We developed natural language processing (NLP) and machine learning (ML) solutions that presented customers with better image search results.
- ProCogia analyzed user queries to identify the main categories of failed searches, which were: Ambiguous search intent, Incorrect parsing of search queries, Inaccurate or missing metadata.
- We designed and implemented the search parsing algorithm – matching search queries to concepts in the search ontology – that eliminated or mitigated several major causes of suboptimal results.
- We created and analyzed a representative dataset of suboptimal searches and used this to identify and address systematic sources of search failure in image metadata, such as incorrect lemmatization.
- ProCogia’s engagement continues with the client to improve searching by deploying foundational models and multimodal embeddings to capture highly relevant contextual information between the query and the returned images. This ensures irrelevant images are not presented to the user.
- Through rigorous hypothesis testing we have helped our client identify the underlying causes of the search challenges.
- We developed and deployed a range of linguistics and AI solutions to improve query understanding, and improve the searching and browsing experience for users, while driving key conversion metrics (customer engagement and asset purchase).
- Our NLP machine learning model uses cosine similarity to identify best matches between search queries and images in the database, greatly improving the accuracy of ambiguous queries.
- ProCogia’s solution allows our client’s customers to see a higher number of relevant search results (improved recall) and reduces incorrect results (greater precision).