Development of a prototype of the mechanism for searching for goods on the selection portal
As part of the hackathon, the task to be implemented is: Analyze a typical functional product search system (search in the course of a set of names - hints, contextual synonyms, transliteration, oki, ranking of hints and issuance, search taking into account properties, accounting for word permutation, search history, search for goods and goods ). etc.); Implement the system; Test the system, compiling its work within the framework of the implementation of metrics; About the round performance of the project; Identify opportunities for scaling the solution and propose improvements.
- server.py - Flask backend, connects ML to the front
- search.py - building information search
- nlp.py - nlp processing of datasets and outgoing files
- main.py - ml part with Bert language model
- autocorrect.py - autocorrect source files
- autocomplete.py - autocomplete files at runtime
- template and statics - front part
- bert - language model training