Method and apparatus for determining item name, computer device, and storage medium
A method includes: obtaining a first image including a target item; selecting a plurality of reference images corresponding to the first image from a database; performing word segmentation on item text information corresponding to the plurality of reference images to obtain a plurality of words; and extracting a key word meeting a reference condition from the plurality of words, and determining the extracted key word as an item name of the target item.
1. A method for determining an item name of a target item, applied to a computing device, the method comprising:
obtaining a first image including the target item;
selecting a plurality of reference images according to the first image from a database, the database including a plurality of images and item information corresponding to the plurality of images, the item information corresponding to the images being used for describing items included in the images, and the item information corresponding to the images including item text information;
performing word segmentation on the item text information corresponding to the plurality of reference images to obtain a plurality of words; and
extracting a key word meeting a reference condition from the plurality of words, and determining the extracted key word as the item name of the target item,
wherein performing the word segmentation comprises:
performing clustering processing on the plurality of reference images to obtain a plurality of image stacks, each image stack including at least two reference images;
obtaining a similarity between each image stack and the first image according to similarities between reference images in each image stack and the first image;
selecting an image stack having a highest similarity with the first image from the plurality of image stacks, and determining the selected image stack as a target image stack; and
performing word segmentation on item text information corresponding to the plurality of reference images in the target image stack to obtain a plurality of words.
2. The method according to claim 1 , wherein extracting the key word comprises:
determining an average value of word vectors of the plurality of words as a center vector; and
determining a distance between each word vector in the word vectors and the center vector, and determining a word corresponding to a word vector having a smallest distance as the key word meeting the reference condition.
3. The method according to claim 1 , wherein performing the word segmentation further comprises:
performing word segmentation on the item text information corresponding to the plurality of reference images according to a plurality of reference lengths, to obtain a word whose length is equal to each reference length separately.
4. The method according to claim 1 , wherein extracting the key word comprises:
determining appearance frequency of the plurality of words in the item text information corresponding to the plurality of reference images; and
selecting, from the plurality of words, a word having a largest length and whose appearance frequency is higher than a first threshold, and determining the selected word as the key word meeting the reference condition; or
selecting, from the plurality of words, a word having the highest appearance frequency, and determining the selected word as the key word meeting the reference condition.
5. The method according to claim 1 , wherein obtaining the first image comprises:
obtaining an original image including the target item;
performing item detection on the original image to determine a region in which the target item is located in the original image; and
extracting an image of the region from the original image to obtain the first image.
6. The method according to claim 1 , wherein the database includes images belonging to a plurality of categories; and selecting the plurality of reference images comprises:
determining a target category to which the first image belongs; and
selecting a plurality of reference images belonging to the target category and according to the first image from the database.
7. The method according to claim 6 , wherein determining the target category comprises:
obtaining a similarity between each category in the plurality of categories and the first image separately; and
determining a category having a highest similarity with the first image in the plurality of categories as the target category.
8. The method according to claim 7 , wherein the database includes a plurality of sub-databases, different sub-databases correspond to different categories, and each sub-database includes at least one image belonging to a corresponding category and item information corresponding to the at least one image; and
obtaining the similarity comprises:
for each sub-database in the plurality of sub-databases, performing:
obtaining a similarity between each image in the sub-database and the first image;
selecting a plurality of second images from the sub-database according to the similarity between each image and the first image, a similarity between the second image and the first image being higher than a similarity between another image in the sub-database and the first image; and
determining an average similarity corresponding to the plurality of second images, and determining the average similarity as a similarity between a category corresponding to the sub-database and the first image.
9. The method according to claim 1 , wherein performing the clustering processing comprises:
establishing an association relationship between any two reference images in response to that a similarity between any two reference images in the plurality of reference images is higher than a second threshold; and
forming reference images having an association relationship in the plurality of reference images into an image stack, to obtain the plurality of image stacks.
10. The method according to claim 1 , further comprising:
performing searching according to the item name, to obtain item information corresponding to the item name; and
displaying the item information in a current display interface.
11. The method according to claim 10 , further comprising:
selecting a reference image having a highest quality score from the plurality of reference images, and using the selected reference image as a presentation image of the target item; and
displaying the presentation image in the display interface.
12. An apparatus for determining an item name of a target item, the apparatus comprising: a memory storing computer program instructions; and a processor coupled to the memory and configured to execute the computer program instructions and perform:
obtaining a first image including the target item;
selecting a plurality of reference images according to the first image from a database, the database including a plurality of images and item information corresponding to the plurality of images, the item information corresponding to the images being used for describing items included in the images, and the item information corresponding to the images including item text information;
performing word segmentation on the item text information corresponding to the plurality of reference images to obtain a plurality of words; and
extracting a key word meeting a reference condition from the plurality of words, and determining the extracted key word as the item name of the target item,
wherein performing the word segmentation comprises:
performing clustering processing on the plurality of reference images to obtain a plurality of image stacks, each image stack including at least two reference images;
obtaining a similarity between each image stack and the first image according to similarities between reference images in each image stack and the first image;
selecting an image stack having a highest similarity with the first image from the plurality of image stacks, and determining the selected image stack as a target image stack; and
performing word segmentation on item text information corresponding to the plurality of reference images in the target image stack to obtain a plurality of words.
13. The apparatus of claim 12 , wherein extracting the key word includes:
determining an average value of word vectors of the plurality of words as a center vector; and
determining a distance between each word vector in the word vectors and the center vector, and determining a word corresponding to a word vector having a smallest distance as the key word meeting the reference condition.
14. The apparatus of claim 12 , wherein performing the word segmentation further includes:
performing word segmentation on the item text information corresponding to the plurality of reference images according to a plurality of reference lengths, to obtain a word whose length is equal to each reference length separately.
15. The apparatus of claim 12 , wherein extracting the key word includes:
determining appearance frequency of the plurality of words in the item text information corresponding to the plurality of reference images; and
selecting, from the plurality of words, a word having a largest length and whose appearance frequency is higher than a first threshold, and determining the selected word as the key word meeting the reference condition; or
selecting, from the plurality of words, a word having the highest appearance frequency, and determining the selected word as the key word meeting the reference condition.
16. The apparatus of claim 12 , wherein obtaining the first image includes:
obtaining an original image including the target item;
performing item detection on the original image to determine a region in which the target item is located in the original image; and
extracting an image of the region from the original image to obtain the first image.
17. The apparatus of claim 12 , wherein the database includes images belonging to a plurality of categories; and selecting the plurality of reference images includes:
determining a target category to which the first image belongs; and
selecting a plurality of reference images belonging to the target category and according to the first image from the database.
18. The apparatus of claim 12 , wherein determining the target category includes:
obtaining a similarity between each category in the plurality of categories and the first image separately; and
determining a category having a highest similarity with the first image in the plurality of categories as the target category.
19. A non-transitory computer-readable storage medium storing computer program instructions executable by at least one processor to perform:
obtaining a first image including the target item;
selecting a plurality of reference images according to the first image from a database, the database including a plurality of images and item information corresponding to the plurality of images, the item information corresponding to the images being used for describing items included in the images, and the item information corresponding to the images including item text information;
performing word segmentation on the item text information corresponding to the plurality of reference images to obtain a plurality of words; and
extracting a key word meeting a reference condition from the plurality of words, and determining the extracted key word as the item name of the target item,
wherein performing the word segmentation comprises:
performing clustering processing on the plurality of reference images to obtain a plurality of image stacks, each image stack including at least two reference images;
obtaining a similarity between each image stack and the first image according to similarities between reference images in each image stack and the first image;
selecting an image stack having a highest similarity with the first image from the plurality of image stacks, and determining the selected image stack as a target image stack; and
performing word segmentation on item text information corresponding to the plurality of reference images in the target image stack to obtain a plurality of words.