Systems and methods for intelligent extraction of attributes from product titles
Some aspects of the present disclosure are directed to computerized methods for extracting attributes from product titles. The method may include: retrieving a title associated with a product listing and historical product title data; refining the title; determining at least one tag associated with an attribute; generating, based on the at least one extracted tag and the historical title data, a first combination of one or more attributes; determining whether the title includes at least one plurality of product options, and if so: determining, for each product option hi the plurality of product options, a second combination of one or more attributes by removing attributes associated with alternative product options from the first combination; and generating a product identifier based on the second combination; and if the title does not include at least one plurality of product options, generating, the product identifier based on the first combination.
1. A computer-implemented system for extracting attributes from product titles, the system comprising:
a memory storing instructions; and
at least one processor configured to execute the instructions to perform operations comprising:
retrieving a title associated with a product listing and historical product title data;
preprocessing the title into identifiable attributes;
extracting at least one tag;
generating, using the at least one tag and a machine-learning algorithm that relies on the historical title data to determine an optimal combination of attributes, a first combination of one or more attributes;
determining whether the product listing includes at least one plurality of product options, and if so:
determining, for each product option in the plurality of product options, a second combination of one or more attributes; and
generating a plurality of product identifiers, wherein each determined second combination is associated with one of the plurality of product identifiers; and
if the title does not include at least one plurality of product options, generating a product identifier associated with the first combination.
2. The system of claim 1 , wherein extracting the at least one tag comprises refining the title to convert it to a standard format.
3. The system of claim 1 , wherein extracting at least one tag comprises extracting multiple tags from at least one single origin word.
4. The system of claim 1 , wherein extracting the at least one tag comprises converting at least one origin word in the title to a representative word associated with the at least one tag.
5. The system of claim 1 , wherein extracting the at least one tag comprises recognizing at least one pattern and extracting the at least one tag based on the at least one pattern.
6. The system of claim 5 , wherein extracting the at least one tag based on the at least one pattern comprises converting a unit value in the title into a base unit.
7. The system of claim 1 , wherein extracting the at least one tag comprises:
consulting a dictionary stored in the at least one data structure; and
extracting at least one word from the title, the at least one word being registered in the dictionary and associated with the at least one tag.
8. The system of claim 1 , wherein determining the second combination comprises removing attributes associated with alternative product options from the first combination to arrive at the second combination.
9. The system of claim 1 , wherein the machine-learning algorithm is at least partially informed based on a sequence of tags included in the title.
10. The system of claim 1 , wherein the machine-learning algorithm is a Viterbi algorithm with a probabilistic table.
11. A computer-implemented method for extracting attributes from product titles, the method comprising:
storing instructions in a memory; and
performing operations, by at least one processor configured to execute the instructions, comprising:
retrieving a title associated with a product listing and historical product title data;
preprocessing the title into identifiable attributes;
extracting at least one tag;
generating, using the at least one tag and a machine-learning algorithm that relies on the historical title data to determine an optimal combination of attributes, a first combination of one or more attributes;
determining whether the product listing includes at least one plurality of product options, and if so:
determining, for each product option in the plurality of product options, a second combination of one or more attributes; and
generating a plurality of product identifiers, wherein each determined second combination is associated with one of the plurality of product identifiers; and
if the title does not include at least one plurality of product options, generating a product identifier associated with the first combination.
12. The method of claim 11 , wherein extracting the at least one tag comprises refining the title to convert it to a standard format.
13. The method of claim 11 , wherein extracting the at least one tag comprises extracting multiple tags from at least one single origin word.
14. The method of claim 11 , wherein extracting the at least one tag comprises converting at least one origin word in the title to a representative word associated with the at least one tag.
15. The method of claim 11 , wherein extracting the at least one tag comprises recognizing at least one pattern and extracting the at least one tag based on the at least one pattern.
16. The method of claim 15 , wherein extracting the at least one tag based on the at least one pattern comprises converting a unit value in the title into a base unit.
17. The method of claim 11 , wherein extracting the at least one tag comprises:
consulting a dictionary stored in the at least one data structure; and
extracting at least one word from the title, the at least one word being registered in the dictionary and associated with the at least one tag.
18. The method of claim 11 , wherein determining the second combination comprises removing attributes associated with alternative product options from the first combination to arrive at the second combination.
19. The method of claim 11 , wherein the machine-learning algorithm is at least partially informed based on a sequence of tags included in the title.
20. The method of claim 11 , wherein the machine-learning algorithm is a Viterbi algorithm with a probabilistic table.