# US 10,872,204 B2 · Generating natural language recommendations based on an industrial language model > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 10,872,204 B2 - Document ID: US10872204B2 - Status: Granted - Kind: B2 - Application number: 16/256,316 - Filed: 2019-01-24 - Published: 2019-08-01 - Granted: 2020-12-22 - Earliest claimed: 2018-01-26 - Applicants: GE Inspection Technologies, LP - Inventors: Xuan Zhu; Arun Karthi Subramaniyan; Haiming Zhao; Zhengjie Xu - Web page: https://patentgrep.com/patent/US10872204B2 ## CPC - G06F 40/284: Lexical analysis, e.g. tokenisation or collocates - G06F 40/216: using statistical methods - G06F 40/30: Semantic analysis - G06N 3/0895: Weakly supervised learning, e.g. semi-supervised or self-supervised learning - G06N 20/00: Machine learning - G06N 20/20: Ensemble learning - G06N 3/045: Combinations of networks - G06N 3/126: Evolutionary algorithms, e.g. genetic algorithms or genetic programming - G06N 5/01: Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound - G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks - G06N 20/10: using kernel methods, e.g. support vector machines [SVM] - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract Systems, methods, and a computer readable medium are provided for generating natural language recommendations based on an industrial language model. Input data, including a plurality of natural language input units is received from a first computing device and transmitted to a second computing device. The second computing device can determine an output including a plurality of natural language outputs using a predictive model trained to generate the output corresponding to an energy exploration lexicon. The second computing device can provide the output to the first computing device for display. ## Claims ### Claim 1 A method comprising: receiving an input provided to a first computing device, including a data processor, the input describing a task performed within an oil and gas production environment and including a plurality of natural language input units and an image of a piece of machinery associated with the task; transmitting the input to a second computing device including a data processor, the transmission performed by the first computing device over a network; determining an output, by the second computing device, the output determined using the transmitted input and a predictive model trained to receive natural language input units and, in response to the receiving, generate an output corresponding to a lexicon of the oil and gas production environment, the output including a plurality of natural language output units predicted to correspond to the input and to the image, the plurality of natural language output units included in the lexicon; and transmitting the output, by the second computing device, to the first computing device, the transmission causing the first computing device to provide the output for display. ### Claim 2 (depends on 1) The method of claim 1, further comprising a second predictive model, the second predictive model trained to determine a second output based on a second input, the second output including a different plurality of natural language output units. ### Claim 3 (depends on 1) The method of claim 1, wherein the plurality of natural language input units includes a word, a part of a word, a character, a sequence of characters, a number, or a sequence of numbers. ### Claim 4 (depends on 1) The method of claim 1, wherein the predictive model has been trained to determine a distribution of words in the input that corresponds to the lexicon based on receiving natural language input units as training input data to a word embedding model in a training phase of a machine learning process, the predictive model trained in the training phase of the machine learning process. ### Claim 5 (depends on 1) The method of claim 1, wherein the predictive model has been trained to classify the input into one or more categories associated with the lexicon based on receiving natural language input units as training input data to a text classification model in a training phase of a machine learning process, the predictive model trained in the training phase of the machine learning process. ### Claim 6 (depends on 1) The method of claim 1, wherein the predictive model has been trained to receive the input and automatically generate outputs as phrases and sentences including the plurality of natural language output units included in the lexicon based on receiving natural language input units as training input data to a text generation model in a training phase of a machine learning process, the predictive model trained in the training phase of the machine learning process. ### Claim 7 (depends on 6) The method of claim 6, wherein the predictive model has been trained to rank the automatically generated outputs based on user preferences using one or more text ranking algorithms in a training phase of a machine learning process, the predictive model trained in the training phase of the machine learning process. ### Claim 8 (depends on 1) The method of claim 1, wherein the first computing device includes a domain-based, auto-suggestion and spelling correction interface configured to receive the input and provide the output for display, the output determined based on a word embedding model and a language model. ### Claim 9 (depends on 1) The method of claim 1, wherein the first computing device includes a contextually aware recommendation agent configured to receive the input and provide the output for display, the output determined based on one or more text ranking algorithms. ### Claim 10 (depends on 1) The method of claim 1, wherein the first computing device includes a domain-specific, semantic search interface configured to receive the input and provide the output for display, the output determined based on a word embedding model and a language model. ### Claim 11 (depends on 1) The method of claim 1, wherein the first computing device includes a content-based text generation interface configured to receive the input and provide the output for display, the output determined based on a text generation model. ### Claim 12 (depends on 1) The method of claim 1, wherein the output includes one or more output units requiring further input to cause execution of executable content configured on the first and/or second computing devices. ### Claim 13 A system comprising: a first computing device, including a data processor and configured to receive and transmit an input describing a task performed within an oil and gas production environment; and a second computing device coupled to the first computing device via a network, the second computing device including a processor and a memory storing computer-readable instructions and a plurality of prediction models, the processor configured to execute the computer-readable instructions, which when executed, cause the processor to perform operations including receiving the input transmitted from the first computing device, the input including a plurality of natural language input units and an image of a piece of machinery associated with the task; determining an output, the output determined using the transmitted input and a predictive model trained to receive natural language input units and, in response to the receiving, generate an output corresponding to a lexicon of the oil and gas production environment, the output including a plurality of natural language output units predicted to correspond to the input and to the image, the plurality of natural language output units included in the lexicon; and transmitting the output to the first computing device, the transmission causing the first computing device to provide the output for display. ### Claim 14 (depends on 13) The system of claim 13, wherein the memory further comprises a second predictive model, the second predictive model trained to determine a second output based on a second input, the second output including a different plurality of natural language output units. ### Claim 15 (depends on 13) The system of claim 13, wherein the plurality of natural language input units includes a word, a part of a word, a character, a sequence of characters, a number, or a sequence of numbers. ### Claim 16 (depends on 13) The system of claim 13, wherein the predictive model has been trained to determine a distribution of words in the input that corresponds to the lexicon based on receiving natural language input units as training input data to a word embedding model in a training phase of a machine learning process, the predictive model trained in the training phase of the machine learning process. ### Claim 17 (depends on 13) The system of claim 13, wherein the predictive model has been trained to classify the input into one or more categories associated with the lexicon based on receiving natural language input units as training input data to a text classification model in a training phase of a machine learning process, the predictive model trained in the training phase of the machine learning process. ### Claim 18 (depends on 13) The system of claim 13, wherein the predictive model has been trained to receive the input and automatically generate outputs as phrases and sentences including the plurality of natural language output units included in the lexicon based on receiving natural language input units as training input data to a text generation model in a training phase of a machine learning process, the predictive model trained in the training phase of the machine learning process. ### Claim 19 (depends on 18) The system of claim 18, wherein the predictive model has been trained to rank the automatically generated outputs based on user preferences using one or more text ranking algorithms in a training phase of a machine learning process, the predictive model trained in the training phase of the machine learning process. ### Claim 20 (depends on 13) The system of claim 13, wherein the first computing device includes a domain-based, auto-suggestion and spelling correction interface configured to receive the input and provide the output for display, the output determined based on a word embedding model and a language model. ### Claim 21 (depends on 13) The system of claim 13, wherein the first computing device includes a contextually aware recommendation agent configured to receive the input and provide the output for display, the output determined based on one or more text ranking algorithms. ### Claim 22 (depends on 13) The system of claim 13, wherein the first computing device includes a domain-specific, semantic search interface configured to receive the input and provide the output for display, the output determined based on a word embedding model and a language model. ### Claim 23 (depends on 13) The system of claim 13, wherein the first computing device includes a content-based text generation interface configured to receive the input and provide the output for display, the output determined based on a text generation model. ### Claim 24 (depends on 13) The system of claim 13, wherein the output includes one or more output units requiring further input to cause execution of executable content configured on the first and/or second computing devices. ### Claim 25 A non-transitory computer readable storage medium containing program instructions, which when executed by at least one data processor causes the at least one data processor to perform operations comprising: receiving an input provided to a first computing device, including a data processor, the input describing a task performed within an oil and gas production environment and including a plurality of natural language input units and an image of a piece of machinery associated with the task; transmitting the input, to a second computing device including a data processor, the transmission performed by the first computing device over a network; determining an output, by the second computing device, the output determined using the transmitted input and a predictive model trained to receive natural language input units and, in response to the receiving, generate an output corresponding to a lexicon of the oil and gas production environment, the output including a plurality of natural language output units predicted to correspond to the input and to the image, the plurality of natural language output units included in the lexicon; and transmitting the output, by the second computing device, to the first computing device, the transmission causing the first computing device to provide the output for display.