# US 2019/0318263 A1 · Classifying text to determine a goal type used to select machine learning algorithm outcomes > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2019/0318263 A1 - Document ID: US20190318263A1 - Status: Published application - Kind: A1 - Application number: 16/453,221 - Filed: 2019-06-26 - Published: 2019-10-17 - Earliest claimed: 2018-04-13 - Assignees: International Business Machines Corporation - Applicants: International Business Machines Corporation - Inventors: Aysu Ezen Can; David Contreras; Bob Delima; Corville O. Allen - Web page: https://patentgrep.com/patent/US20190318263A1 ## CPC - G06F 40/20: Natural language analysis (semantic analysis of natural language G06F40/30) - G06F 16/3344: {using natural language analysis} - G06F 16/353: into predefined classes - G06F 40/205: Parsing - G06N 20/00: Machine learning - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract Provided are a computer program product, system, and method for classifying text to determine a goal type used to select machine learning algorithm outcomes. Natural language processing of text is performed to determine features in the text and their relationships. A classifier classifies the text based on the relationships and features to determine a goal type. The determined features and relationships from the text are inputted into a plurality of different machine learning algorithms to generate outcomes. For each of the machine learning algorithms, a determination is made of performance measurements resulting from the machine learning algorithms generating the outcomes. A determination is made of at least one machine learning algorithm having performance measurements that are highly correlated to the determined goal type. An outcome is determined from at least one of the outcomes. ## Claims ### Claim at position 1 1-16. (canceled) ### Claim 17 A method implemented in a computer system for determining an outcome for text from machine learning algorithms, comprising: performing natural language processing of text to determine features in the text and their relationships; classifying, with a classifier, the text based on the relationships and features of the text to determine a goal type of a plurality of goal types; inputting the determined features and relationships from the text into a plurality of different machine learning algorithms to generate outcomes for the text; for each of the machine learning algorithms, determining performance measurements resulting from the machine learning algorithms; determining at least one machine learning algorithm of the machine learning algorithms having performance measurements that are highly correlated to the determined goal type; and determining an outcome from at least one of the outcomes of the determined at least one machine learning algorithm. ### Claim 18 (depends on 17) The method of claim 17, wherein the plurality of goal types include at least multiple of accuracy, precision, recall and execution time, and wherein the determined at least one machine learning algorithm comprises the at least one machine learning algorithm having performance measurements most highly correlated to the goal type. ### Claim 19 (depends on 18) The method of claim 18, wherein the determining the goal type comprises scoring each of the goal types by performing natural language processing of the text to determine features in the text , wherein the determined goal type comprises the goal type having a highest score of the goal types. ### Claim 20 (depends on 19) The method of claim 19, wherein the scoring the goal types comprises performing at least one determination that is a member of a set of determinations comprising: determining whether the text includes a feature providing a high degree of specificity in an ontology depth, wherein the precision goal type is scored relatively higher to other of the goal types in response to determining that the feature provides a high degree of specificity in the ontology depth for the feature; determining whether the text includes a feature having a high degree of generality in an ontology depth, wherein the recall goal type is scored relatively higher to other of the goal types in response to determining that the feature provides a high degree of generality in the ontology depth for the feature; determining whether the text includes features comprising a comparison or a feature having a high weighting, wherein the precision goal type is scored relatively higher to other of the goal types in response to determining that the feature includes a comparison or and the accuracy goal type is scored relatively higher to other of the goal types in response to determining that the feature has a high weighting; determining whether the text has a feature comprising a numerical quantity that relates to a size of a related feature, including a medical condition, wherein the accuracy goal type is scored relatively higher than other of the goal types in response to determining that the feature comprises a numerical quantity relating to a size of a medical condition; and determining whether the text has features indicating a deep search or a specific answer, wherein the execution time goal type is scored higher in response to determining that the features indicate a specific answer, and wherein the execution time goal type is scored lower in response to determining that the features indicate a deep search. ### Claim 21 (depends on 18) The method of claim 18, further comprising: in response to determining the goal type, processing the determined features in the text to determine features unrelated to the determined goal type; and removing the determined features in the text unrelated to the goal type, wherein the inputting the determined features comprises inputting the determined features not removed from the text. ### Claim 22 (depends on 17) The method of claim 17, wherein the text comprises a question and wherein the outcomes comprise answers to the question. ### Claim 23 (depends on 19) The method of claim 19, wherein the scoring the goal types comprises: determining whether the text includes a feature providing a high degree of specificity in an ontology depth, wherein the precision goal type is scored relatively higher to other of the goal types in response to determining that the feature provides a high degree of specificity in the ontology depth for the feature. ### Claim 24 (depends on 19) The method of claim 19, wherein the scoring the goal types comprises: determining whether the text includes a feature having a high degree of generality in an ontology depth, wherein the recall goal type is scored relatively higher to other of the goal types in response to determining that the feature provides a high degree of generality in the ontology depth for the feature. ### Claim 25 (depends on 19) The method of claim 19, wherein the scoring the goal types comprises: determining whether the text includes features comprising a comparison or a feature having a high weighting, wherein the precision goal type is scored relatively higher to other of the goal types in response to determining that the feature includes a comparison or and the accuracy goal type is scored relatively higher to other of the goal types in response to determining that the feature has a high weighting. ### Claim 26 (depends on 19) The method of claim 19, wherein the scoring the goal types comprises: determining whether the text has a feature comprising a numerical quantity that relates to a size of a related feature, including a medical condition, wherein the accuracy goal type is scored relatively higher than other of the goal types in response to determining that the feature comprises a numerical quantity relating to a size of a medical condition. ### Claim 27 (depends on 19) The method of claim 19, wherein the scoring the goal types comprises: determining whether the text has features indicating a deep search or a specific answer, wherein the execution time goal type is scored higher in response to determining that the features indicate a specific answer, and wherein the execution time goal type is scored lower in response to determining that the features indicate a deep search. ### Claim 28 (depends on 17) The method of claim 17, wherein the operations further comprise: combining outcomes from a plurality of machine learning algorithms producing results satisfying the determined goal type, and selecting from the combined outcomes using at least one of scoring the results, majority voting or boosting.