Client's motivation for launching the project:
The need to create a unified jewelry database revealed limitations in the current process of collating product data from different suppliers, which created difficulties for business growth.
Description of the initial situation:
Project goals:
creation of a model for comparing reference products and products from suppliers with a high level of reliability and resistance to paraphrases.
MIL Team solution:
the use of the team’s existing solutions in the field of natural language processing, as well as pre-trained neural network embeddings, allowed the customer to obtain a prototype system for matching products with the required level of quality.
The need to create a unified jewelry database revealed limitations in the current process of collating product data from different suppliers, which created difficulties for business growth.
Description of the initial situation:
- creating a unified and complete database of jewelry involves comparing various databases from different suppliers in a single, structured format;
- Bringing all fields and ways of writing them to a single format is a labor-intensive task due to differences in the possible spellings of structured fields.
- high accuracy of comparison of reference products and products from suppliers is ensured using manual marking;
- manual marking does not allow you to scale your business and expand the reference database of jewelry.
Project goals:
creation of a model for comparing reference products and products from suppliers with a high level of reliability and resistance to paraphrases.
MIL Team solution:
the use of the team’s existing solutions in the field of natural language processing, as well as pre-trained neural network embeddings, allowed the customer to obtain a prototype system for matching products with the required level of quality.