Making neural content meaningful and truthful
Boris Galitsky, Author of 150+ publications, 50+ patents and 6 books
Abstract: A text obtained by a deep learning (DL) content generation system (raw text) usually has major issues in terms of randomness and incorrectness. We build a content improvement system specifically oriented towards repairing these errors by finding correct and consistent sentences from various sources and substituting problematic entities, phrases, and sentences in raw content with the correct ones. We use text mining to identify correct corresponding sentences and the syntactic and semantic generalization procedure adopted to the content improvement task. We observed that raw content produced by a DL system like GPT-3 can be substantially improved for factual correctness and meaningfulness
Bio
Boris Galitsky contributed linguistic and machine learning technologies to Silicon Valley startups as well as companies like eBay and Oracle for over 25 years. Boris’ information extraction and sentiment analysis techniques assisted a number of acquisitions, such as Xoopit by Yahoo, Uptake by Groupon, Loglogic by Tibco and Zvents by eBay. His security-related technologies of document analysis contributed to acquisition of Elastica by Semantec. https://github.com/bgalitsky/relevanc...
As an architect of the Intelligent Bots project at Oracle, Boris developed a discourse analysis technique user for dialogue management and published in the book "Developing Enterprise Chatbots”. He also published a two-volume monograph “AI for CRM”, based on his experience developing Oracle Digital Assistant. Boris is Apache committer to OpenNLP where he created OpenNLP.Similarity component which is a basis for a semantically-enriched search engine and chatbot development.
Galitsky’s exploration and formalization of human seasoning culminated in the book “Computational Autism” broadly used by parents of children with autistic reasoning and rehabilitation personnel. Boris focus on medical domain led to another research monograph, “AI for Health Applications and Management”. https://www.amazon.com/Books-Boris-Ga...
An Author of 150+ publications, 50+ patents and 6 books, Boris’s focus now is on improving content generation quality.
0:00 Introduction
2:24 Presentation
4:05 Overview
6:24 Objective
7:43 Motivations 2
8:40 What is "content writing" and what transformers do
10:55 Applications of GPT-3
11:33 What GPT-3 wrote…
12:23 Contributions
12:34 Correction of Entity Values
17:24 Correcting the lies
22:06 Raw and True content in health
29:56 Fact-checking: substituting raw entities with true entities
30:30 Fact-checking, correction and extension of the raw content
32:45 Personalized drug recommendation
36:49 Fact checking via web mining
39:49 Coordination between a discourse tree for raw and true texts
46:52 System architecture
48:31 COMPETITION: OPEN AI recent paper
54:56 Evaluation of entity value correction in health domain
56:01 Correction characteristics
57:09 The error rate of repaired generated content
58:41 A manual assessment of error rate and ablation study
59:57 Discourse Incoherence
1:00:23 Discourse Tree Structure
1:00:31 Discourse Incoherence in generated texts
1:01:26 Results for learning-based discourse improvement
1:01:51 OpenNLP.Similarity.Chatbot
1:02:47 Conclusions
1:03:50 Q&A