Text reduction and analysis interface to a text generation modeling system
A query request identifying a query and a plurality of text portions for determining an answer to the query may be received. Relevance scores corresponding with respective ones of the text portions may be determined based on application of one or more machine learning models to the respective text portion and the query. A subset of the text portions may be selected based on the relevance scores. A response message to the query request including an answer to the query in natural language text generated by a large language model based on the first subset of text portions may be determined.
1. A method comprising:
receiving via a communication interface a request to generate a timeline based on a plurality of input documents;
determining a plurality of event identification prompts via a processor, an event identification prompt of the plurality of event identification prompts being determined by replacing a designated one or more fillable portions of an event identification prompt template with a respective text portion of the plurality of input documents, the event identification prompt and the event identification prompt template each including a natural language event identification instruction to identify events based on the respective text portion;
identifying an initial plurality of event entries based on novel text generated by a generative language model executing the natural language event identification instruction in the plurality of event identification prompts;
determining a deduplicated plurality of event entries based on the initial plurality of event entries; and
transmitting to a client machine a timeline generation message that includes an instruction for generating the timeline, the instruction including information for representing the deduplicated plurality of event entries on the timeline.
2. The method recited in claim 1 , wherein the event identification prompt includes a natural language notability instruction instructing the generative language model to generate text characterizing notability of the events.
3. The method recited in claim 2 , wherein an initial event entry of the initial plurality of event entries includes a notability text portion characterizing notability of the initial event entry.
4. The method recited in claim 1 , wherein the event identification prompt includes a date portion that identifies a date on which the event identification prompt is created.
5. The method recited in claim 1 , wherein identifying the initial plurality of event entries involves parsing structured text generated in accordance with a predetermined output format identified in a natural language output format portion of the event identification prompt template.
6. The method recited in claim 1 , the method further comprising:
transmitting the plurality of event identification prompts to a remote text generation system; and
receiving from the remote text generation system a corresponding plurality of completed event identification prompts generated by the generative language model based on the plurality of event identification prompts.
7. The method recited in claim 1 , wherein determining the deduplicated plurality of event entries comprises:
determining an event deduplication prompt via the processor by replacing one or more fillable portions of an event deduplication prompt template with the initial plurality of event entries, the event deduplication prompt and the event deduplication prompt template each including a natural language event deduplication instruction to deduplicate the initial plurality of event entries.
8. The method recited in claim 7 , wherein the deduplicated plurality of event entries are determined based on deduplicated event text generated by the generative language model executing the natural language event deduplication instruction in the event deduplication prompt.
9. The method recited in claim 7 , the method further comprising:
transmitting the event deduplication prompt to a remote text generation system; and
receiving from the remote text generation system a completed event deduplication prompt by the generative language model based on the event deduplication prompt.
10. A system comprising:
a communication interface configured to receive a request to generate a timeline based on a plurality of input documents;
a processor configured to determine a plurality of event identification prompts, an event identification prompt of the plurality of event identification prompts being determined by replacing a designated one or more fillable portions of an event identification prompt template with a respective text portion of the plurality of input documents, the event identification prompt and the event identification prompt template each including a natural language event identification instruction to identify events based on the respective text portion;
memory configured to store an initial plurality of event entries identified based on novel text generated by a generative language model executing the natural language event identification instruction in the plurality of event identification prompts, wherein the processor is further configured to determine a deduplicated plurality of event entries based on the initial plurality of event entries, and wherein the communication interface is further configured to transmit to a client machine a timeline generation message that includes an instruction for generating the timeline, the instruction including information for representing the deduplicated plurality of event entries on the timeline.
11. The system recited in claim 10 , wherein the event identification prompt includes a natural language notability instruction instructing the generative language model to generate text characterizing notability of the events.
12. The system recited in claim 11 , wherein an initial event entry of the initial plurality of event entries includes a notability text portion characterizing notability of the initial event entry.
13. The system recited in claim 10 , wherein the event identification prompt includes a date portion that identifies a date on which the event identification prompt is created.
14. The system recited in claim 10 , wherein identifying the initial plurality of event entries involves parsing structured text generated in accordance with a predetermined output format identified in a natural language output format portion of the event identification prompt template.
15. The system recited in claim 10 , wherein determining the deduplicated plurality of event entries comprises:
determining an event deduplication prompt via the processor by replacing one or more fillable portions of an event deduplication prompt template with the initial plurality of event entries, the event deduplication prompt and the event deduplication prompt template each including a natural language event deduplication instruction to deduplicate the initial plurality of event entries.
16. The system recited in claim 15 , wherein the deduplicated plurality of event entries are determined based on deduplicated event text generated by the generative language model executing the natural language event deduplication instruction in the event deduplication prompt.
17. One or more non-transitory computer-readable media having instructions stored thereon for performing a method, the method comprising:
receiving via a communication interface a request to generate a timeline based on a plurality of input documents;
determining a plurality of event identification prompts via a processor, an event identification prompt of the plurality of event identification prompts being determined by replacing a designated one or more fillable portions of an event identification prompt template with a respective text portion of the plurality of input documents, the event identification prompt and the event identification prompt template each including a natural language event identification instruction to identify events based on the respective text portion;
identifying an initial plurality of event entries based on novel text generated by a generative language model executing the natural language event identification instruction in the plurality of event identification prompts;
determining a deduplicated plurality of event entries based on the initial plurality of event entries; and
transmitting to a client machine a timeline generation message that includes an instruction for generating the timeline, the instruction including information for representing the deduplicated plurality of event entries on the timeline.
18. The non-transitory computer-readable media recited in claim 17 , wherein the event identification prompt includes a natural language notability instruction instructing the generative language model to generate text characterizing notability of the events, wherein an initial event entry of the initial plurality of event entries includes a notability text portion characterizing notability of the initial event entry.
19. The non-transitory computer-readable media recited in claim 17 , the method further comprising:
transmitting the plurality of event identification prompts to a remote text generation system; and
receiving from the remote text generation system a corresponding plurality of completed event identification prompts generated by the generative language model based on the plurality of event identification prompts.
20. The non-transitory computer-readable media recited in claim 17 , wherein determining the deduplicated plurality of event entries comprises:
determining an event deduplication prompt via the processor by replacing one or more fillable portions of an event deduplication prompt template with the initial plurality of event entries, the event deduplication prompt and the event deduplication prompt template each including a natural language event deduplication instruction to deduplicate the initial plurality of event entries, wherein the deduplicated plurality of event entries are determined based on deduplicated event text generated by the generative language model executing the natural language event deduplication instruction in the event deduplication prompt, the method further comprising:
transmitting the event deduplication prompt to a remote text generation system; and
receiving from the remote text generation system a completed event deduplication prompt by the generative language model based on the event deduplication prompt.