Discover unidirectional associations among terms or documents

Patent metadata

Field	Value
Publication number	US-11526806-B2
Application number	US-202016784954-A
Country	US
Kind code	B2
Filing date	Feb 7, 2020
Priority date	Feb 7, 2020
Publication date	Dec 13, 2022
Grant date	Dec 13, 2022

Abstract

An approach is provided in which the approach trains a machine learning model using reference entries included in a reference dataset. During the training, the machine learning model learns a first set of unidirectional associations between the reference entries. The approach inputs a user dataset into the trained machine learning model and generates a second set of unidirectional associations between user dataset entries included in the user dataset. The approach builds a hierarchical relationship of the user dataset based on the second set of unidirectional associations and manages the user dataset based on the hierarchical relationship.

First claim

The invention claimed is:

1. A method implemented by an information handling system that includes a memory and a processor, the method comprising:
training a machine learning model using a reference dataset comprising a plurality of reference entries, wherein the machine learning model learns a first set of unidirectional associations between the plurality of reference entries;
preparing a user dataset for input into the trained machine learning model, wherein the user dataset comprises a plurality of user dataset entries, wherein the preparing comprises:
creating a first one of the plurality of user dataset entries that assigns a first term as a parent to a second term;
creating a second one of the plurality of user dataset entries that assigns the second term as a parent to the first term; and
applying a set of lexical relations feature learning algorithms to the first user dataset entry and the second user dataset entry to generate a first set of user dataset feature learning results, wherein the first set of user dataset feature learning results includes a feature value result for each of the user dataset entries;

generating a second set of unidirectional associations between the plurality of user dataset entries included in the user dataset in response to inputting the prepared user dataset into the trained machine learning model, wherein the second set of unidirectional associations are based on the feature value results;
building a hierarchical relationship of the user dataset based on the second set of unidirectional associations; and
managing the user dataset based on the hierarchical relationship.

2. The method of claim 1 further comprising:
inputting the first set of user dataset feature learning results into the trained machine learning model to generate a portion of the second set of unidirectional associations.

3. The method of claim 1 wherein at least one of the set of lexical relations feature learning algorithms is selected from a group consisting of a hypernym feature learning algorithm, a hyponym feature learning algorithm, a holonym feature learning algorithm, and a meronym feature learning algorithm.

4. The method of claim 1 wherein the user dataset comprises a first description corresponding to the first user term and comprises a second description corresponding to the second user term, the method further comprising:
combining the first term and the first description into a first document;
combining the second term and the second description into a second document;
creating a third one of the plurality of user dataset entries that assigns the first document as a parent to the second document;
creating a fourth one of the plurality of user dataset entries that assigns the second document as a parent to the first document;
applying the set of lexical relations feature learning to the third data entry and the fourth data entry to generate a second set of user dataset feature learning results; and
inputting the second set of user dataset feature learning results into the trained machine learning model to generate a portion of the second set of unidirectional associations.

5. The method of claim 1 wherein the reference dataset comprises a subset of related entries and a subset of unrelated entries, the method further comprising:
removing the subset of unrelated entries from the reference dataset to create a prepared reference dataset;
applying a set of lexical relations feature learning algorithms to the prepared reference dataset to generate a set of reference dataset feature learning results; and
performing the training of the machine learning model using the set of reference dataset feature learning results.

6. The method of claim 1 wherein the user dataset is devoid of classification information and is also devoid of data association information prior to the generating of the second set of unidirectional associations.

7. An information handling system comprising:
one or more processors;
a memory coupled to at least one of the processors;
a set of computer program instructions stored in the memory and executed by at least one of the processors in order to perform actions of:
training a machine learning model using a reference dataset comprising a plurality of reference entries, wherein the machine learning model learns a first set of unidirectional associations between the plurality of reference entries;
preparing a user dataset for input into the trained machine learning model, wherein the user dataset comprises a plurality of user dataset entries, wherein the preparing comprises:
creating a first one of the plurality of user dataset entries that assigns a first term as a parent to a second term;
creating a second one of the plurality of user dataset entries that assigns the second term as a parent to the first term; and
applying a set of lexical relations feature learning algorithms to the first user dataset entry and the second user dataset entry to generate a first set of user dataset feature learning results, wherein the first set of user dataset feature learning results includes a feature value result for each of the user dataset entries;

generating a second set of unidirectional associations between the plurality of user dataset entries included in the user dataset in response to inputting the prepared user dataset into the trained machine learning model, wherein the second set of unidirectional associations are based on the feature value results;
building a hierarchical relationship of the user dataset based on the second set of unidirectional associations; and
managing the user dataset based on the hierarchical relationship.

8. The information handling system of claim 7 wherein the processors perform additional actions comprising:
inputting the first set of user dataset feature learning results into the trained machine learning model to generate a portion of the second set of unidirectional associations.

9. The information handling system of claim 7 wherein at least one of the set of lexical relations feature learning algorithms is selected from a group consisting of a hypernym feature learning algorithm, a hyponym feature learning algorithm, a holonym feature learning algorithm, and a meronym feature learning algorithm.

10. The information handling system of claim 7 wherein the user dataset comprises a first description corresponding to the first user term and comprises a second description corresponding to the second user term, and wherein the processors perform additional actions comprising:
combining the first term and the first description into a first document;
combining the second term and the second description into a second document;
creating a third one of the plurality of user dataset entries that assigns the first document as a parent to the second document;
creating a fourth one of the plurality of user dataset entries that assigns the second document as a parent to the first document;
applying the set of lexical relations feature learning to the third data entry and the fourth data entry to generate a second set of user dataset feature learning results; and
inputting the second set of user dataset feature learning results into the trained machine learning model to generate a portion of the second set of unidirectional associations.

11. The information handling system of claim 7 wherein the reference dataset comprises a subset of related entries and a subset of unrelated entries, and wherein the processors perform additional actions comprising:
removing the subset of unrelated entries from the reference dataset to create a prepared reference dataset;
applying a set of lexical relations feature le

Assignees

• IBM

Inventors

• Singh Ritu
• Ranjan Rakesh
• Liu Simao

Classifications

• G06F18/241

  relating to the classification model, e.g. parametric or non-parametric approaches · CPC title

• G06F18/2431

  Multiple classes · CPC title

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06N5/022

  Knowledge engineering; Knowledge acquisition · CPC title

• G06N20/00 Primary

  Machine learning · CPC title