Abstract: Can we learn dialog structure from existing dialogs without ontology or domain assumptions. Understanding dialog structures from existing task oriented human human dialogs can help us automate these dialogues in a better way. Traditionally dialog structures have been created using ontologies that are created by domain experts. However, in our experience getting the ontology right is difficult and time consuming. Like other such tasks an unsupervised approach may do better than hand crafted rules. We propose an unsupervised dialog structure discovery approach that is based on SCAN (Semantic Clustering using Nearest Neighbors). Our approach comprises of two steps, the first being creating clusters of utterances and the second being creation of a structure using inter-cluster transition probabilities. Our main contribution in this paper is the adaptation of SCAN on text data. Unlike the SCAN approach for images, for text we did not train a separate pretext model and were able to use BERT for the same. Similarly for neigbor discovery, instead of augmentation we were able to leverage data variety. Evaluation metrics on dialog structures are a bit subjective, so we have used statistical measures as proxies for structure quality. We have also included our results on an internal human human task oriented 100k dialog dataset. We think SCAN like approaches are very promising for problems that use embedding similarities and should be further explored.Abstract: Can we learn dialog structure from existing dialogs without ontology or domain assumptions. Understanding dialog structures from existing task oriented human human dialogs can help us automate these dialogues in a better way. Traditionally dialog structures have been created using ontologies that are created by domain experts. However, in our exper...Show More
Abstract: Solidification cracking is a significant problem during the welding of fully austenitic stainless steels. The present work is considered as the first trial to investigate and propose a mechanism of hot cracking formation when welding the Fan-shaped cracking test specimen, using the pulsed current gas tungsten arc welding process (PCGTAW). The specimen lateral expansions perpendicular to welding line due to thermal effects, plus the transverse expansion due to crack opening are sensed and recorded to detect the crack behavior with time. The stages of crack formation are filmed by a high-speed photography of the weld pool and solidification process at a speed of about 1000 fps. Additionally, some microscopic examinations using Scanning Electron Microscope (SEM) and Electron Probe Micro-Analyzer (EPMA) are performed on the welds. The results helped in establishing a proposed mechanism for the formation of hot cracks in full–austenitic stainless steel welds done on a Fan-shaped test specimen. The proposed mechanism suggests three stages during hot cracking formation; the crack initiation, propagation, and ceasing. The occurrence of a hot crack during welding mainly depends on the way by which the molten zone solidifies, and which solid phase will primarily solidify. This affects, in turn, the segregation of the chemical elements, which found to have a great role in crack initiation. Moreover, the weld metal structure type, together with the thermal stresses in conjunction with the applied strains on the weld joint play a great role in the crack expansion and ceasing. The present work is considered the first trial done to propose a mechanism of hot-cracking formation during welding the Fan-Shaped test specimen using Pulsed-Current Gas Tungsten Arc welding process.Abstract: Solidification cracking is a significant problem during the welding of fully austenitic stainless steels. The present work is considered as the first trial to investigate and propose a mechanism of hot cracking formation when welding the Fan-shaped cracking test specimen, using the pulsed current gas tungsten arc welding process (PCGTAW). The speci...Show More