KR20190083143A - Sensory evaluation method and apparatus - Google Patents

Abstract

Disclosed are a sensory evaluation method with regard to products and a device thereof. The sensory evaluation method comprises the steps of: (a) collecting product review data in accordance with input conditions online or on SNS; (b) analyzing the product review data based on a preset taste expression dictionary and selecting an evaluation candidate factor word; (c) after deducting a vector value with regard to the selected evaluation candidate factor word, embedding the vector value in a vector space, individually calculating similarity between embedded vector values and characteristic keyword vector values, and selecting n (natural number) number of evaluation factor words having high similarity; and (d) classifying the evaluation factor words in accordance with emotion based on the taste expression dictionary to generate a final sensory evaluation result with regard to products.

Description

[0001] The present invention relates to a sensory evaluation method and apparatus,

The present invention relates to a sensory evaluation method and apparatus therefor.

There are face-to-face surveys conducted by food companies that invite or visit consumers and use their own tastes in order to evaluate the taste and characteristics of products.

This approach is marginal and costly for the size of the recruitment group for evaluation and is accompanied by difficulties in making words for appropriate questions to assess the consumer's taste. In addition, the conventional method has a problem that the subjective intention of the questioner who writes the question (question) is intervened.

The present invention is to provide a sensory evaluation method and apparatus for a product.

The present invention is intended to provide a sensory evaluation method and device for automatically selecting a comparison item for sensory evaluation on a review on the on-line or the SNS, thereby performing a sensory evaluation on the product.

The present invention is to provide a sensory evaluation method and device that can prevent the subjective intention of the questioner from being reflected in the conventional sensory evaluation and can provide an objective sensory evaluation.

According to one aspect of the present invention, there is provided a sensory evaluation method for a product.

According to an embodiment of the present invention, there is provided a sensory evaluation method for a product, comprising the steps of: (a) collecting product review data according to an input condition on-line or in an SNS; (b) analyzing the product review data based on a preset taste expression dictionary to select evaluation candidate element words; (c) a vector value for the selected evaluation candidate element word is derived and embedded in the vector space, and similarity rolls between the embedded vector value and the feature keyword vector values are calculated, and the top n (natural number) evaluation elements Selecting words; And (d) sorting the evaluation element word according to emotion based on the taste expression dictionary to generate a final sensory evaluation result for the product.

The step (d) may classify the evaluation element word according to emotion, based on positive or negative word data defined in the taste expression dictionary.

Wherein the step (b) includes automatically selecting the evaluation candidate element word based on the frequency of each word included in the product review data, wherein the evaluation candidate element word is automatically selected by the sensory evaluation comparison element to be.

In the step (b), the evaluation candidate element word is extracted through morphological analysis using the taste expression dictionary. The product review data is morphologically analyzed on a sentence-by-sentence basis, wherein a frequency among words corresponding to a noun, an adjective, A high number of words can be extracted as the evaluation candidate word, and the words corresponding to the search and personal pronouns can be excluded.

In the step (c), before selecting the evaluation element word, a deep learning model is learned by using a vector value of a word having a high frequency among the embedded vector values, and the deep learning model is classified into the selected deep learning model It is possible to input the vector value of the evaluation candidate word and remove the vector value of the meaningless word.

The characteristic keyword vector value may be changed dynamically according to the detailed comparison element.

According to another aspect of the present invention, there is provided an apparatus for sensory evaluation of a product.

According to an embodiment of the present invention, there is provided an information processing system comprising: a collecting unit for collecting product review data according to input conditions on-line or in an SNS; A word extracting unit for analyzing the product review data based on a preset taste expression dictionary to select evaluation candidate element words; The vector values for the selected evaluation candidate element words are derived and embedded in the vector space. The similarity degree rolls between the embedded vector values and the feature keyword vector values are calculated, and the top n (natural number) evaluation element words having high similarity are selected And an analyzer for sorting the evaluation element word according to emotion based on the taste expression dictionary to generate a final sensory evaluation result on the product.

According to another embodiment of the present invention, there is provided a computing device comprising: a memory for storing at least one application; And a processor coupled to the memory and executing an application stored in the memory, wherein the application executed by the processor comprises: (a) collecting product review data according to an input condition on-line or in an SNS; (b) analyzing the product review data based on a preset taste expression dictionary to select evaluation candidate element words; (c) a vector value for the selected evaluation candidate element word is derived and embedded in the vector space, and similarity rolls between the embedded vector value and the feature keyword vector values are calculated, and the top n (natural number) evaluation elements Selecting words; And (d) generating the final sensory evaluation result for the product by classifying the evaluation element word according to emotion based on the taste representation dictionary.

By providing the sensory evaluation method and apparatus according to an embodiment of the present invention, there is an advantage that the sensory evaluation can be performed on the product by automatically selecting the comparison items for sensory evaluation on a review on the on-line or the SNS.

Accordingly, the present invention can prevent the subjective intention of the questioner from being reflected in the conventional sensory evaluation, and it is possible to more objectively perform the sensory evaluation on the product.

1 is a flowchart showing a sensory evaluation method for a food company product according to an embodiment of the present invention;
FIG. 2 illustrates a detailed process of collecting review data (product review data) on a product in an embodiment of the present invention or on the SNS; FIG.
3 is a diagram illustrating words selected through morphological analysis according to an embodiment of the present invention.
FIG. 4 illustrates a continuation schemegram according to an embodiment of the present invention. FIG.
5 illustrates an output layer according to a weighting application according to an embodiment of the present invention.
Figure 6 illustrates a final evaluation element word output in accordance with an embodiment of the present invention.
7 is a graphical representation of the values of the last selected evaluation element word by applying a weight to a vector value corresponding to the feature keyword;
8 is a block diagram of a product sensory evaluation device according to an embodiment of the present invention.
9 is a diagram illustrating a word embedding process based on CNN model using a deep learning technique according to an embodiment of the present invention.

As used herein, the singular forms "a", "an" and "the" include plural referents unless the context clearly dictates otherwise. In this specification, the terms "comprising ", or" comprising ", etc. should not be construed as necessarily including the various elements or steps described in the specification, Or may be further comprised of additional components or steps. Also, the terms "part," " module, "and the like described in the specification mean units for processing at least one function or operation, which may be implemented in hardware or software or a combination of hardware and software .

The present invention relates to an invention capable of automatically performing a food sensory test using a Deep Learning technique after automatically extracting an evaluation target item based on big data collected online. Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 is a flowchart showing a sensory evaluation method for a food company product according to an embodiment of the present invention. FIG. 2 is a flowchart illustrating a method of collecting review data (product review data) FIG. 3 is a diagram showing words selected through morpheme analysis according to an embodiment of the present invention, FIG. 4 is a diagram illustrating a continuous schemegram according to an embodiment of the present invention, and FIG. And FIG. 5 is a diagram illustrating an output layer according to an embodiment of the present invention. FIG. 6 is a diagram illustrating a final evaluation element word output according to an embodiment of the present invention. And the value of the finally selected evaluation element word by applying a weight to the vector value.

In step 110, the sensory evaluation apparatus 100 collects review data (hereinafter referred to as product review data) on products on-line or on the SNS.

At this time, the sensory evaluation apparatus 100 can collect product review data that meets the input conditions. The sensory evaluation apparatus 100 can receive product reviews data corresponding to a specific product (for example, product A, product B, etc.) as an input condition, and collect the corresponding product review data on-line or on the SNS.

For example, it is assumed that the product A and the product B are inputted to the sensory evaluation apparatus 100 under the input conditions. The sensory evaluation apparatus 100 can collect and store review data on each of the products A and B on-line and on the SNS.

In addition, the sensory evaluation apparatus 100 generates a sub-item category for the creators on-line or on the SNS such as sex, age, and area when collecting the product review data according to the input conditions on-line or on the SNS, For example.

There are many ways in which the sensory evaluation apparatus 100 collects product review data on-line or on the SNS. However, in the embodiment of the present invention, it is assumed that product review data (sentences) are collected on-line or on the SNS using a crawling technique This will be mainly described below.

For example, based on the data crawling process, the crawler can be implemented using the Python language to collect data according to input conditions. Since the crawling process itself is obvious to those skilled in the art, a detailed description thereof will be omitted.

FIG. 2 illustrates a detailed process of collecting review data (product review data) on products on-line or on the SNS.

A method of collecting product review data will be briefly described with reference to FIG.

In step 210, the sensory evaluation apparatus 100 inputs the input condition (or a search word) to the seed URL to search for a product review (taste, flavor, evaluation, analysis, and the like) to acquire a FULL URL.

In step 215, the sensory evaluation apparatus 100 determines whether or not the obtained URL is opened. If not, proceed to step 210.

However, if the URL is opened, the sensory evaluation apparatus 100 accesses the page through the obtained URL, crawls the page, and stores the crawled page in the DB as product review data.

For example, the sensory evaluation apparatus 100 can parse a page of the obtained URL based on an HTML tag, crawl product review data (sentence), and store it in the DB.

The above process is repeated until the last URL is accessed, and data can be collected.

The data crawling process illustrated in FIG. 2 is shown for the convenience of understanding and explanation, and it is natural that the product review data according to input conditions can be collected through other methods besides data crawling.

Referring again to FIG. 1, the product review data collected by the sensory evaluation apparatus 100 may each be in the form of a sentence.

In step 115, the sensory evaluation apparatus 100 analyzes the product review data based on the preset taste expression dictionary, and selects evaluation candidate element words.

At this time, the sensory evaluation apparatus 100 may morphologically analyze the product review data based on the preset taste expression dictionary to select evaluation candidate element words. The sensory evaluation apparatus 100 can remove the useless morpheme (e.g., examination) in the process of morphological analysis of the product review data based on the taste expression dictionary.

It is common to use only nouns in morphological analysis. However, the sensory evaluation apparatus 100 according to an embodiment of the present invention is for sensory evaluation for comparison between specific products or products, and may include words meaningful to adjectives and verbs in addition to nouns.

Thus, in one embodiment of the present invention, evaluation candidate element words including nouns, adjectives, and verbs can be selected from each of the product review data during morphological analysis.

In this way, more selective use of adjectives and verbs in addition to nouns makes the meaning of nouns different by adjectives and verbs, which makes it useful to analyze emotions of selected nouns. At this time, a Korean analysis package (KoNLPy) can be used for morphological analysis.

Further, in morphological analysis of the product review data, the sensory evaluation apparatus 100 selects meaningful evaluation candidate word words suitable for the sensory evaluation by limiting the tastant expression dictionary to taste or flavor related words in order to increase the calculation speed .

In the embodiment of the present invention, the sensory evaluation for foods is assumed, and thus the taste expression dictionary is used. However, it is a matter of course that the dictionary to be used may be changed when the sensory evaluation object is changed.

Here, the taste expression dictionary collects data from the food evaluation site such as the taste evaluation data of the SNS and the menu plate, the yogi yo, the delivery trough, and the nation of delivery gathered using the keywords capable of expressing a general taste such as a restaurant, And a dictionary summarizing expressions related to flavors that are commonly used based on frequency.

Therefore, the dictionary of taste expressions has various words about taste and incense for food, food and so on.

In one embodiment of the present invention, the morphological analysis of the product review data based on the tastant expression dictionary, and the significant word used in the evaluation and analysis of the food or food are selected as the evaluation candidate word, There is an advantage.

FIG. 3 shows an example of words selected through morphological analysis.

As shown in FIG. 3, one of the collected product review data is assumed to be "I ate XX chanpon ramen noodles, which is rich in seafood flavor and is excellent in broth flavor. &Quot;

The sensory evaluation apparatus 100 can select words based on nouns, adjectives, and verbs as shown in FIG. 3 through morphological analysis based on the taste expression dictionary. At this time, the sensory evaluation apparatus 100 may not select the pronoun designated by the exclusion word because the pronoun according to the person is insignificant in sensory evaluation.

In step 120, the sensory evaluation apparatus 100 derives a vector value for the selected evaluation candidate element word, and then embeds the vector value in the vector space.

For example, the sensory evaluation apparatus 100 may derive the vector values for the selected evaluation candidate element words and then embed them in the vector space. At this time, the vector value for each evaluation candidate element word can be converted into M (M is a natural number) dimension. For example, the vector value for each evaluation candidate element word may be transformed within 10 to 1000 dimensions.

In one embodiment of the present invention, it is assumed that the word-embedding uses a continuous skip-gram model developed by Tomas Mikolov. However, it goes without saying that other known techniques may be applied in addition to this embodiment.

As shown in FIG. 4, the continuous skip-graph model includes three layers such as an input layer, a projection layer, and a hidden layer. The continuous skip-gram model can predict words (words) that can appear in the neighborhood of the current word (word). Using the corpus constructed from the morphological analysis, the same two skipgram models are used to train the skipgram model, and each word can be represented by a 64-dimensional vector.

The sensory evaluation apparatus 100 may change the weights according to detailed comparison items (hereinafter referred to as detailed comparison items) when embedding them in the vector space based on the vector values of the evaluation candidate element words. At this time, the sensory evaluation apparatus 100 may be given the same weight according to the detailed comparison item, or may be applied differently according to each detailed comparison item (FIG. 5).

Further, the sensory evaluation apparatus 100 can exclude words having a meaningless value or a neutral value in the word embedding process for each evaluation candidate element word under detection.

For this purpose, the sensory evaluation apparatus 100 can use a deep learning technique.

For example, the sensory evaluation apparatus 100 learns the word embedding value (vector value) of the evaluation candidate element word having a high frequency in the deep learning model, and then other embedding values are classified and removed through the learned deep learning model can do.

The sensory evaluation apparatus 100 may preprocess review data on-line or on the SNS to calculate the frequency of each word, and then add emotion (emotion) values of the respective words by reflecting the psychological dictionary. In this way, the emotion (emotion) value of each word is included, and a vector value for word embedding can be performed based on the emotion value. The detailed processor for this is shown in Fig.

In step 125, the sensory evaluation apparatus 100 derives the degree of similarity between the characteristic keyword and the evaluation candidate element word, and selects the top n (n is a natural number) evaluation element words having a high degree of similarity.

For example, the sensory evaluation apparatus 100 can derive the degree of similarity between the characteristic keyword and each evaluation candidate element word using the cosine similarity as described in Equation (1) below.

Where A and B represent the vector to be compared, N represents the total number of dimensions, and i represents the current dimension.

Generally, when evaluating food, they are mainly interested in the taste or flavor of food, and users can share experiences through online reviews. Accordingly, in one embodiment of the present invention, when a sensory evaluation is performed on "taste" or "incense" through an online review, a word "taste" or "incense" is selected as a feature keyword, . Thus, the degree of similarity between the characteristic keyword and each word can be derived using each derived vector value, and the top n words having high similarity can be extracted as the evaluation element word.

6 shows an example in which n evaluation element words having a high degree of similarity to feature keywords for product A and product B and having a high frequency of use are finally selected.

FIG. 7 shows an example of visually showing the value of the finally selected evaluation element word by applying a weight to the vector value according to the feature keyword. As shown in FIG. 7, n evaluation element words having a high degree of similarity can be selected differently according to the feature keyword.

In step 130, the sensory evaluation apparatus 100 generates the sensory evaluation results for the product by classifying the selected evaluation element words according to the emotion based on the taste expression dictionary. For example, the sensory evaluation apparatus 100 classifies the evaluation element words selected by applying emotion analysis techniques according to emotion. Let us suppose that the emotion classification is classified as positive and negative. The sensory evaluation apparatus 100 can classify the evaluation element word into a positive word and a negative word according to the emotion analysis technique.

Based on the word analysis used in the word used in the recommendation (affirmative) article of the taste evaluation site and the word analysis used in the dissertation (dissent) article, the positive and negative data in the taste expression dictionary are used to use the word embedding value The positive and negative values of a particular word and its neighboring words are examined and the result can be used to analyze the positive or negative feature.

In this way, the emotional analysis can analyze the people 's evaluation about the unspecified food to the consumer' s feelings, so that it is possible to make more objective and accurate sensory evaluation than what the food has.

8 is a block diagram of a product sensory evaluation device according to an embodiment of the present invention.

8, the sensory evaluation apparatus 100 for a product according to an embodiment of the present invention includes a collecting unit 810, a word extracting unit 815, an analyzing unit 820, a memory 825, 830).

The collecting unit 810 is a means for collecting product review data according to input conditions on-line or in the SNS.

The word extracting unit 815 is means for analyzing the product review data based on the preset taste expression dictionary to select evaluation candidate element words.

The analysis unit 820 derives the vector values for the selected evaluation candidate element words and then embeds the vector values in the vector space and calculates the similarity rolls between the embedded vector values and the feature vector vector values to obtain the top n (natural number) And a means for classifying the evaluation element word according to emotion based on the taste representation dictionary to generate a final sensory evaluation result on the product. Here, the characteristic keyword may vary depending on the element to be compared in detail.

The memory 825 is a means for storing a command, an application, various data derived from the process, and the like necessary for performing a sensory evaluation on a product according to an embodiment of the present invention.

The processor 830 may include internal components (e.g., a collection unit 810, a word extraction unit 815, an analysis unit 820, a memory 810, and a memory 810) of the sensory evaluation apparatus 100 according to an embodiment of the present invention. 825), and the like.

In FIG. 8, it is assumed that each of the functions is included in the sensory evaluation apparatus 100 as a single configuration, and the description is made based on this assumption. The respective functions are the same as those described in detail with reference to FIG. 1, and a duplicate description will be omitted.

In addition, an application capable of automatically performing a food sensory test using a Deep Learning technique after automatically extracting an evaluation target item based on the big data collected online can be implemented. In such a case, the application may be stored in the memory of each computing device, and the respective application may be executed by the processor, which is a component of the computing device, so that the individual steps as described in FIG. 1 may be performed. Since this is the same as that described above with reference to FIG. 1, redundant description will be omitted.

In addition, embodiments of the present invention may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like, alone or in combination. The program instructions recorded on the medium may be those specially designed and constructed for the present invention or may be available to those skilled in the art of computer software. Examples of computer-readable media include magnetic media such as hard disks, floppy disks and magnetic tape; optical media such as CD-ROMs and DVDs; magnetic media such as floppy disks; Examples of program instructions, such as magneto-optical and ROM, RAM, flash memory and the like, can be executed by a computer using an interpreter or the like, as well as machine code, Includes a high-level language code. The hardware devices described above may be configured to operate as one or more software modules to perform operations of one embodiment of the present invention, and vice versa.

As described above, the present invention has been described with reference to particular embodiments, such as specific elements, and limited embodiments and drawings. However, it should be understood that the present invention is not limited to the above- Various modifications and variations may be made thereto by those skilled in the art to which the present invention pertains. Accordingly, the spirit of the present invention should not be construed as being limited to the embodiments described, and all of the equivalents or equivalents of the claims, as well as the following claims, belong to the scope of the present invention .

100: Sensory evaluation device
810:
815: Word extracting unit
820:
825: Memory
830: Processor

Claims (9)

In a sensory evaluation method for a product,
(a) collecting product review data according to an input condition on-line or in an SNS;
(b) analyzing the product review data based on a preset taste expression dictionary to select evaluation candidate element words;
(c) a vector value for the selected evaluation candidate element word is derived and embedded in the vector space, and similarity rolls between the embedded vector value and the feature keyword vector values are calculated, and the top n (natural number) evaluation elements Selecting words; And
(d) classifying the evaluation element word according to emotion based on the taste expression dictionary to generate a final sensory evaluation result for the product.
The method according to claim 1,
The step (d)
Wherein the evaluation element word is classified according to emotion based on positive or negative word data defined in the taste expression dictionary.
The method according to claim 1,
The step (b)
Automatically selecting the rating candidate element word based on the frequency of each word included in the product review data,
Wherein the evaluation candidate element word is a sensory evaluation comparative element of a product that is automatically selected.
The method according to claim 1,
The step (b)
The morphological analysis is performed to extract the evaluation candidate element words using the taste expression dictionary. The product review data is morphologically analyzed on a sentence-by-sentence basis, and a plurality of words having a high frequency among nouns, adjectives, Element words,
And the words corresponding to the survey and personal pronouns are excluded.
The method according to claim 1,
In the step (c), before the evaluation element word is selected,
Learning a deep learning model using vector values of words having a high frequency among the embedded vector values, inputting vector values of the selected evaluation candidate element words into the learned deep learning model, and removing vector values of meaningless words Wherein the sensory evaluation method comprises the steps of:
The method according to claim 1,
Wherein the feature vector vector value is dynamically changed according to the detailed comparison element.
A computer readable recording medium on which program codes necessary for carrying out the method according to any one of claims 1 to 6 are recorded.
A collection unit for collecting product review data according to input conditions on-line or in a SNS;
A word extracting unit for analyzing the product review data based on a preset taste expression dictionary to select evaluation candidate element words;
The vector values for the selected evaluation candidate element words are derived and embedded in the vector space. The similarity degree rolls between the embedded vector values and the feature keyword vector values are calculated, and the top n (natural number) evaluation element words having high similarity are selected And an analyzer for classifying the evaluation element word according to an emotion based on the taste representation dictionary to generate a final sensory evaluation result on the product.
A computing device comprising:
A memory for storing at least one application;
A processor coupled to the memory and executing an application stored in the memory,
The application executed by the processor,
(a) collecting product review data according to an input condition on-line or in an SNS;
(b) analyzing the product review data based on a preset taste expression dictionary to select evaluation candidate element words;
(c) a vector value for the selected evaluation candidate element word is derived and embedded in the vector space, and similarity rolls between the embedded vector value and the feature keyword vector values are calculated, and the top n (natural number) evaluation elements Selecting words; And
(d) classifying the evaluation element word according to an emotion based on the taste expression dictionary to generate a final sensory evaluation result for the product.

Images