An ultra-short character category intention prediction method, device and storage medium

Abstract

The application discloses a kind of super-short character's category intention prediction method, equipment and storage medium.Method includes the following steps: encoding processing to input character, generate feature data;Calculate the prior probability of each category and the conditional probability of feature data in each category;According to the prior probability of each category and the conditional probability of feature data in each category, the final probability of feature data in each category is obtained;The one or more categories with maximum final probability are output as the category prediction result of input character.The application classifies feature data based on one-bit effective coding and two-bit effective coding, unifies the data structure of feature data, and can overcome the problem of disordered input data structure of natural language text.The application uses the calculation of prior probability, conditional probability and final probability judgment to predict the category of feature data, which can efficiently process a large amount of data sets while avoiding the compatibility problem of character matching rules.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data classification prediction in machine learning, and in particular to a super-short character category intention prediction method, device and storage medium. BACKGROUND

[0002] Natural language text has the characteristics of large content, multiple categories, disordered input data structure, and non-uniform data, which makes the classification of natural language text a problem to be solved. In the prior art, a dictionary is usually added manually, and the input natural language text is classified by preset character matching rules. Although manual operation can accurately meet the implementation of the classification method in some requirements, due to the disordered input data structure of natural language text, text classification cannot be correctly completed in some cases. In addition, manual operation to add classification rules not only occupies human resources, and the change of classification rules may conflict with the previously preset classification rules. SUMMARY

[0003] Therefore, the present application provides a super-short character category intention prediction method, device and storage medium.

[0004] The first aspect of the present application provides a super-short character category intention prediction method, comprising the following steps:

[0005] encoding the input character to generate feature data;

[0006] calculating the prior probability of each category and the conditional probability of the feature data in each category;

[0007] obtaining the final probability of the feature data in each category according to the prior probability of each category and the conditional probability of the feature data in each category;

[0008] outputting one or more categories with the maximum final probability as the category prediction result of the input character.

[0009] Further, the encoding of the input character to generate feature data specifically comprises the following steps:

[0010] one-bit effective encoding of the input character to obtain a first array, a second array and a third array;

[0011] two-bit effective encoding of the input character to obtain a two-bit array;

[0012] assigning respective weights to the first array, the second array, the third array and the two-bit array according to a preset weight allocation rule to generate feature data corresponding to the input character.

[0013] Further, the one-bit effective encoding processing of the input character specifically comprises the following steps.

[0014] According to the preset separation rule, the input character is separated to obtain a plurality of character strings;

[0015] The one-bit effective encoding processing is performed on the character string with a length of no more than three characters to obtain a first array of the input character;

[0016] The one-bit effective encoding processing is performed on the character string with a length of four characters to obtain a second array of the input character;

[0017] The one-bit effective encoding processing is performed on the first character and the last character of the character string to obtain a third array of the input character.

[0018] Further, the one-bit effective encoding processing is based on an encoding dictionary containing 36 character identifiers of English letters a-z and numbers 0-9; when the one-bit effective encoding processing is performed, the corresponding character identifier of the character in the encoding dictionary is set to 1, and the rest of the character identifiers are set to 0, thereby completing the one-bit effective encoding processing.

[0019] Further, the two-bit effective encoding is based on an encoding dictionary containing 1369 character identifiers of English letters a-z, numbers 0-9 and the character '+' in any two combinations; when the two-bit effective encoding processing is performed, the two adjacent characters in the input character are taken as a character whole, the corresponding character identifier of the character whole in the encoding dictionary is set to 1, and the rest of the character identifiers are set to 0, thereby completing the two-bit effective encoding processing.

[0020] Further, the calculation of the prior probability of each category and the conditional probability of the feature data in each category specifically comprises the following steps:

[0021] The prior probability of the category c is calculated by the following formula:

[0022]

[0023] Wherein, P(C=c) represents the prior probability of the category c, N represents the total number of documents in the classification data set, the classification data set is provided with m categories (a, b, c, …, m), and Nc represents the number of documents belonging to the target category c in the classification data set;

[0024] The conditional probability of the feature data in the target category c is calculated by the following formula:

[0025]

[0026] Wherein, P(w i |c) represents the conditional probability of the character w iThe conditional probability in the category c, ΣW(Nc) represents the sum of the number of characters in the document belonging to the category c, Σw i (Nc) represents the sum of the number of characters in the document belonging to the category c, Σw i The sum of the number of occurrences in the document belonging to the category c;

[0027] The prior probability of each category (a, b, c, …, m) and the conditional probability of each character (w1, w2, …, w i ) in the feature data in each category (a, b, c, …, m) is calculated by the above formula; the prior probability P(C=a), P(C=b), P(C=c), … P(C=m) of each category and the conditional probability P(w1|a), P(w1|b), …, P(w1|m), P(w2|a), …, P(w2|m), …, P(w i |m) of the feature data in each category are obtained.

[0028] Further, the final probability of the feature data in each category is obtained according to the prior probability of each category and the conditional probability of the feature data in each category, and specifically includes the following steps:

[0029] The prior probability of each category is multiplied by the conditional probability of each character in the feature data in each category to obtain the final probability of the feature data in each category:

[0030] P(a) = P(C=a) × P(w1|a) × P(w2|a) × … × P(w i |a);

[0031] P(b) = P(C=b) × P(w1|b) × P(w2|b) × … × P(w i |b);

[0032] …

[0033] P(m) = P(C=m) × P(w1|m) × P(w2|m) × … × P(w i |m);

[0034] Where P(a), P(b), …, P(m) represent the final probability of the feature data in categories a, b, …, m, respectively.

[0035] Further, before the one or more categories with the maximum final probability are output as the category prediction result of the input character, the following steps are further included:

[0036] The final probability of the one or more categories with the maximum final probability is compared with a preset probability threshold, and the one or more categories with a final probability less than the preset probability threshold are removed.

[0037] The second aspect of the present application discloses an electronic device comprising a processor and a memory;

[0038] The memory is configured to store a program;

[0039] The processor executes the program to implement a category intention prediction method of an ultra-short character.

[0040] The third aspect of the present application discloses a computer readable storage medium, the storage medium stores a program, and the program is executed by a processor to implement a category intention prediction method of an ultra-short character.

[0041] The embodiments of the present application have the following beneficial effects: the present application classifies feature data based on one-bit effective coding and two-bit effective coding, unifies the data structure of the feature data, and can overcome the problem of disordered input data structure of natural language text. The present application uses the calculation of prior probability, conditional probability and final probability judgment to predict the category of the feature data, can efficiently process a large amount of data sets, and at the same time, converts the demand for adding character matching rules into the demand for using new classified data sets, avoiding the compatibility problem of character matching rules.

[0042] Additional aspects and advantages of the present application will be described in the following description part, some of which will become apparent from the following description, or will be understood by those skilled in the art through practice of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0044] Figure 1 is a basic flow diagram of a category intention prediction method of an ultra-short character, device and storage medium of the present application. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solutions and advantages of the present application more clear, the present application will be further described in detail in the following with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0046] In the case of sufficient data and correct data processing and model selection debugging, the machine learning model can also achieve a certain level of natural language classification ability. Using the machine learning model can reduce the consumption of human resources while maintaining high accuracy, improve the ability to process messy natural language input, and process a large amount of data quickly, and can continue to learn by inputting new data to improve accuracy.

[0047] Thus, the embodiment provides a category intention prediction method for ultra-short characters, which mainly includes the following steps as shown in the formula (1): Figure 1

[0048] S1. Encode the input characters to generate feature data.

[0049] S2. Calculate the prior probability of each category and the conditional probability of the feature data in each category.

[0050] S3. According to the prior probability of each category and the conditional probability of the feature data in each category, the final probability of the feature data in each category is obtained.

[0051] S4. Output one or more categories with the maximum final probability as the category prediction result of the input characters.

[0052] The category prediction of the ultra-short character in the embodiment mainly includes two aspects, the data encoding aspect of step S1 and the probability calculation aspect of steps S2-S3, and finally the prediction result is output in step S4. The execution process of each step of the embodiment is discussed in detail as follows:

[0053] S1. Encode the input characters to generate feature data.

[0054] Since natural language, whether Chinese, English or other characters, cannot be directly put into the model for calculation, and the essence of model classification is to use a certain mathematical formula and function to calculate a certain classification, the input characters need to be encoded before being input into the model. Appropriate encoding can improve the accuracy of the model to a certain extent in solving actual classification problems. The encoding method used in the embodiment mainly has two types, which are one-bit effective encoding and two-bit effective encoding.

[0055] ​One-Hot Encoding: One-Hot Encoding is also known as One-Hot Encoding. The principle is to give all possible single characters a mark. When the input character is One-Hot encoded, the encoding dictionary is traversed, and the corresponding mark position of the input character in the mark group is 1, and the rest of the mark positions are 0. Before One-Hot encoding, the input character is first separated according to the delimiter to form a string to improve the encoding efficiency. In this embodiment, spaces are used as delimiters for input characters. For example, the input character "apple pie" contains two strings "apple" and "pie".

[0056] One-Hot Encoding is mainly used to emphasize some special or important characters. In this embodiment, three important characters are selected for One-Hot encoding: the first type is a character with a string length of not more than 3. This type of string usually directly helps the model to point to part of the classification result; the second type is a character with a string length of 4. This type of string also directly helps the model to point to part of the classification result and does not coincide with the first type of string; the third type is the first and last characters of the string. Each category will form an array after encoding, respectively denoted as the first array, the second array and the third array.

[0057] The encoding dictionary of One-Hot Encoding contains English letters a-z and numbers 0-9, a total of 36 character mark bits, that is, {'a': 0, 'b': 1, 'c': 2, 'd': 3, 'e': 4, 'f': 5, 'g': 6, 'h': 7, 'i': 8, 'j': 9, 'k': 10, 'l': 11,'m': 12, 'n': 13, 'o': 14, 'p': 15, 'q': 16, 'r': 17,'s': 18, 't': 19, 'u': 20, 'v': 21, 'w': 22, 'x': 23, 'y': 24, 'z': 25, '0': 26, '1': 27, '2': 28, '3': 29, '4': 30, '5': 31, '6': 32, '7': 33, '8': 34, '9': 35}.

[0058] The following is an example of one-bit efficient coding of "apple pie". The string "apple pie" is divided into two strings "apple" and "pie" by the space as a separator. The length of "apple" is 5, which does not meet the classification requirements of the first and second categories. The length of "pie" is 3, which meets the classification requirement of the first category. Therefore, in the one-bit efficient coding of "apple pie", the character identification positions of "p", "i" and "e" in the first array are 1, and the rest are 0. The second array is all 0 because there is no string in the input characters that meets the classification condition (length of 4). In the third array, the character identification positions of the first characters "a" and "p" and the last character "e" of "apple" and "pie" are 1, and the rest are 0. Thus, the one-bit efficient coding of the character "apple pie" is completed.

[0059]

[0060] The principle of two-bit efficient coding is similar to that of one-bit efficient coding, with the difference being that two-bit efficient coding identifies two adjacent characters as a whole to obtain a two-bit array. For example, the two-bit efficient coding of the string "you" will have the corresponding identification positions of "yo" and "ou" as 1.

[0061] The encoding dictionary of two-bit effective encoding includes two-by-two arbitrary combinations of English letters a~z, numbers 0~9, and the character '+' of a total of 1369 character identification bits, that is, {'aa': 0, 'ab': 1, 'ac': 2, 'ad': 3, 'ae': 4, 'af': 5, 'ag': 6, 'ah': 7, 'ai': 8, 'aj': 9, 'ak': 10, 'al': 11, 'am': 12, 'an': 13, 'ao': 14, 'ap': 15, 'aq': 16, 'ar': 17, 'as': 18, 'at': 19, 'au': 20, 'av': 21, 'aw': 22, 'ax': 23, 'ay': 24, 'az': 25, 'ba': 26, 'bb': 27, 'bc': 28, 'bd': 29, 'be': 30, 'bf': 31, 'bg': 32, 'bh': 33, 'bi': 34, 'bj': 35, 'bk': 36, 'bl': 37, 'bm': 38, 'bn': 39, 'bo': 40, 'bp': 41, 'bq': 42, 'br': 43, 'bs': 44, 'bt': 45, 'bu': 46, 'bv': 47, 'bw': 48, 'bx': 49, 'by': 50, 'bz': 51, 'ca': 52, 'cb': 53, 'cc': 54, 'cd': 55, 'ce': 56, 'cf': 57, 'cg': 58, 'ch': 59, 'ci': 60, 'cj': 61, 'ck': 62, 'cl': 63, 'cm': 64, 'cn': 65, 'co': 66, 'cp': 67, 'cq': 68, 'cr': 69, 'cs': 70, 'ct': 71, 'cu': 72, 'cv': 73, 'cw': 74, 'cx': 75, 'cy': 76, 'cz': 77, 'da': 78, 'db': 79, 'dc': 80, 'dd': 81, 'de': 82, 'df': 83, 'dg': 84, 'dh': 85, 'di': 86, 'dj': 87, 'dk': 88, 'dl': 89, 'dm': 90, 'dn': 91, 'do': 92, 'dp': 93, 'dq': 94, 'dr': 95, 'ds': 96, 'dt': 97, 'du': 98, 'dv': 99...}.

[0062] The following is an example of "apple pie" to illustrate the process of two-bit effective encoding. In the character "apple pie", there is a space as a delimiter, so it is divided into "apple" and "pie". Since two-bit effective encoding identifies two adjacent characters as a whole, the character identification bit needs to be set to 1 for "ap", "pp", "pl", "le", "pi", and "ie". Since "apple" and "pie" are separated, "ep" does not need to be set to 1.

[0063] This embodiment mainly uses two-bit effective encoding for encoding, so it needs to allocate weights to the first array, the second array, the third array, and the two-bit array. In this embodiment, the weight of the two-bit array is 0.7, the weight of the first array and the second array is 0.2 (the first array and the second array do not overlap); the weight of the third array is 0.1. In this way, the data encoding process of the input character is completed.

[0064] In some embodiments, three-bit effective encoding or bag-of-words encoding can be used instead of the two-bit effective encoding in this embodiment. Three-bit effective encoding refers to identifying three adjacent strings as a whole, which has the same principle as two-bit effective encoding in this embodiment. Bag-of-words encoding refers to identifying the entire string as a whole.

[0065] S2. Calculate the prior probability of each category and the conditional probability of the feature data in each category;

[0066] S3. According to the prior probability of each category and the conditional probability of the feature data in each category, the final probability of the feature data in each category is obtained.

[0067] Steps S2-S3 are mainly based on the multinomial naive Bayes model (MultinomialNB). The multinomial naive Bayes model assumes that the probability of each result category follows a simple multinomial distribution, i.e. each category is independent and does not interfere with each other. The simplest example is coin tossing, the probabilities of heads and tails are independent and do not interfere with each other. In one experiment, if a head appears, a tail cannot appear. The training set used in this application corresponds to only one classification result for each input. For each input, there is only one corresponding result, so it is very suitable to use the multinomial naive Bayes model. Even in the actual use process, a single input has multiple results, the model can also identify the probability of each result and find the higher probability result to get multiple output results.

[0068] The specific implementation method is as follows:

[0069] The prior probability of category c is calculated by the following formula:

[0070]

[0071] wherein P(C=c) represents the prior probability of the category c, N represents the total number of documents in the classification data set, m categories (a, b, c, …, m) are provided in the classification data set, and Nc represents the number of documents belonging to the target category c in the classification data set;

[0072] The conditional probability of the feature data in the target category c is calculated by the following formula:

[0073]

[0074] wherein P(w i |c) represents the conditional probability of the character w i in the feature data in the category c, and∑W(Nc) represents the sum of the number of characters in the documents belonging to the category c, and∑W i (Nc) represents the sum of the number of characters w i in the documents belonging to the category c.

[0075] The prior probability of each category (a, b, c, …, m) and the conditional probability of each character (w1, w2, …, w i ) in the feature data in each category (a, b, c, …, m) are calculated by the above formula respectively; the prior probability P(C=a), P(C=b), P(C=c), …, P(C=m) of each category and the conditional probability P(w1|a), P(w1|b), …, P(w1|m), P(w2|a), …, P(w2|m), …, P(w i |m) of the feature data in each category are obtained.

[0076] The prior probability of each category is multiplied by the conditional probability of each character in the feature data in each category to obtain the final probability of the feature data in each category:

[0077] P(a)=P(C=a)×P(w1|a)×P(w2|a)×…×P(w i |a);

[0078] P(b)=P(C=b)×P(w1|b)×P(w2|b)×…×P(w i |b);

[0079] …

[0080] P(m)=P(C=m)×P(w1|m)×P(w2|m)×…×P(w i |m);

[0081] P(a), P(b), …, P(m) represent the final probability of the feature data in the categories a, b, …, m respectively.

[0082] The polynomial naive Bayes model uses the prior probability of the category c to multiply the conditional probability of each word of the input character in the category c to obtain the final probability of the input character in the category c. The category with the maximum final probability is output as the result after traversing all categories.

[0083] S4. Output one or more categories with the maximum final probability as the category prediction result of the input character.

[0084] In the embodiment, a preset probability threshold can be set before the category prediction result is output, and one or more categories with a final probability less than the preset probability threshold are removed through comparison of the probability threshold and the category prediction result, which helps to improve the reliability of model prediction.

[0085] The embodiment of the application further discloses a computer program product or a computer program, which comprises computer instructions stored in a computer readable storage medium. A processor of a computer device can read the computer instructions from the computer readable storage medium, and the processor executes the computer instructions to enable the computer device to execute the method shown in the embodiment. Figure 1

[0086] In some alternative embodiments, the functions / operations mentioned in the block diagram can not occur in the order mentioned in the operation diagram. For example, depending on the functions / operations involved, two blocks shown in succession can actually be executed substantially simultaneously or the blocks can sometimes be executed in reverse order. In addition, the embodiments presented and described in the flowcharts of the application are provided by way of example, with the purpose of providing a more comprehensive understanding of the technology. The disclosed method is not limited to the operations and logical flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and in which sub-operations described as part of larger operations are independently executed.

[0087] ​Furthermore, although the present application has been described in the context of functional modules, it is to be understood that one or more of the functions and / or features described can be integrated in a single physical device and / or software module, or one or more functions and / or features can be implemented in separate physical devices or software modules. It will also be appreciated that specific details are set forth in order to provide a thorough understanding of the application. However, it will be apparent to one of ordinary skill in the art that the application can be practiced without these details being present. In other instances, well-known methods, procedures and components have not been described in detail so as not to unnecessarily obscure aspects of the application. Further, it will be apparent to one of ordinary skill in the art that the application, as described herein, can be practiced in the absence of any element that is not specifically disclosed. In other instances, various articles, materials, and components have been described in detail so as to provide a thorough understanding of the application. However, it will be apparent to one of ordinary skill in the art that the application, as described herein, can be practiced in the absence of these specifically disclosed articles, materials, and components. Also, it will be appreciated by persons of the art that the present application is not limited to what has been described in the above description. Rather, the words "include," "comprise," "no exclude," "in addition to," and "contain" and variations thereof, mean "including but not limited to," and do not exclude other moieties, additives, components, integers or steps. Additionally, the terms "coupled" and "connected" and variations thereof are intended to

[0088] In the description of the specification, reference has been made to the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" and the like meant to include a specific feature, structure, material or characteristic described in connection with at least one embodiment or example. The appearances of such terms in various places of the specification are not necessarily referring to the same embodiment or example. Furthermore, the described specific features, structures, materials or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0089] While embodiments of the application have been shown and described, it is to be understood that various additional modifications and substitutions can be undertaken by one of ordinary skill in the art without departing from the spirit of the present application. The scope of the application is not to be understood as being limited to the particular embodiments described and as indicated by the claims below.

[0090] The above is a specific description of the preferred embodiments of the present application, but the present application is not limited to the above-described embodiments, and those skilled in the art can make various equivalent modifications or replacements without departing from the spirit of the present application, and these equivalent modifications or replacements are included in the scope of the present application defined by the claims.

Claims

1. A method for predicting category intent using ultra-short characters, characterized in that, The method comprises the following steps: encoding the input character to generate feature data; calculating the prior probability of each category and the conditional probability of the feature data in each category; obtaining the final probability of the feature data in each category according to the prior probability of each category and the conditional probability of the feature data in each category; outputting one or more categories with the maximum final probability as the category prediction result of the input character; The encoding of the input character to generate feature data specifically comprises the following steps: one-bit effective encoding of the input character to obtain a first array, a second array and a third array; two-bit effective encoding of the input character to obtain a two-bit array; assigning corresponding weights to the first array, the second array, the third array and the two-bit array according to a preset weight allocation rule to generate feature data corresponding to the input character; The two-bit effective encoding is based on an encoding dictionary containing 1369 character identifier bits of any two arbitrary combinations of English letters a-z, numbers 0-9 and characters +. When performing two-bit effective encoding, two adjacent characters in the input character are taken as a character whole, and the character identifier position corresponding to the character whole in the encoding dictionary is set to 1, and the rest of the character identifier positions are set to 0, thereby completing the two-bit effective encoding. 2.The super-short character category intent prediction method of claim 1, wherein, The one-bit effective encoding of the input character specifically comprises the following steps: separating the input character according to a preset separation rule to obtain a plurality of strings; one-bit effective encoding of a string with a length of no more than three characters as the first array of the input character; one-bit effective encoding of a string with a length of four characters as the second array of the input character; one-bit effective encoding of the first character and the last character of the string as the third array of the input character. 3.The super-short character category intent prediction method of claim 2, wherein, The one-bit effective encoding is based on an encoding dictionary containing 36 character identifier bits of English letters a-z and numbers 0-9. When performing one-bit effective encoding, the character identifier position corresponding to the character in the encoding dictionary is set to 1, and the rest of the character identifier positions are set to 0, thereby completing the one-bit effective encoding. 4.The method of claim 1, wherein, The calculation of the prior probability of each category and the conditional probability of the feature data in each category specifically comprises the following steps: The prior probability of category c is calculated by the following formula: Wherein, P(C=c) represents the prior probability of category c, N represents the total number of documents in the classification data set, the classification data set is provided with m categories (a, b, c, …, m), and Nc represents the number of documents belonging to the target category c in the classification data set; The conditional probability of the feature data in the target category c is calculated by the following formula: Wherein, P(wi|c) represents the conditional probability of character wi in the feature data in category c, ΣW(Nc) represents the sum of the number of characters in the documents belonging to category c, and Σwi(Nc) represents the sum of the number of occurrences of character wi in the documents belonging to category c; The prior probability of each category (a, b, c,..., m) and the conditional probability of each character (w1, w2,..., wn) in the feature data in each category (a, b, c,..., m) are calculated by the above formulas; the prior probability of each category P(C=a), P(C=b), P(C=c),..., P(C=m) and the conditional probability of the feature data in each category P(w1|a), P(w1|b),..., P(w1|m), P(w2|a),..., P(w2|m),..., P(wn|a), P(wn|b),..., P(wn|m) are obtained. i i ​​ 5. The method of claim 4, wherein the category intent prediction of an ultra-short character is performed by using a neural network model. The calculation of the final probability of the feature data in each category according to the prior probability of each category and the conditional probability of the feature data in each category specifically comprises the following steps: The prior probability of each category is multiplied by the conditional probability of each character in the feature data in each category to obtain the final probability of the feature data in each category: ； ； …… ； Wherein, P(a), P(b), …, P(m) represent the final probability of the feature data in categories a, b, …, m respectively.

6. The method of claim 1, wherein the category intent prediction of an ultra-short character is performed by using a neural network model. Before outputting the one or more categories with the maximum final probability as the category prediction result of the input character, the following steps are further included: The final probability of the one or more categories with the maximum final probability is compared with a preset probability threshold, and the one or more categories with the final probability less than the preset probability threshold are removed.

7. An electronic device, comprising: It comprises a processor and a memory; The memory is used to store a program; The processor executes the program to realize the method of any one of claims 1-6.

8. A computer-readable storage medium, characterized in that, The storage medium stores a program, and the program is executed by the processor to realize the method of any one of claims 1-6. The storage medium stores a program, and the program is executed by the processor to realize the method of any one of claims 1-6.

Images