Learning approximate translations of unfamiliar measurement units during deep question answering system training and usage

Patent metadata

Field	Value
Publication number	US-11556758-B2
Application number	US-201916552379-A
Country	US
Kind code	B2
Filing date	Aug 27, 2019
Priority date	Aug 27, 2019
Publication date	Jan 17, 2023
Grant date	Jan 17, 2023

Abstract

A method learns approximate translations of unfamiliar measurement units during deep question answering (DeepQA) system training and usage. The DeepQA system receives a training set containing Question-Answer (QA) pairs having known unit-of-measurement terms, where each QA pair contains an answer having a known numeric value for a corresponding question from the QA pair. The DeepQA system receives a question from each QA pair from the training set to the DeepQA system in order to find answers and passage phrases to the question from each QA pair, and then identifies all found answers and passage phrases having values that are within a predetermined range of answer values of the training set, where one or more of the identified all found answers and passage phrases contain unfamiliar unit-of-measurement terms, in order to learn approximate translations of the unfamiliar unit-of-measurement terms.

First claim

What is claimed is:

1. A method comprising:
receiving, by a deep question answering (DeepQA) system, a training set containing Question-Answer (QA) pairs having known unit-of-measurement terms, wherein each QA pair contains an answer having a known numeric value;
submitting a question from each QA pair from the training set to the DeepQA system in order to find answers and passage phrases to the question from each QA pair;
identifying, by the DeepQA system, all found answers and passage phrases having values that are within a predetermined range of answer values of the training set, wherein one or more of the identified all found answers and passage phrases contain unfamiliar unit-of-measurement terms, wherein unit-of-measurement terms describe concepts associated with the answer values, wherein the unfamiliar unit-of-measurement terms are unknown measurement terms that are unknown to the DeepQA system, and wherein the unknown measurement terms are not found in an answer key used by the DeepQA system;
defining, by the DeepQA system, the identified all found answers and passage phrases as hypothetical unit answers for answering a particular type of question;
for each hypothetical unit answer, inferring, by the DeepQA system, a hypothetical numeric translation value based on the known numeric value from the answer in the training set;
comparing, by the DeepQA system, the hypothetical numeric translation value for each hypothetical unit answer to numeric values from hypothetical numeric translation values that are applied to other answers and passage phrases from the identified all found answers and passage phrases;
evaluating, by the DeepQA system, how consistent the hypothetical numeric translation value is for each hypothetical unit answer from the identified all found answers and passage phrases;
utilizing, by the DeepQA system, how consistent the hypothetical numeric translation value is for each hypothetical unit answer to compute an overall translation value for the identified all found answers and passage phrases;
utilizing the answer key to train the DeepQA system to search for passages that answer the particular type of question, wherein the trained DeepQA system outputs an answer key value and an answer key measurement unit that is associated with the answer key value;
identifying a candidate answer that includes a candidate passage containing the answer key value but not the answer key measurement unit, wherein a candidate passage measurement unit in the candidate passage is associated with the answer key value;
matching the answer key measurement unit to the candidate passage measurement unit based on the answer key measurement unit and the candidate passage measurement unit both being associated with the answer key value;
retraining the trained DeepQA system to retrieve answers, for the particular type of question, that include one or more of the answer key measurement unit and the candidate passage measurement unit, wherein the retrained DeepQA system controls a physical aircraft;
receiving, by the retrained DeepQA system, the particular type of question;
retrieving, by the retrained DeepQA system, a passage that includes the answer key value and one or more of the answer key measurement unit and the candidate passage measurement unit;
answering, by the retrained DeepQA system, the particular type of question with the passage that includes the answer key value and one or more of the answer key measurement unit and the candidate passage measurement unit;
recognizing, by the retrained DeepQA system, that the physical aircraft is flying below an identified minimum approach speed while landing, wherein the minimum approach speed is identified by the retrained DeepQA system, and wherein an auto-assist hydraulic system is required to control a rudder of the physical aircraft if the physical aircraft is flying below the identified minimum approach speed while landing; and
in response to recognizing that the physical aircraft is flying below the identified minimum approach speed while landing, automatically activating, by the retrained DeepQA system, the auto-assist hydraulic system to control a rudder of the physical aircraft while the physical aircraft is landing.

2. The method of claim 1 , further comprising:
utilizing, by the DeepQA system, the overall translation value to update an answer key with newly discovered appropriate answers and passage phrases.

3. The method of claim 2 , wherein an answer key value in the answer key and an answer value in one or more of the identified all found answers and passage phrases that contain the unfamiliar unit-of-measurement terms are exact matches.

4. The method of claim 2 , wherein an answer key value in the answer key and an answer value in one or more of the identified all found answers and passage phrases that contain the unfamiliar unit-of-measurement terms are within a predefined range.

5. The method of claim 1 , further comprising:
receiving a new question by a trained DeepQA system that has been trained to recognize the unfamiliar unit-of-measurement terms as being equivalent to the known unit-of-measurement terms;
analyzing the new question to determine a lexical answer type (LAT) for the new question;
generating answers, to the new question, from one or more of the identified all found answers and passage phrases that contain the unfamiliar unit-of-measurement terms; and
selecting an answer based on a matching to the LAT and the hypothetical numeric translation value.

6. The method of claim 1 , wherein the concept described by the unit-of-measurement terms is speed, and wherein different unit-of-measurement terms use different scales to describe a same concept associated with the answer values.

7. A computer system comprising one or more processors, one or more computer readable memories, and one or more computer readable non-transitory storage mediums, and program instructions stored on at least one of the one or more computer readable non-transitory storage mediums for execution by at least one of the one or more processors via at least one of the one or more computer readable memories, the stored program instructions executed to perform a method comprising:
receiving a training set containing Question-Answer (QA) pairs having known unit-of-measurement terms, wherein each QA pair contains an answer having a known numeric value;
receiving a question from each QA pair from the training set in order to find answers and passage phrases to the question from each QA pair;
identifying all found answers and passage phrases having values that are within a predetermined range of answer values of the training set, wherein one or more of the identified all found answers and passage phrases contain unfamiliar unit-of-measurement terms, wherein unit-of-measurement terms describe concepts associated with the answer values, wherein the unfamiliar unit-of-measurement terms are unknown measurement terms that are unknown to the DeepQA system, and wherein the unknown measurement terms are not found in an answer key used by the DeepQA system;
defining the identified all found answers and passage phrases as hypothetical unit answers for answering a particular type of question;
for each hypothetical unit answer, inferring a hypothetical numeric translation value based on the known numeric value from the answer in the training set;
comparing the hypothetical numeric translation value for each hypothetical unit answer to numeric values from other answers and passage phrases from the identified all answers and passage phrases;
evaluating how consistent the hypothetical numeric translation value is for each hypothetical unit answer from the identified all found answers and passage phrases; and
utilizing how consistent

Assignees

• IBM

Inventors

• Katz Edward G
• Beller Charles E
• Boxwell Stephen A
• Summers Kristen M

Classifications

• G06N3/084 Primary

  Backpropagation, e.g. using gradient descent · CPC title

• G06N3/042 Primary

  Knowledge-based neural networks; Logical representations of neural networks · CPC title

• G06N5/041

  Abduction · CPC title

• G06N3/08

  Learning methods · CPC title

• G06N3/045

  Combinations of networks · CPC title