# US 2026/0311692 A1 · Sentiment-based conversation hotspot detection

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0311692 A1
- Document ID: US20260311692A1
- Status: Published application
- Kind: A1
- Application number: 19/539,869
- Filed: 2026-02-13
- Published: 2026-10-08
- Earliest claimed: 2023-09-29
- Assignees: Amazon Technologies, Inc.
- Applicants: Amazon Technologies, Inc.
- Inventors: Gizem Tabak; Masahito Togami; Michael Mark Goodwin; Amalavoyal Chari; Siddhartha Shankara Rao
- Web page: https://patentgrep.com/patent/US20260311692A1

## CPC

- G10L 15/1815: {Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning}
- G10L 15/04: Segmentation; Word boundary detection
- G10L 15/22: Procedures used during a speech recognition process, e.g. man-machine dialogue
- G08B 21/182: {Level alarms, e.g. alarms responsive to variables exceeding a threshold}
- G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G08B: SIGNALLING SYSTEMS, e.g. PERSONAL CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS

## Abstract

A system may include machine learning models. A system may receive audio information representing an utterance of a conversation session. A system may divide the audio information into a plurality of audio portions. A system may evaluate a first audio portion using a tone-based sentiment analysis model to generate sentiment probabilities. A system may determine a first positive sentiment probability exceeds a threshold. A system may generate a textual representation of the first audio portion. A system may evaluate the textual representation using a topic identification model to generate a topic result indicating a topic. A system may evaluate a second audio portion using the tone-based sentiment analysis model to generate second sentiment probabilities. A system may determine a second positive sentiment probability does not exceed the threshold.

## Claims

### Claim 1

(canceled)

### Claim 2

A computer-implemented method comprising:
under control of a computing device comprising one or more processors configured to execute specific instructions,
receiving audio information representing one or more utterances;
identifying, using a first machine learning model and based on a textual representation of the one or more utterances, a topic instance;
determining, using a second machine learning model, a sentiment associated with the topic instance;
generating, based on the sentiment associated with the topic instance, a sentiment score associated with the topic instance;
ranking, based on the sentiment score associated with the topic instance, the topic instance relative to a second topic instance associated with the one or more utterances; and
generating output data comprising the sentiment score and a topic associated with the topic instance.

### Claim 3 (depends on 2)

The computer-implemented method of claim 2, further comprising generating the textual representation of the one or more utterances.

### Claim 4 (depends on 2)

The computer-implemented method of claim 2, wherein using the second machine learning model comprises using a text-based sentiment detection model.

### Claim 5 (depends on 2)

The computer-implemented method of claim 2, wherein using the second machine learning model comprises using an audio-based sentiment detection model.

### Claim 6 (depends on 2)

The computer-implemented method of claim 2, wherein identifying the topic instance comprises determining a time window indicating a time in the one or more utterances associated with the topic instance.

### Claim 7 (depends on 2)

The computer-implemented method of claim 2, wherein identifying the topic instance comprises generating an indication of the topic.

### Claim 8 (depends on 2)

The computer-implemented method of claim 2, further comprising determining, using a text-based topic detection model and a portion of the textual representation associated with the topic instance, the topic.

### Claim 9 (depends on 2)

The computer-implemented method of claim 2, further comprising determining, using a text-based topic detection model and a portion of the textual representation associated with the topic instance, a topic category associated with the topic instance, wherein the topic category is selected from a plurality of existing topic categories.

### Claim 10 (depends on 2)

The computer-implemented method of claim 2, further comprising:
identifying, using the first machine learning model and based on the textual representation of the one or more utterances, the second topic instance;
determining, using the second machine learning model, a second sentiment associated with the second topic instance; and
generating, based on the second sentiment, a second sentiment score associated with the second topic instance,
wherein ranking the topic instance relative to the second topic instance is further based on the second sentiment score.

### Claim 11 (depends on 2)

The computer-implemented method of claim 2, wherein generating the output data further comprises generating an indication of the ranking of the topic instance relative to the second topic instance.

### Claim 12 (depends on 2)

The computer-implemented method of claim 2, wherein the one or more utterances are associated with a first session, and wherein the computer-implemented method further comprises:
receiving second audio information representing a second one or more utterances;
identifying, using the first machine learning model and based on a second textual representation of the second one or more utterances, a third topic instance, wherein the third topic instance is associated with the topic;
determining, using the second machine learning model, a second sentiment associated with the third topic instance; and
generating, based on the second sentiment associated with the third topic instance, a second sentiment score associated with the third topic instance;
analyzing the sentiment score and the second sentiment score to determine a most common sentiment associated with the topic,
wherein the output data further comprises an indication of the most common sentiment.

### Claim 13

A system comprising:
a computer-readable memory storing computer-executable instructions; and
one or more processors configured to execute the computer-executable instructions to at least:
receive audio information representing one or more utterances;
identify, using a first machine learning model and based on a textual representation of the one or more utterances, a topic instance;
determine, using a second machine learning model, a sentiment associated with the topic instance;
generate, based on the sentiment associated with the topic instance, a sentiment score associated with the topic instance;
rank, based on the sentiment score associated with the topic instance, the topic instance relative to a second topic instance associated with the one or more utterances; and
generate output data comprising the sentiment score and a topic associated with the topic instance.

### Claim 14 (depends on 13)

The system of claim 13, wherein the computer executable instructions, when executed by the one or more processors, further cause the one or more processors to generate the textual representation of the one or more utterances

### Claim 15 (depends on 13)

The system of claim 13, wherein the second machine learning model comprises a text-based sentiment detection model.

### Claim 16 (depends on 13)

The system of claim 13, wherein the second machine learning model comprises an audio-based sentiment detection model.

### Claim 17 (depends on 13)

The system of claim 13, wherein the topic instance comprises a time window indicating a time in the one or more utterances associated with the topic instance.

### Claim 18 (depends on 13)

The system of claim 13, wherein the topic instance comprises an indication of the topic.

### Claim 19 (depends on 13)

The system of claim 13, wherein the output data comprises an indication of the ranking of the topic instance relative to the second topic instance.

### Claim 20 (depends on 13)

The system of claim 13, wherein the computer executable instructions, when executed by the one or more processors, further cause the one or more processors to determine, using a text-based topic detection model and a portion of the textual representation associated with the topic instance, a topic category associated with the topic instance, wherein the topic category is one of a plurality of existing topic categories.

### Claim 21 (depends on 12)

The system of claim 12, wherein the topic is determined using a text-based topic detection model and a portion of the textual representation associated with the topic instance.
