Shubhangi Tandon

A multimodal embedding modifier generates a modified seed search selection embedding for providing a set of search results. The multimodal embedding modifier enhances the ability and accuracy of identifying a user's true intent when searching the online marketplace. For example, embodiments disclosed herein can allow a user to navigate multiple modalities for an item. In some embodiments, a user may select a search result corresponding to an initial search query, and further modify the selected search result by inputting a modifier (e.g., a textual modifier). The multimodal embedding modifier can be trained using a training dataset including a text embedding, an image embedding, another type of embedding, or a combination thereof.

A search engine provides diverse search results in response to a search query using a trained diversity ranker. The diverse results may be generated by determining ranking positions of at least some search results based on delta feature values that are dependent on higher-ranked results. Delta feature values indicate a degree or magnitude of difference between a particular result and other results with respect to a particular feature, such as price, category, or shipping type. In providing the ranked results for presentation to a user at a computing device, a diversity feature indicator for at least some results may be generated. The diversity feature is a feature that distinguishes a given result from other results. As such, the diversity feature indicator represents that diversity feature for the particular result being different from other results and distinguishes the diversity feature from other features in the particular result.