AUTOMATION FOR CONDUCTING INTERVIEWS
Interviewing, for generation of training data from skilled personnel, is automated in various ways. The training data is intended to train other ML tools to emulate workflow elicited from the skilled personnel. A trainee machine-learning (ML) tool can be provided representations of interviews and trained to perform a prediction function. The trained tool can be deployed to participate in conducting additional interviews of skilled personnel. The deployed tool can act as an independent interviewer or as a partner to other interviewers or evaluators, and can give feedback to other interviewers or receive feedback from evaluators. Interview representations can be annotated, e.g. with workflow maps, topic allocations, or commentary. Training of interviewers, evaluators, and annotators is disclosed, as also synthesis of interview representations.
1 . A method, comprising:
training a copilot to attain first and second capabilities, using a training corpus comprising respective representations of one or more interviews, wherein:
each interview comprises an alternating sequence of (i) prompts from one or more interviewers to one or more subjects and (ii) responses from the one or more subjects;
the first capability is to perform a prediction task;
at least a given interview of the one or more interviews is represented in the training corpus at least in part by a static or dynamic first workflow map derived from the given interview by an annotator, the first workflow map describing a job function performed by the respective one or more subjects; and
the second capability is to derive second workflow maps from representations of further interviews; and
deploying one or more instances of the trained copilot to participate in conducting additional interviews of skilled personnel.
2 . The method of claim 1 , further comprising:
by a given instance of the deployed instance(s) of the trained copilot, conducting the additional interviews.
3 . The method of claim 2 , further comprising:
by the given instance of the deployed instance(s) of the trained copilot, receiving feedback or suggestion from a partner monitoring at least one of the additional interviews.
4 . The method of claim 3 , wherein the feedback or suggestion is received after the at least one additional interview via reinforcement learning.
5 . The method of claim 1 , further comprising:
by a given instance of the deployed instance(s) of the trained copilot, suggesting prompts to a partner interviewer during at least one of the additional interviews.
6 . The method of claim 1 , wherein the copilot comprises a network of microservices including an expansion microservice, a retrieval microservice, one or more core microservices, and one or more evaluation microservices.
7 . The method of claim 1 , wherein the copilot comprises at least one large language model (LLM) or at least one deep neural network (DNN).
8 . The method of claim 1 , wherein the given interview is a first given interview, and wherein at least a second given interview of the one or more interviews is represented at least in part by a script in the training corpus.
9 . The method of claim 8 , wherein the script comprises a text document authored by a human as a creative act, distinct from being a transcription, and without the second interview taking, or having taken, place.
10 . The method of claim 8 , wherein the script comprises a transcript of the second given interview, and the method further comprises:
extracting the transcript from an audio or video representation of the second given interview.
11 . The method of claim 1 , wherein the prediction task is to predict a next prompt.
12 . The method of claim 1 , wherein the training is performed in multiple phases comprising three or more phases targeting progressively narrower knowledge domains.
13 . The method of claim 1 , wherein the one or more subjects comprise at least one first subject having a first job description and at least one second subject having a second job description distinct from but related to the first job description.
14 . (canceled)
15 . The method of claim 14 , wherein the copilot comprises a core microservice incorporating a long short-term memory (LSTM).
16 . The method of claim 1 , wherein the skilled personnel are interviewees in the additional interviews, the trained copilot is a first copilot, and the method further comprises:
generating training data from the additional interviews;
using the training data to train a second copilot to emulate workflow of at least one of the skilled personnel; and
causing the trained second copilot to be deployed to perform at least part of the emulated workflow.
17 . The method of claim 16 , wherein the generated training data comprises workflow maps generated by the trained first copilot for the additional interviews.
18 . A system, comprising:
one or more hardware processors with memory coupled thereto; and
computer-readable media storing instructions executable by one or more hardware processors, the instructions comprising:
a training module which, upon execution, is configured to train a copilot to attain first and second capabilities using a training corpus comprising respective representations of one or more interviews, wherein:
each interview comprises an alternating sequence of (i) prompts from one or more interviewers to one or more subjects and (ii) responses from the one or more subjects;
the first capability is to perform a prediction task;
at least a given interview of the one or more interviews is represented in the training corpus at least in part by a static or dynamic first workflow map derived from the given interview by an annotator, the first workflow map describing a job function performed by the respective one or more subjects; and
the second capability is to derive second workflow maps from representations of further interviews; and
a run-time module in the trained copilot which, upon execution in a deployed instance of the trained copilot, is configured to participate in conducting additional interviews of skilled personnel.
19 . The system of claim 18 , wherein the run-time module is configured to conduct the additional interviews.
20 . The system of claim 18 , wherein the run-time module is configured to monitor the additional interviews, and provide feedback or suggestions to one or more interviewers conducting the additional interviews.
21 . The system of claim 18 , wherein the copilot comprises a network of microservices including an expansion microservice, a retrieval microservice, one or more core microservices, and one or more evaluation microservices.