253] In some alternative implementations, the processor 1901 specifically performs:

[0254] Slot information with the same attributes is grouped into the same initial slot information combination;

[0255] For any initial slot information combination, if the parallel conjunction is associated with slot information in the initial slot information combination, then it is determined whether the slot information associated with the parallel conjunction is at a different level; wherein, the slot information associated with the parallel conjunction is the slot information adjacent to the parallel conjunction.

[0256] If so, the slot information before the parallel conjunction and the slot information after the parallel conjunction in the initial slot information combination are divided into different sub-slot information combinations; otherwise, the initial slot information combination is not divided; wherein, the slot information is sorted according to its position in the text to be processed.

[0257] In some alternative implementations, the processor 1901 specifically performs:

[0258] The slot fields are sorted according to the position of the target slot information in the text to be processed, and adjacent slot fields with the same attributes are bound together; wherein, the target slot information is the slot information corresponding to the lowest level among all slot fields;

[0259] For any slot field, if there is another slot field with the same attribute as the slot field in the previous initial field combination, then the slot field is assigned to the next initial field combination; otherwise, the slot field is assigned to the previous initial field combination.

[0260] For any initial field combination, the target slot field of the missing attribute is added to the initial field combination, and the slot field bound in the initial field combination is combined with the slot fields of other attributes in the initial field combination to obtain the target field combination.

[0261] In some alternative implementations, the target slot field for the missing attribute is determined in the following manner:

[0262] For any initial field combination, if the first field combination contains a slot field corresponding to the missing attribute, then the last slot field corresponding to the missing attribute in the first field combination is determined as the target slot field.

[0263] If there is no slot field corresponding to the missing attribute in the first field combination, then the first slot field corresponding to the missing attribute in the second field combination is determined as the target slot field.

[0264] Wherein, the first field combination is another initial field combination preceding the initial field combination; the second field combination is another initial field combination following the initial field combination.

[0265] In some alternative implementations, the processor 1901 specifically performs:

[0266] Based on the word vectors and part-of-speech vectors of the text to be processed, the fusion feature vector of the text to be processed is determined;

[0267] Based on the fused feature vector, the feature information corresponding to the text to be processed is identified.

[0268] In some optional implementations, the feature information further includes intent information; after allocating the plurality of slot fields to obtain a plurality of target field combinations corresponding to the text to be processed, the processor 1901 further executes:

[0269] For any combination of target fields, the combination of target fields is matched with a preset question corresponding to the intent information, and a target question that is successfully matched is selected from the preset question.

[0270] The response information corresponding to the target question will be notified through a preset notification method.

[0271] Since the electronic device is the same electronic device that executes the method in the embodiments of this application, and the principle of the electronic device in solving the problem is similar to that of the method, the implementation of the electronic device can refer to the implementation of the method, and the repeated parts will not be described again.

[0272] This application provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the program implements the steps of the slot information processing method described above. The readable storage medium can be a non-volatile readable storage medium.

[0273] In some possible implementations, various aspects of this application may also be implemented as a program product comprising program code that, when the program product is run on an electronic device, causes the processor of the electronic device to perform the steps of the slot information processing method according to the various exemplary embodiments of this application described above.

[0274] The program product may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0275] like Figure 20 As shown, a program product 2000 according to an embodiment of this application is described, which may employ a portable compact disc read-only memory (CD-ROM) and include program code, and may run on an electronic device. However, the program product of this application is not limited thereto. In this document, a readable storage medium may be any tangible medium that contains or stores a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0276] The present application has been described above with reference to block diagrams and / or flowcharts illustrating methods, apparatus (systems), and / or computer program products according to embodiments of the present application. It should be understood that a block of a block diagram and / or flowchart, as well as combinations of blocks of block diagrams and / or flowcharts, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, and / or other programmable means to produce a machine, such that the instructions, executable via the computer processor and / or other programmable means, create methods for implementing the functions / actions specified in the blocks of the block diagrams and / or flowcharts.

[0277] Accordingly, this application can also be implemented using hardware and / or software (including firmware, resident software, microcode, etc.). Furthermore, this application can take the form of a computer program product on a computer-usable or computer-readable storage medium, having computer-usable or computer-readable program code implemented in the medium for use by or in conjunction with an instruction execution system. In the context of this application, a computer-usable or computer-readable medium can be any medium that can contain, store, communicate, transmit, or deliver a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0278] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0279] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for processing slot information, characterized in that, The method includes: Upon receiving the text to be processed, the feature information corresponding to the text to be processed is determined; wherein, the feature information includes slot information and coordinating conjunctions; Based on the slot information and the parallel conjunctions, multiple slot fields are determined; wherein any slot field contains slot information with the same attributes. The multiple slot fields are assigned to obtain multiple target field combinations corresponding to the text to be processed; wherein, any target field combination contains one slot field of each preset attribute; Based on the slot information and the parallel conjunctions, multiple slot fields are determined, including: Based on the attributes of each slot information and the parallel conjunctions, all slot information is divided into slot information combinations corresponding to the attributes; For any combination of slot information, the slot information is combined based on the preset hierarchical relationship corresponding to the attributes of the slot information combination to obtain at least one slot field; Based on the attributes of each slot information and the aforementioned parallel conjunctions, all slot information is divided into slot information combinations corresponding to the attributes, including: Slot information with the same attributes is grouped into the same initial slot information combination; For any initial slot information combination, if the parallel conjunction is associated with slot information in the initial slot information combination, then it is determined whether the slot information associated with the parallel conjunction is at a different level; wherein, the slot information associated with the parallel conjunction is the slot information adjacent to the parallel conjunction. If so, the slot information before the parallel conjunction and the slot information after the parallel conjunction in the initial slot information combination are divided into different sub-slot information combinations; otherwise, the initial slot information combination is not divided; wherein, the slot information is sorted according to its position in the text to be processed.

2. The method as described in claim 1, characterized in that, The multiple slot fields are assigned to obtain multiple target field combinations corresponding to the text to be processed, including: The slot fields are sorted according to the position of the target slot information in the text to be processed, and adjacent slot fields with the same attributes are bound together; wherein, the target slot information is the slot information corresponding to the lowest level among all slot fields; For any slot field, if there is another slot field with the same attribute as the slot field in the previous initial field combination, then the slot field is assigned to the next initial field combination; otherwise, the slot field is assigned to the previous initial field combination. For any initial field combination, the target slot field of the missing attribute is added to the initial field combination, and the slot field bound in the initial field combination is combined with the slot fields of other attributes in the initial field combination to obtain the target field combination.

3. The method as described in claim 2, characterized in that, The target slot field for the missing attribute is determined using the following method: For any initial field combination, if the first field combination contains a slot field corresponding to the missing attribute, then the last slot field corresponding to the missing attribute in the first field combination is determined as the target slot field. If there is no slot field corresponding to the missing attribute in the first field combination, then the first slot field corresponding to the missing attribute in the second field combination is determined as the target slot field. Wherein, the first field combination is another initial field combination preceding the initial field combination; the second field combination is another initial field combination following the initial field combination.

4. The method as described in claim 1, characterized in that, Determining the feature information corresponding to the text to be processed includes: Based on the word vectors and part-of-speech vectors of the text to be processed, the fusion feature vector of the text to be processed is determined; Based on the fused feature vector, the feature information corresponding to the text to be processed is identified.

5. The method according to any one of claims 1 to 4, characterized in that, The feature information also includes intent information; after allocating the multiple slot fields to obtain a combination of multiple target fields corresponding to the text to be processed, it also includes: For any combination of target fields, the combination of target fields is matched with a preset question corresponding to the intent information, and a target question that is successfully matched is selected from the preset question. The response information corresponding to the target question will be notified through a preset notification method.

6. An electronic device, characterized in that, Includes communication units and processors; The communication unit is used to receive the text to be processed; The processor is configured to, upon receiving a text to be processed, determine the feature information corresponding to the text to be processed; wherein the feature information includes slot information and coordinating conjunctions; based on the slot information and the coordinating conjunctions, determine multiple slot fields; wherein any slot field contains slot information with the same attributes; allocate the multiple slot fields to obtain multiple target field combinations corresponding to the text to be processed; wherein any target field combination contains a slot field for each preset attribute; The determination of multiple slot fields based on the slot information and the parallel conjunctions includes: Based on the attributes of each slot information and the parallel conjunctions, all slot information is divided into slot information combinations corresponding to the attributes; For any combination of slot information, the slot information is combined based on the preset hierarchical relationship corresponding to the attributes of the slot information combination to obtain at least one slot field; Based on the attributes of each slot information and the aforementioned parallel conjunctions, all slot information is divided into slot information combinations corresponding to the attributes, including: Slot information with the same attributes is grouped into the same initial slot information combination; For any initial slot information combination, if the parallel conjunction is associated with slot information in the initial slot information combination, then it is determined whether the slot information associated with the parallel conjunction is at a different level; wherein, the slot information associated with the parallel conjunction is the slot information adjacent to the parallel conjunction. If so, the slot information before the parallel conjunction and the slot information after the parallel conjunction in the initial slot information combination are divided into different sub-slot information combinations; otherwise, the initial slot information combination is not divided; wherein, the slot information is sorted according to its position in the text to be processed.

Images