| CPC G06Q 30/018 (2013.01) [G06F 40/205 (2020.01); G06N 3/006 (2013.01); G06N 5/04 (2013.01); G06N 20/00 (2019.01); G06Q 30/016 (2013.01); G06Q 50/18 (2013.01); G06Q 20/127 (2013.01)] | 3 Claims |
|
1. A system for automatically interviewing and evaluating prospective clients for engagement with a law firm, the system comprising:
a virtual agent interface coupled via a network interface to be accessible to prospective clients of a law firm, wherein the virtual agent interface includes a natural language processor configured to parse input messages received from the prospective clients via natural language inputs and to extract key words and/or strings from said natural language inputs; and
a server running a predictive case engine, said predictive case engine communicatively coupled to the virtual agent interface to receive the key words and/or strings extracted from the input messages by the virtual agent interface, and to present interrogatories to the prospective clients via the virtual agent interface, the virtual agent interface thus configured as a front end for the predictive case engine to engage the prospective clients of the law firm in a conversation to elicit facts concerning potential causes of action, wherein the predictive case engine comprises:
a case matching module configured to:
identify potential causes of action from the key words and/or strings extracted by the virtual agent interface by using said key words and/or strings as indices to data structures that associate causes of action with the key words and/or strings, and
upon identifying a specific potential cause of action, engage the prospective clients in a virtual conversation via the virtual agent interface to elicit facts concerning the specific potential cause of action; and
one or more machine learning modules configured to:
identify factual distinctions associated with the potential causes of action that affect outcomes in judicial decisions and settlements by accessing one or more databases storing prior case information, and generate, using a convolutional neural network, case fact pattern vector representations from the prior case information by pooling results of convolution operations on n-dimension vectors that represent extracted facts from individual cases included in the prior case information, and
use identified factual distinctions to train the case matching module to engage the prospective clients with specific questions when the potential causes of action are recognized from the key words and/or strings extracted from the input messages,
a case scoring module to score the potential causes of action by evaluating descriptions provided by the prospective clients through the virtual conversation against stored case information in the one or more databases, each respective evaluation producing an outcome with an associated confidence level and those outcomes rising above a predetermined threshold being deemed to be an engageable prospect,
said predictive case engine thus configured to evaluate said facts for assessing engagement of the prospective clients by the law firm using scorings assigned according to stored information concerning prior judicial decisions and settlements.
|