Granted Patent B1

US 12,547,654 · App. 19/073,901 · Granted Feb 10, 2026

Inventors: Zachary Michael Ziegler (Cambridge, MA); Daniel Joseph Nadler (Nassau, BS)

Assignee: OpenEvidence Inc.

G06F16/35G06F16/338G06F16/345G06F40/20

Abstract

Claims (68)

1 . A method performed by one or more computers, the method comprising:

receiving query data defining a query from a user;

processing a network input that comprises the query data, using a language model neural network, to generate a set of planning parameters of plan to be executed by the language model neural network to generate a response to the query from the user, wherein the set of planning parameters define at least:

(i) a natural language filtering criterion for filtering a set of documents to identify documents relevant to the query; and

(ii) a set of natural language fields to be extracted from each document that has been identified to be relevant to the query, wherein each natural language field in the set of natural language fields is expressed in a natural language;

executing the plan defined by the set of planning parameters generated by the language model neural network, comprising:

performing a classification operation that comprises classifying, for each document in a set of documents, whether the document is relevant to the query by processing a network input comprising:

(i) the document, and (ii) the natural language filtering criterion previously generated by the language model neural network, using the language model neural network to generate a network output that classifies whether the document is relevant to the query; and

performing an extraction operation that comprises extracting, from each document that is classified as being relevant to the query, natural language data from the document that is relevant to each natural language field in the set of natural language fields generated by the language model neural network; and

wherein latency is reduced and computational efficiency is increased during execution of the plan by:

(a) partitioning the set of documents into a plurality of document subsets, and

(b) assigning each document subset to a respective computing unit in a collection of computing units, wherein each computing unit operates independently and in parallel with the other computing units and performs classification operations and extraction operations on the document subset assigned to the computing unit; and

generating, using the language model neural network, and based on the data extracted from the documents classified as being relevant to the query, a response to the query.

2 . The method of claim 1 , wherein the set of documents are stored in a knowledge base external to the language model neural network.

3 . The method of claim 1 , wherein the set of planning parameters identifies one or more natural language fields in the set of natural language fields based on which documents that have been identified to be relevant to the query should be clustered; and

wherein executing the plan defined by the set of planning parameters further comprises:

generating one or more document clusters each comprising one or more documents based on one or more defined natural language fields.

4 . The method of claim 3 , wherein generating the one or more document clusters comprises generating the one or more document clusters using the language model neural network.

5 . The method of claim 3 , wherein the set of planning parameters define a respective summarization instruction for each document cluster in the one or more document clusters; and

wherein executing the plan defined by the set of planning parameters further comprises:

generating, using the language model neural network and for each document cluster in the one or more document clusters, a respective summary for the document cluster based on processing the respective summarization instruction for the document cluster and at least a portion of each of the one or more documents included in the document cluster.

6 . The method of claim 5 , wherein generating, using the language model neural network and for each document cluster in the one or more document clusters, the respective summary for the document cluster comprises processing, by the language model neural network, the data extracted from each document included in the document cluster.

7 . The method of claim 5 , wherein the respective summary for the document cluster includes, for each of one or more documents included in the document cluster, a citation indicator that identifies the document.

8 . The method of claim 5 , wherein generating the response to the query comprises processing, by the language model neural network, the respective summary for each document cluster in the one or more document clusters.

9 . The method of claim 1 , wherein receiving the query data defining the query from the user comprise receiving the query from the user and by way of a user interface presented to the user on a display of a user device, and wherein the method further comprises:

presenting, by way of the user interface and on the display of the user device, the response to the query.

10 . A system comprising one or more computers and one or more storage devices storing instructions that, when executed by the one or more computers, cause the one or more computers to perform operations comprising:

receiving query data defining a query from a user;

processing a network input that comprises the query data, using a language model neural network, to generate a set of planning parameters of plan to be executed by the language model neural network to generate a response to the query from the user, wherein the set of planning parameters define at least:

(i) a natural language filtering criterion for filtering a set of documents to identify documents relevant to the query; and

(ii) a set of natural language fields to be extracted from each document that has been identified to be relevant to the query, wherein each natural language field in the set of natural language fields is expressed in a natural language;

executing the plan defined by the set of planning parameters generated by the language model neural network, comprising:

performing a classification operation that comprises classifying, for each document in a set of documents, whether the document is relevant to the query by processing a network input comprising:

(i) the document, and (ii) the natural language filtering criterion previously generated by the language model neural network, using the language model neural network to generate a network output that classifies whether the document is relevant to the query; and

performing an extraction operation that comprises extracting, from each document that is classified as being relevant to the query, natural language data from the document that is relevant to each natural language field in the set of natural language fields generated by the language model neural network; and

wherein latency is reduced and computational efficiency is increased during execution of the plan by:

(a) partitioning the set of documents into a plurality of document subsets, and

(b) assigning each document subset to a respective computing unit in a collection of computing units, wherein each computing unit operates independently and in parallel with the other computing units and performs classification operations and extraction operations on the document subset assigned to the computing unit; and

generating, using the language model neural network, and based on the data extracted from the documents classified as being relevant to the query, a response to the query.

11 . The system of claim 10 , wherein the set of documents are stored in a knowledge base external to the language model neural network.

12 . The system of claim 10 , wherein the set of planning parameters identifies one or more natural language fields in the set of natural language fields based on which documents that have been identified to be relevant to the query should be clustered; and

wherein executing the plan defined by the set of planning parameters further comprises:

generating one or more document clusters each comprising one or more documents based on one or more defined natural language fields.

13 . The system of claim 12 , wherein generating the one or more document clusters comprises generating the one or more document clusters using the language model neural network.

14 . The system of claim 12 , wherein the set of planning parameters define a respective summarization instruction for each document cluster in the one or more document clusters; and

wherein executing the plan defined by the set of planning parameters further comprises:

generating, using the language model neural network and for each document cluster in the one or more document clusters, a respective summary for the document cluster based on processing the respective summarization instruction for the document cluster and at least a portion of each of the one or more documents included in the document cluster.

15 . The system of claim 14 , wherein generating, using the language model neural network and for each document cluster in the one or more document clusters, the respective summary for the document cluster comprises processing, by the language model neural network, the data extracted from each document included in the document cluster.

16 . The system of claim 14 , wherein the respective summary for the document cluster includes, for each of one or more documents included in the document cluster, a citation indicator that identifies the document.

17 . The system of claim 14 , wherein generating the response to the query comprises processing, by the language model neural network, the respective summary for each document cluster in the one or more document clusters.

18 . The system of claim 10 , wherein receiving the query data defining the query from the user comprise receiving the query from the user and by way of a user interface presented to the user on a display of a user device, and wherein the operations further comprises:

presenting, by way of the user interface and on the display of the user device, the response to the query.

19 . One or more non-transitory computer storage media storing instructions that, when executed by one or more computers, cause the one or more computers to perform operations comprising:

receiving query data defining a query from a user;

processing a network input that comprises the query data, using a language model neural network, to generate a set of planning parameters of plan to be executed by the language model neural network to generate a response to the query from the user, wherein the set of planning parameters define at least:

(i) a natural language filtering criterion for filtering a set of documents to identify documents relevant to the query; and

(ii) a set of natural language fields to be extracted from each document that has been identified to be relevant to the query, wherein each natural language field in the set of natural language fields is expressed in a natural language;

executing the plan defined by the set of planning parameters generated by the language model neural network, comprising:

performing a classification operation that comprises classifying, for each document in a set of documents, whether the document is relevant to the query by processing a network input comprising:

(i) the document, and (ii) the natural language filtering criterion previously generated by the language model neural network, using the language model neural network to generate a network output that classifies whether the document is relevant to the query; and

performing an extraction operation that comprises extracting, from each document that is classified as being relevant to the query, natural language data from the document that is relevant to each natural language field in the set of natural language fields generated by the language model neural network; and

wherein latency is reduced and computational efficiency is increased during execution of the plan by:

(a) partitioning the set of documents into a plurality of document subsets, and

(b) assigning each document subset to a respective computing unit in a collection of computing units, wherein each computing unit operates independently and in parallel with the other computing units and performs classification operations and extraction operations on the document subset assigned to the computing unit; and

generating, using the language model neural network, and based on the data extracted from the documents classified as being relevant to the query, a response to the query.

20 . The non-transitory computer storage media of claim 19 , wherein the set of planning parameters identifies one or more natural language fields in the set of natural language fields based on which documents that have been identified to be relevant to the query should be clustered; and

wherein executing the plan defined by the set of planning parameters further comprises:

generating one or more document clusters each comprising one or more documents based on one or more defined natural language fields.

Assignments (2)

References Cited (11)