# US 10,482,183 B1 · Device and method for natural language processing through statistical model comparison > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 10,482,183 B1 - Document ID: US10482183B1 - Status: Granted - Kind: B1 - Application number: 16/144,718 - Filed: 2018-09-27 - Granted: 2019-11-19 - Earliest claimed: 2018-09-27 - Assignees: Babylon Partners Limited - Applicants: Babylon Partners Limited - Inventors: Francisco Vargas; Kamen Brestnichki; Dane Grant Sherburn; Vitalii Zhelezniak; Nils Yannick Hammerla - Web page: https://patentgrep.com/patent/US10482183B1 ## CPC - G06F 40/30: Semantic analysis - G06F 40/40: Processing or translation of natural language (natural language analysis G06F40/20; semantic analysis G06F40/30) - G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks - G10L 15/22: Procedures used during a speech recognition process, e.g. man-machine dialogue - G16H 50/20: for computer-aided diagnosis, e.g. based on medical expert systems - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING - G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA ## Abstract A computer-implemented method comprising: receiving the first set of words and the second set of words, wherein each of the first and second sets of words; calculating a first likelihood-based measure representing how well a first model can be fit to the first and second sets of words, the first model comprising a shared parametric distribution representing both the first and second sets of words; calculating a second likelihood-based measure representing how well a second model can be fit to the first and second sets of words, the second model comprising a first parametric distribution representing the first set of words and a second parametric distribution representing the second set of words; calculating a similarity score based on a ratio of the first likelihood measure to the second likelihood measure, the similarity score being representative of the similarity between the first and second sets of words; and outputting the similarity score. ## Claims ### Claim 1 A computer-implemented method for natural language processing for a human-machine interface through statistical model comparison, comprising: receiving a first set of words and a second set of words; calculating a first likelihood-based measure representing how well a first model can be fit to the first and second sets of words, the first model comprising a shared parametric distribution representing both the first and second sets of words; calculating a second likelihood-based measure representing how well a second model can be fit to the first and second sets of words, the second model comprising a first parametric distribution representing the first set of words and a second parametric distribution representing the second set of words; calculating a similarity score based on a ratio of the first likelihood measure to the second likelihood measure, the similarity score being representative of the similarity between the first and second sets of words; outputting the similarity score to the human-machine interface to determine a semantic similarity between the first set of words and the second set of words; and associating, through the human-machine interface, the first set of words with the second set of words based on the semantic similarity to determine a response. ### Claim 2 (depends on 1) The method of claim 1 wherein the first, second and shared parametric distributions are first, second and shared likelihood functions respectively, each defined by a respective set of parameters. ### Claim 3 (depends on 2) The method of claim 2 wherein the first and second sets of words are first and second sets of word embeddings constrained to lie on the surface of a unit hypersphere, and wherein the first, second and shared likelihood functions are in accordance with the von Mises-Fisher distribution. ### Claim 4 (depends on 2) The method of claim 2 further comprising determining maximum likelihood estimates for the parameters for each likelihood function. ### Claim 5 (depends on 4) The method of claim 4 wherein the first likelihood-based measure is calculated based on a maximum value for the shared likelihood function and the second likelihood-based measure is calculated based on maximum values for the first and second likelihood functions. ### Claim 6 (depends on 5) The method of claim 5 wherein the first and second information criteria are in accordance with the Takeuchi Information Criterion or the Akaiki Information Criterion. ### Claim 7 (depends on 1) The method of claim 1 wherein the first and second likelihood based measures are first and second information criteria respectively. ### Claim 8 (depends on 1) The method of claim 1 wherein the first and second likelihood-based measures each comprise a respective penalty based on complexity for the respective model. ### Claim 9 A system for natural language processing for a human-machine interface through statistical model comparison comprising one or more processors configured to: receive a first set of words and a second set of words; calculate a first likelihood-based measure representing how well a first model can be fit to the first and second sets of words, the first model comprising a shared parametric distribution representing both the first and second sets of words; calculate a second likelihood-based measure representing how well a second model can be fit to the first and second sets of words, the second model comprising a first parametric distribution representing the first set of words and a second parametric distribution representing the second set of words; calculate a similarity score based on a ratio of the first likelihood measure to the second likelihood measure, the similarity score being representative of the similarity between the first and second sets of words; output the similarity score to the human-machine interface to determine a semantic similarity between the first set of words and the second set of words; and associate, through the human-machine interface, the first set of words with the second set of words based on the semantic similarity to determine a response. ### Claim 10 (depends on 9) The system of claim 9 wherein the first, second and shared parametric distributions are first, second and shared likelihood functions respectively, each defined by a respective set of parameters. ### Claim 11 (depends on 10) The system of claim 10 wherein the first and second sets of words are first and second sets of word embeddings constrained to lie on the surface of a unit hypersphere, and wherein the first, second and shared likelihood functions are in accordance with the von Mises-Fisher distribution. ### Claim 12 (depends on 10) The system of claim 10 wherein the one or more processors are further configured to determine maximum likelihood estimates for the parameters for each likelihood function. ### Claim 13 (depends on 12) The system of claim 12 wherein the first likelihood-based measure is calculated based on a maximum value for the shared likelihood function and the second likelihood-based measure is calculated based on respective maximum values for the first and second likelihood functions. ### Claim 14 (depends on 9) The system of claim 9 wherein the first and second likelihood based measures are first and second information criteria respectively. ### Claim 15 (depends on 14) The system of claim 14 wherein the first and second information criteria are in accordance with the Takeuchi Information Criterion or the Akaiki Information Criterion. ### Claim 16 (depends on 9) The system of claim 9 wherein the first and second likelihood-based measures each comprise a respective penalty based on complexity for the respective model. ### Claim 17 A non-transient computer readable medium containing programming instructions for natural language processing for a human-machine interface through statistical model comparison that, when executed by a computer, cause the computer to: receive a first set of words and a second set of words; calculate a first likelihood-based measure representing how well a first model can be fit to the first and second sets of words, the first model comprising a shared parametric distribution representing both the first and second sets of words; calculate a second likelihood-based measure representing how well a second model can be fit to the first and second sets of words, the second model comprising a first parametric distribution representing the first set of words and a second parametric distribution representing the second set of words; calculate a similarity score based on a ratio of the first likelihood measure to the second likelihood measure, the similarity score being representative of the similarity between the first and second sets of words; output the similarity score to the human-machine interface to determine a semantic similarity between the first set of words and the second set of words; and associate, through the human-machine interface, the first set of words with the second set of words based on the semantic similarity to determine a response.