Systems and methods for providing machine-learning enabled user-specific evaluations are disclosed. Implementations include obtaining a first set of evaluation data from a user interface, obtaining a first set of target-descriptive data including target-specific characteristics objectively describing the evaluation targets, and training, with a machine-learning algorithm, a user-specific evaluation profile indicating evaluation patterns relative to the first set of evaluation data and the first set of target-specific characteristics. Implementations include applying the user-specific evaluation profile to a second set of target-descriptive data to predict a user-specific evaluation.
BACKGROUND (1) Product and service review tools, such as YELP, ROTTEN TOMATOES, AMAZON, TRIP ADVISOR, NETFLIX, and other similar review systems, provide users with aggregated review information. For example, the review system may provide an average number of “stars,” points, happy faces, tomatoes, or other relative ranking indicator for all reviews submitted. The same evaluation scores and descriptions may be provided to users interested in the particular good or service being reviewed, regardless of the users' individualized preferences and tastes. For example, a user who is particular about certain types of fried chicken may perceive and review a restaurant that serves chicken differently than a user who does not like chicken, or is ambiguous to chicken. In another example involving relative price rating systems, one user might think $5 for a meal is expensive, while another user might think $5 for a meal is cheap. While some ranking systems do attempt to learn a user's overall preference and recommend products or services (such as NETFLIX), those ranking systems are generally binary or numeric (e.g., thumbs up or thumbs down, or rank on a scale of 1 to 5). BRIEF SUMMARY OF EMBODIMENTS (2) According to various embodiments of the disclosed technology, systems and methods for enhancing evaluations and reviews with machine learning are described. In particular, the disclosed technology uses a machine learning algorithm to train a user-specific evaluation profile based on the u
1. A computer implemented method of providing evaluations, the method comprising: obtaining a first set of evaluation data indicating qualitative descriptive feedback from a first user regarding one or more initial evaluation targets; obtaining a first set of target-descriptive data comprising, for each initial evaluation target, one or more target-specific characteristics objectively describing the initial evaluation target; training, with a user evaluation profiling logical circuit, a user-specific evaluation profile model indicating evaluation patterns relative to the first set of evaluation data and the one or more target-specific characteristics; predicting, using the user-specific evaluation profile model, a user review for a review target from a review category; receiving a second set of evaluation data for a second evaluation target, the second set of evaluation data corresponding to one or more users different from the first user; filtering the second set of evaluation data based at least in part on the predicted user review to generate a filtered set of second evaluation data, wherein the filtering comprises: determining a first probability that the one or more users different from the first user will prefer the review target as compared with other possible review targets from the review category; determining a second probability that the one or more users different from the first user will prefer the other possible review targets from the review category; and filtering a list of the other possible review targets based on the second probability that the one or more users different from the first user will prefer the other possible review targets in the list; and displaying information based at least in part on the filtered set of second evaluation data. ||
10. The method claim 1, wherein the one or more target-specific characteristics objectively describing the initial evaluation target comprise a general review category, a target location, a target proximity to the first user, a medium of discovery of the initial evaluation target, or a combination thereof. ||
11. An evaluation system comprising: a user interface, a data store, and a user evaluation profiling logical circuit, the user evaluation prediction logical circuit comprising a processor and a non-transitory medium with computer executable instructions embedded thereon, the computer executable instructions configured to: obtain a first set of evaluation data indicating qualitative descriptive feedback from a first user regarding one or more initial evaluation targets; obtain a first set of target-descriptive data comprising, for each initial evaluation target, one or more target-specific characteristics objectively describing the initial evaluation target; train, with the user evaluation profiling logical circuit, a user-specific evaluation profile model indicating evaluation patterns relative to the first set of evaluation data and one or more target-specific characteristics; predict, using the user-specific evaluation profile model, a user review for a review target from a review category; receive a second set of evaluation data for a second evaluation target, the second set of evaluation data corresponding to one or more users different from the first user; filter the second set of evaluation data based at least in part on the predicted user review to generate a filtered set of second evaluation data, wherein the computer executable instructions performing the filtering are configured to: determine a first probability that the one or more users different from the first user will prefer the review target as compared with other possible review targets from the review category; determine a second probability that the one or more users different from the first user will prefer the other possible review targets from the review category; and filter a list of the other possible review targets based on the second probability that the one or more users different from the first user will prefer the other possible review targets in the list; and display information based at least in part on the filtered set of second evaluation data.