Automated, parameter-pattern-driven, data mining system based on customizable chain of machine-learning-structures providing an automated data-processing pipeline, and method thereof
Proposed is a parameter pattern-driven, data mining system and corresponding method with a knowledge extraction engine based on a customizable chain of machine-learning-structures providing an automated pipeline for data processing of complex data structures with a hidden pattern detection for triggering automated under-writing processes. A plurality of digital risk-transfer policies is assessed via a data interface and storable captured by a persistence repository unit of the parameter pattern-driven, data mining system. The digital policy at least comprises premium parameter values and/or deducible parameter values and/or risk-transfer type definition parameter values and/or policy limits parameter values and/or exclusion parameter values and/or riders/addit parameter values. The parameter pattern-driven, data mining system comprises a chained series of machine learning modeling structures automatically assessing and parsing digital risk-transfer policies of a policyholder, and automatically translating contractual language of the digital policy into actionable offers for the policyholder by generating appropriate new digital risk-transfer policies for automated under writing by the policyholder.
1 . A parameter pattern-driven, digital, data mining system, the system comprising:
a knowledge extraction engine having,
a customizable chained series of machine learning modeling structures, wherein each machine learning modeling structure of the chained series of machine learning modeling structures passes an output to a next machine learning modeling structure of the chained series of machine learning modeling structures to provide an automated pipeline for data processing of complex data structures, and
a hidden pattern detection unit configured to trigger automated under-writing processes;
a data interface configured to assess a plurality of digital risk-transfer policies; and
a persistence repository unit configured to store the plurality of digital risk-transfer policies, wherein each of the plurality of digital risk-transfer policies at least includes premium parameter values, and
wherein the chained series of machine learning modeling structures are configured to automatically assess and parse a digital risk-transfer policy of the plurality of digital risk-transfer policies, and automatically translate contractual language of the digital risk-transfer policy into a set of actions.
2 . The system according to claim 1 , wherein, for the assessing and the parsing, the chained series of machine learning modeling structures include one or more parser structures configured to select one of the plurality of digital risk-transfer policies of the persistence repository unit and parse assessable characters, words, and string of words of the selected one of the plurality of digital risk-transfer policies into digital constituents by providing and storing a parse tree that at least includes location attributes of every character in the selected one of the plurality of digital risk-transfer policies.
3 . The system according to claim 2 , further comprising a learner component configured to provide an open extractor process with a self-supervised learning of semantic relations during processing of the plurality of digital risk-transfer policies stored in the persistence repository unit.
4 . The system according to claim 3 , wherein the learner component comprises a pattern learner configured to classify whether a shortest dependency path between two strings of words or pattern of words indicate a semantic relation.
5 . The system according to claim 3 , further comprising a matcher configured to construct training and/or labeling data for the learner component by heuristically matching attribute-value pairs from the plurality of digital risk-transfer policies containing the assessable characters, words, and string of words.
6 . The system according to claim 5 , wherein the matcher is configured to seek a unique pattern to match an attribute value, and to produce a best training set, the matcher is configured to perform at least:
(i) skipping an attribute value completely when multiple parts of a digital policy match the attribute value or an equivalent attribute,
(ii) rejecting a pattern when the attribute value is not heads of the phrases containing them, and
(iii) discarding a pattern when the attribute value does not appear in the same clause or in a parent/child clause in the parse tree.
7 . The system according to claim 2 , wherein the chained series of machine learning modeling structures include one or more recover structures configured to recover a structure of the selected digital one of the digital risk-transfer policies by grouping the words into coherent units of text at least including document artefacts or characteristics based on an output of the one or more parser structures.
8 . The system according to claim 7 , wherein the document artefacts comprise at least one of the group comprising: section headings, paragraphs, and tables.
9 . The system according to claim 1 , wherein the chained series of machine learning modeling structures include one or more identifier structures configured to identify elements of the selected one of the digital risk-transfer policies containing language elements defining condition parameters indicating offers providable to a policyholder of the selected one of the digital risk-transfer policies based on a signaling output of the one or more identifier structures.
10 . The system according to claim 9 , wherein the chained series of machine learning modeling structures includes one or more linker structures configured to translate and map the identified elements to one or more parameterized queries which are executed against a standardized database of customer data.
11 . The system according to claim 1 , wherein the digital risk-transfer policies include life risk-transfer structures at least including term life risk-transfer structures and/or health risk-transfer structures.
12 . The system according to claim 11 , wherein the digital risk-transfer policy is associated with a policyholder, and wherein,
the policyholder, as a risk-exposed individual captured by a cover provided by one of the life risk-transfer structures, is associated with wearables and/or is associable with a measuring parameter,
measuring parameter values of the wearables are assessed via the data interface and stored by the persistence repository unit, and
the chained series of machine learning modeling structures is configured to automatically assess and process said measuring parameter values to automatically translate the processed data into the set of actions.
13 . The system according to claim 12 , wherein the policy holder is associated with at least one of the group comprising: bodily and environmental sensory devices, wearable telematics sensory devices, a measured laboratory parameter, and a clinical measuring parameter.
14 . The system according to claim 1 , wherein each of the digital risk-transfer policies includes at least one of the group comprising: deducible parameter values, risk-transfer type definition parameter values, policy limits parameter values, exclusion parameter values, and riders/addit parameter values.
15 . A digital, parameter pattern-driven method for a data mining system with a knowledge extraction engine based on a customizable chained series of machine learning structures providing an automated pipeline for data processing of complex data structures with a hidden pattern detection for triggering automated underwriting processes, the method comprising:
assessing a plurality of digital risk-transfer policies via a data interface;
storing the digital risk-transfer policies in a persistence repository unit of the data mining system, each of the digital risk-transfer policies at least including premium parameter values;
automatically assessing and parsing a digital risk-transfer policy of the plurality of digital risk-transfer policies by the chained series of machine learning modeling structures, wherein each machine learning modeling structure of the chained series of machine learning modeling structures passes an output to a next machine learning modeling structure of the chained series of machine learning modeling structures; and
automatically translating contractual language of the digital risk-transfer policy into a set of actions to be executed.
16 . The method according to claim 15 , wherein, for the assessing and the parsing, the chained series of machine learning modeling structures comprise one or more parser structures for selecting one of the digital risk-transfer policies of the persistence repository unit and parsing assessable characters, words, and string of words of the selected one of the digital risk-transfer policies into digital constituents by providing and storing a parse tree that at least includes location attributes of every character in the selected one of the digital risk-transfer policies.
17 . The method according to claim 16 , wherein the chained series of machine learning modeling structures include one or more recover structures recovering a structure of the selected one of the digital risk-transfer policies by grouping the words into coherent units of text at least including document artefacts or characteristics based on an output of the one or more parser structures.
18 . The method according to claim 15 , wherein the chained series of machine learning modeling structures include one or more identifier structures identifying elements of the selected one of the digital risk-transfer policies containing language elements defining condition parameters indicating offers providable to a policy holder of the selected one of the digital risk-transfer policies based on the output of the one or more identifier structures.
19 . The method according to claim 18 , wherein the chained series of machine learning modeling structures include one or more linker structures translating and mapping the identified elements to one or more parameterized queries which are executed against a standardized database of customer data.
20 . The digital, parameter pattern-driven method according to claim 15 , wherein each of the digital risk-transfer policies includes at least one of the group comprising: deducible parameter values, risk-transfer type definition parameter values, policy limits parameter values, exclusion parameter values, and riders/addit parameter values.