# US 2026/0088014 A1 · Structured description-based chatbot development techniques

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

- Number: US 2026/0088014 A1
- Document ID: US20260088014A1
- Status: Published application
- Kind: A1
- Application number: 19/405,114
- Filed: 2025-12-01
- Published: 2026-03-26
- Earliest claimed: 2022-12-08
- Assignees: Google Inc.
- Applicants: Google LLC
- Inventors: Sasha Goldshtein; Yoav Tzur; Asaf Aharoni; Ofer Ron
- Web page: https://patentgrep.com/patent/US20260088014A1

## CPC

- G06F 40/35: Discourse or dialogue representation
- G10L 13/047: Architecture of speech synthesisers
- G06F 40/30: Semantic analysis
- G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
- G06N 3/045: Combinations of networks
- G10L 13/00: Speech synthesis; Text to speech systems
- G10L 15/22: Procedures used during a speech recognition process, e.g. man-machine dialogue
- H04L 51/02: using automatic reactions or user delegation, e.g. automatic replies or chatbot-generated messages
- G06N 3/006: based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
- G06N 3/0495: Quantised networks; Sparse networks; Compressed networks
- G06N 3/096: Transfer learning
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)

## Abstract

Implementations are directed to receiving unstructured free-form natural language input, generating a chatbot based on the unstructured free-form natural language input and in response to receiving the unstructured free-form natural language input, and causing the chatbot to perform engage in corresponding conversations with additional users. In various implementations, the unstructured free-form natural language input implicitly defines a corresponding dialog state map (e.g., defines corresponding dialog states and/or corresponding dialog state transitions) without defining any explicit dialog states and/or explicit dialog state transitions. In other implementations, the unstructured free-form natural language input is assigned to explicit dialog states and/or explicit dialog state transitions. Nonetheless, the unstructured free-form natural language input may be utilized to fine-tune and/or primed a machine learning model that is already capable of being utilized in conducting generalized conversations. As a result, the chatbot can be generated and deployed in a quick and efficient manner.

## Claims

### Claim 1

A method implemented by one or more processors of a client device, the method comprising:
receiving, at the client device, unstructured free-form natural language input from a user of the client device and on behalf of an entity, the unstructured free-form natural language input including a natural language description of a corresponding dialog state map, the natural language description of the corresponding dialog state map defining implicit dialog states and implicit dialog state transitions of the corresponding dialog state map and without defining any explicit dialog states or any explicit dialog state transitions of the corresponding dialog state map;
in response to receiving the unstructured free-form natural language input that includes the natural language description of the corresponding dialog state map:
generating, based on the unstructured free-form natural language input, a chatbot to engage in a plurality of corresponding conversations with additional users and on behalf of the entity;
causing the chatbot to engage in the plurality of corresponding conversations with additional users and on behalf of the entity, wherein causing the chatbot to engage in a given corresponding conversation, of the plurality of corresponding conversations, with a given additional user, of the additional users, and on behalf of the entity comprises:
causing the chatbot to render a plurality of instances of textual data for presentation to the given additional user, wherein one or more of the plurality of instances of textual data are generated based on the implicit dialog states and/or the implicit dialog state transitions; and
causing responsive content, that is determined responsive to rendering one or more of the plurality of instances of textual data for presentation to the given additional user, to be provided for presentation to the user of the client device.

### Claim 2 (depends on 1)

The method of claim 1, wherein generating the chatbot to engage in the plurality of corresponding conversations with the additional users and on behalf of the entity comprises:
obtaining a previously trained large language model (LLM);
causing the previously trained LLM to be fine-tuned based on the unstructured free-form natural language input to generate a fine-tuned LLM; and
utilizing the fine-tuned LLM as the chatbot.

### Claim 3 (depends on 2)

The method of claim 2, wherein the previously trained LLM is stored in on-device storage of the client device, and wherein the previously trained LLM that is stored in the on-device storage of the client device is a sparsified version of a global previously trained LLM that is available at a remote system communicatively coupled to the client device.

### Claim 4 (depends on 3)

The method of claim 3, wherein the fine-tuned LLM is stored in the on-device storage of the client device.

### Claim 5 (depends on 2)

The method of claim 2, wherein causing the chatbot to render a given instance of textual data, from among the plurality of instances of textual data, for presentation to the given additional user comprises:
processing, using the fine-tuned LLM, one or more features associated with a given implicit dialog state, of the implicit dialog states, to generate the given instance of textual data that reflects a given behavior of the given implicit dialog state; and
transmitting, from the client device and to an additional client device of the given additional user, the given instance of textual data, wherein transmitting the given instance of textual data to the additional client device causes the additional client device to render the given instance of textual data for presentation to the given additional user via the additional client device.

### Claim 6 (depends on 5)

The method of claim 5, further comprising:
processing, using the fine-tuned LLM, and along with one or more of the features associated with a given implicit dialog state, a corresponding context of the corresponding conversation to generate the given instance of textual data that reflects the given behavior of the given implicit dialog state.

### Claim 7 (depends on 3)

The method of claim 3, further comprising:
prior to processing one or more of the features associated with the given implicit dialog state using the fine-tuned LLM:
extracting one or more of the features from the unstructured free-form natural language input.

### Claim 8 (depends on 7)

The method of claim 7, wherein one or more of the features are explicitly included in the unstructured free-form natural language input, and wherein extracting one or more of the features from the unstructured free-form natural language input that are explicitly included in the unstructured free-form natural language input comprises:
utilizing an input parser to extract one or more of the features are explicitly included in the unstructured free-form natural language input.

### Claim 9 (depends on 8)

The method of claim 8, further comprising:
causing the fine-tuned LLM to utilize one or more of the features in generating the given instance of textual data that reflects the given behavior of the given implicit dialog state.

### Claim 10 (depends on 1)

The method of claim 1, wherein the user is not an active participant in the corresponding conversation between the chatbot and the given additional user.

### Claim 11

A method implemented by one or more processors of a client device, the method comprising:
receiving, at the client device, unstructured free-form natural language input from a user of the client device and on behalf of an entity, the unstructured free-form natural language input including a natural language description of a corresponding dialog state map, the natural language description of the corresponding dialog state map defining implicit dialog states and implicit dialog state transitions of the corresponding dialog state map and without defining any explicit dialog states or any explicit dialog state transitions of the corresponding dialog state map;
in response to receiving the unstructured free-form natural language input that includes the natural language description of the corresponding dialog state map, identifying a chatbot to engage in a plurality of corresponding conversations with additional users and on behalf of the entity; and
causing the chatbot to engage in the plurality of corresponding conversations with additional users and on behalf of the entity, wherein causing the chatbot to engage in a given corresponding conversation, of the plurality of corresponding conversations, with a given additional user, of the additional users, and on behalf of the entity comprises:
causing the chatbot to render a plurality of instances of synthesized speech audio data for presentation to the given additional user, wherein one or more of the plurality of instances of synthesized speech are generated based on the implicit dialog states and/or the implicit dialog state transitions; and
causing responsive content, that is determined responsive to rendering one or more of the plurality of instances of synthesized speech for presentation to the given additional user, to be provided for presentation to the user of the client device.

### Claim 12 (depends on 11)

The method of claim 11, wherein identifying the chatbot to engage in the plurality of corresponding conversations with the additional users and on behalf of the entity comprises:
obtaining a previously trained large language model (LLM); and
causing the previously trained LLM to be utilized as the chatbot.

### Claim 13 (depends on 12)

The method of claim 12, wherein the previously trained LLM is stored in on-device storage of the client device, and wherein the previously trained LLM that is stored in the on-device storage of the client device is a sparsified version of a global previously trained LLM that is available at a remote system communicatively coupled to the client device.

### Claim 14 (depends on 12)

The method of claim 12, wherein causing the chatbot to render a given instance of synthesized speech, from among the plurality of instances of synthesized speech, for presentation to the given additional user comprises:
processing, using the previously trained LLM, the unstructured free-form natural language input and one or more features associated with a given implicit dialog state, of the implicit dialog states, to generate an instance of textual data that reflects a given behavior of the given implicit dialog state;
processing, using a text-to-speech (TTS) model, the given instance of textual data that that reflects the given behavior of the given implicit dialog state to generate the given instance of synthesized speech; and
transmitting, from the client device and to an additional client device of the given additional user, the given instance of synthesized speech, wherein transmitting the given instance of synthesized speech to the additional client device causes the additional client device to audibly render the given instance of synthesized speech for presentation to the given additional user via one or more speakers of the additional client device.

### Claim 15 (depends on 12)

The method of claim 12, further comprising:
processing, using the previously trained LLM, and along with one or more of the features associated with a given implicit dialog state, a corresponding context of the corresponding conversation to generate the given instance of textual data that reflects the given behavior of the given implicit dialog state.

### Claim 16 (depends on 12)

The method of claim 12, further comprising:
in response to the given instance of synthesized speech being audibly rendered for presentation to the given additional user via the one or more speakers of the additional client device:
receiving, at the client device and from the additional client device, a given instance of response audio data that includes the responsive content that is responsive to at least the given instance of synthesized speech;
processing, using an automatic speech recognition (ASR) model, the given instance of response audio data to generate a given instance of response textual data; and
determining, based on the given instance of response textual data, whether to:
process, using the previously trained LLM, one or more features associated with the given implicit dialog state and the response textual data to generate an additional instance of textual data that reflects the given behavior of the given implicit dialog state, or
process, using the previously trained LLM, one or more additional features associated with a given additional implicit dialog state, of the implicit dialog states, and the response textual data to generate an additional instance of textual data that reflects a given additional behavior of the given additional implicit dialog state.

### Claim 17 (depends on 14)

The method of claim 14, wherein causing the responsive content to be provided for presentation to the user of the client device includes a result of the corresponding conversation.

### Claim 18 (depends on 12)

The method of claim 12, further comprising:
prior to processing one or more of the features associated with the given implicit dialog state using the previously trained LLM:
extracting one or more of the features from the unstructured free-form natural language input.

### Claim 19 (depends on 18)

The method of claim 18, wherein one or more of the features are explicitly included in the unstructured free-form natural language input, and wherein extracting one or more of the features from the unstructured free-form natural language input that are explicitly included in the unstructured free-form natural language input comprises:
utilizing an input parser to extract one or more of the features are explicitly included in the unstructured free-form natural language input.

### Claim 20 (depends on 19)

The method of claim 19, further comprising:
causing the previously trained LLM to utilize one or more of the features in generating the instance of textual data that reflects the given behavior of the given implicit dialog state.
