Methods, apparatuses and computer program products for parsing temporal expressions in a conversational data-to-text system

Patent metadata

Field	Value
Publication number	US-12254266-B2
Application number	US-202117463031-A
Country	US
Kind code	B2
Filing date	Aug 31, 2021
Priority date	Aug 31, 2020
Publication date	Mar 18, 2025
Grant date	Mar 18, 2025

Abstract

Embodiments provide for a temporal expression parser in a conversational data-to-text system are described herein. An example method may include receiving user query data comprising an input text string; generating, based at least in part on the input text string, a n-gram set comprising a plurality of n-gram elements; traversing each n-gram element in the n-gram set to generate a parse tree list comprising one or more parse trees based on a grammar template associated with the input text string; and generating, based at least in part on a last parse tree of the parse tree list, one or more semantic frames indicating a temporal expression associated with the input text string.

First claim

The invention claimed is:

1. An apparatus comprising at least one processor and at least one non-transitory memory comprising program code, the at least one non-transitory memory and the program code configured to, with the at least one processor, cause the apparatus to at least:
receive user query data comprising an input text string;
generate, based at least in part on the input text string, a n-gram set comprising a plurality of n-gram elements, wherein an arrangement of the plurality of n-gram elements in the n-gram set is determined based at least in part on a location of each n-gram element with respective to the input text string and a size of each n-gram element;
traverse each n-gram element in the n-gram set to generate a parse tree list comprising one or more parse trees based at least in part on a grammar template associated with the input text string;
generate, based at least in part on one or more of a node of a last parse tree of the parse tree list representing at least one stop word or joining a pair of non-contiguous nodes that represent temporal words in the last parse tree, one or more semantic frames indicating a temporal expression associated with the input text string, wherein the last parse tree comprises a plurality of nodes generated based at least in part on the grammar template;
provide, to a dimensional data model, the one or more semantic frames;
retrieve, by the dimensional data model, from a multi-dimensional database and based at least in part on the one or more semantic frames, one or more multi-dimensional data objects, wherein each multi-dimensional data object of the one or more multi-dimensional data objects is placed in a feature space, wherein the feature space comprises one or more dimensions corresponding to categorial data and measures representing numerical data represented by the multi-dimensional data object; and
generate, based at least in part on executing an analytic query on the one or more multi-dimensional data objects, an output response, wherein the analytic query is determined based at least in part on the temporal expression associated with the input text string.

2. The apparatus of claim 1 , wherein, prior to traversing each n-gram element in the n-gram set, the at least one non-transitory memory and the program code are configured to, with the at least one processor, cause the apparatus to further:
retrieve, from a grammar database, the grammar template corresponding to the input text string, wherein the grammar template defines a grammar structure associated with the input text string.

3. The apparatus of claim 1 , wherein the grammar template is a context-free grammar.

4. The apparatus of claim 1 , wherein the at least one non-transitory memory and the program code are configured to, with the at least one processor, cause the apparatus to traverse each n-gram element in the n-gram set according to the arrangement of the plurality of n-gram elements.

5. The apparatus of claim 1 , wherein the at least one non-transitory memory and the program code are configured to, with the at least one processor, cause the apparatus to further:
when traversing each n-gram in the n-gram set to generate the parse tree list,
determine, for a n-gram element in the n-gram set, whether the grammar template defines a corresponding parse tree for the n-gram element; and
based at least in part on determining that the grammar template defines the corresponding parse tree for the n-gram element, add the corresponding parse tree to the parse tree list.

6. The apparatus of claim 5 , wherein the last parse tree of the parse tree list is defined by the grammar template for a largest n-gram element within the n-gram set.

7. The apparatus of claim 1 , wherein, when generating the one or more semantic frames, the at least one non-transitory memory and the program code are configured to, with the at least one processor, cause the apparatus to further:
determine, based at least in part on the grammar template, whether the last parse tree corresponds to a time point, a time interval, or a variance period.

8. The apparatus of claim 7 , wherein the at least one non-transitory memory and the program code are configured to, with the at least one processor, cause the apparatus to further:
in response to determining that the last parse tree corresponds to the time point:
normalize each node of the last parse tree; and
add at least one inferred node to the last parse tree based at least in part on a current time descriptor and a temporal granularity descriptor associated with the input text string.

9. The apparatus of claim 7 , wherein the at least one non-transitory memory and the program code are configured to, with the at least one processor, cause the apparatus to further:
in response to determining that the last parse tree corresponds to the time interval:
normalize each node of the last parse tree;
add at least one inferred node to the last parse tree based at least in part on a current time descriptor and a temporal granularity descriptor associated with the input text string; and
arrange a first node of the last parse tree that represents a first time point and a second node of the last parse tree that represents a second time point.

10. The apparatus of claim 7 , wherein the at least one non-transitory memory and the program code are configured to, with the at least one processor, cause the apparatus to further:
in response to determining that the last parse tree corresponds to the variance period:
determine a first node of last parse tree that represents a current interval based at least in part on a current time descriptor; and
determine a second node of the last parse tree that represents a comparison interval based at least in part on the current interval, wherein the comparison interval represents a same interval length as the current interval.

11. A computer-implemented method comprising:
receiving user query data comprising an input text string;
generating, based at least in part on the input text string, a n-gram set comprising a plurality of n-gram elements, wherein an arrangement of the plurality of n-gram elements in the n-gram set is determined based at least in part on a location of each n-gram element with respective to the input text string and a size of each n-gram element;
traversing each n-gram element in the n-gram set to generate a parse tree list comprising one or more parse trees based at least in part on a grammar template associated with the input text string;
generating, based at least in part on one or more of a node of a last parse tree of the parse tree list representing at least one stop word or joining a pair of non-contiguous nodes that represent temporal words in the last parse tree, one or more semantic frames indicating a temporal expression associated with the input text string, wherein the last parse tree comprises a plurality of nodes generated based at least in part on the grammar template;
providing, to a dimensional data model, the one or more semantic frames;
retrieving, by the dimensional data model, from a multi-dimensional database and based at least in part on the one or more semantic frames, one or more multi-dimensional data objects, wherein each multi-dimensional data object of the one or more multi-dimensional data objects is placed in a feature space, wherein the feature space comprises one or more dimensions corresponding to categorial data and measures representing numerical data represented by the multi-dimensional data object; and
generating, based at least in part on executing an analytic query on the one or more multi-dimensional data objects, an output response, wherein the analytic query

Assignees

• Arria Data2Text Ltd

Inventors

• De Oliveira Rodrigo Gomes
• Alexander John William

Classifications

• G06F16/245

  Query processing · CPC title

• G06F40/30

  Semantic analysis · CPC title

• G06F40/253

  Grammatical analysis; Style critique · CPC title

• G06F40/211 Primary

  Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars · CPC title

• G06F16/3329

  Natural language query formulation · CPC title