Massive rule-based classification engine
View Patent ↗Systems and methods are disclosed herein for performing classification of documents or performing other tasks based on rules. A rule generator receives a request for a rule that will receive as an input a document and output an outcome such as a classification of the document, addition of the document to a whitelist or blacklist, or occurrence of some other outcome. The rules are applied to a document and the document and outcome of the rules are presented to a rater. A rating of the accuracy of the outcome is received from the rater and the rating is propagated to quality metrics of rules that contributed to the outcome. Rules with a quality metric above a threshold may be added to a production rule set. Rules with a quality metric below a threshold may be removed.
1. A method for rule generation, the method comprising:
receiving, by a computer system, a plurality of classification rules;
for a set of documents in a corpus, performing, by the computer system, for each document in the set, performing repeatedly:
applying a plurality of applicable rules from the plurality of classification rules applicable to the document by a) applying an applicable rule of the plurality of applicable rules to the document to obtain a rule outcome, b) selecting another applicable rule of the plurality of applicable rules according to the rule outcome, c) repeating a) and b) one or more times until a final rule outcome is reached;
presenting the final rule outcome and the document to a rater;
receiving a rating of the final rule outcome; and
updating, for one or more of the plurality of applicable rules, quality metrics corresponding to the one or more of the plurality of applicable rules in accordance with the rating, wherein updating, for the one or more of the plurality of applicable rules, the quality metrics comprises, for each applicable rule of the one or more of the plurality of applicable rules, adjusting the quality metric corresponding to the each applicable rule by an amount that decreases with a number of intervening rules from one of the plurality of applicable rules that produced the final rule outcome;
selecting a first portion of the plurality of applicable rules according to the quality metrics;
adding the first portion of the plurality of applicable rules to a production rule set; and
performing, by the computer system, production document classification in accordance with the production rule set.
2. The method of claim 1 , wherein receiving, by the computer system, the plurality of classification rules further comprises:
selecting an element of the corpus;
transmitting a rule request with the element to a rule generator;
receiving a rule from the rule generator; and
adding the rule to the plurality of classification rules.
3. The method of claim 2 , wherein the rule request includes at least one category to which the element belongs.
4. The method of claim 2 , wherein the element is a web domain.
5. The method of claim 2 , wherein the element is a user identifier.
6. The method of claim 2 , wherein an output of the rule is a category.
7. The method of claim 2 , wherein an output of the rule is a spam flag.
8. The method of claim 2 , wherein an output of the rule is an entity assignment.
9. The method of claim 1 , further comprising removing, by the computer system, at least one rule of the plurality of classification rules from the plurality of classification rules in accordance with a quality metric of the at least one rule.
10. The method of claim 9 , further comprising:
transmitting, by the computer system, the at least one rule to one or more rule generators that generated the at least one rule; and
transmitting, by the computer system, one or more documents inaccurately classified according to the at least one rule to the one or more rule generators that generated the at least one rule.
11. A method for rule generation, the method comprising:
receiving, by a computer system, a plurality of classification rules;
for a set of documents in a corpus, performing, by the computer system, for each document in the set, performing repeatedly:
applying a plurality of applicable rules from the plurality of classification rules applicable to the document;
presenting a final rule outcome and the document to a rater;
receiving a rating of the final rule outcome; and
updating, for one or more of the plurality of applicable rules, quality metrics corresponding to the one or more of the plurality of applicable rules in accordance with the rating, wherein updating, for the one or more of the plurality of applicable rules, the quality metrics comprises, for each applicable rule of the one or more of the plurality of applicable rules, adjusting the quality metric corresponding to the each applicable rule by an amount that decreases with a number of intervening rules from one of the plurality of applicable rules that produced the final rule outcome;
selecting a first portion of the plurality of applicable rules according to the quality metrics;
adding the first portion of the plurality of applicable rules to a production rule set; and
performing, by the computer system, production document classification in accordance with the production rule set.
12. The method of claim 11 , wherein receiving, by the computer system, the plurality of classification rules further comprises:
selecting an element of the corpus;
transmitting a rule request with the element to a rule generator;
receiving a rule from the rule generator; and
adding the rule to the plurality of classification rules.
13. The method of claim 12 , wherein the rule request includes at least one category to which the element belongs.
14. The method of claim 12 , wherein the element is a web domain.
15. The method of claim 12 , wherein the element is a user identifier.
16. The method of claim 12 , wherein an output of the rule is a category.
17. The method of claim 12 , wherein an output of the rule is a spam flag.
18. The method of claim 12 , wherein an output of the rule is an entity assignment.
19. The method of claim 11 , wherein selecting the first portion of the plurality of applicable rules according to the quality metrics comprises:
comparing the quality metrics of the plurality of applicable rules to a rule threshold;
determining that the first portion of the plurality of applicable rules have quality metrics above the threshold;
determining that a second portion of the plurality of applicable rules have quality metrics below the threshold; and
discarding the second portion of the plurality of applicable rules.
20. The method of claim 19 , further comprising:
transmitting, by the computer system, the second portion of the plurality of applicable rules to one or more rule generators that generated the second portion of the plurality of applicable rules; and
transmitting, by the computer system, one or more documents inaccurately classified according to the second portion of the plurality of applicable rules to the one or more rule generators that generated the second portion of the plurality of applicable rules.