The PUNDIT natural-language processing system

Content-Based Image Retrieval by a Fuzzy Scale-Space Approach

“…Figure 4 shows the results of the aforementioned experiment. We consider three different values for w (9,11,13) for the method SS1, while more different values were required for the method SS2 (3,5,7,9,11,13,15,17,19,21,23,25 of cooccurrence matrices, while the adopted measures were uniformity, entropy, contrast, IDM and correlation. Concerning the Gabor features, we adopted the same values (and the same implementation accordingly) fixed in the GIFT software (i.e.…”

“…13,48 At first, the automation was restricted to the processing of textual data. 23,25 When the technological advances allowed the transmission and the storage of other data types like audio, images or video, the need to automate the processing of these data arose again, leading to the so-called Content Based Retrieval (CBR) systems, which became an active research area in the recent years. 52 The interest aroused by these systems can be seen, for example, in the diverse application areas where CBR systems have been used, such as art catalog databases, 14 medical assistance systems 6,37 or Web search robots.…”

A process for predicting manhole events in Manhattan

“…Relatively early work (Hirschman et al 1989) showed it was possible to handle fragmentary text using full syntactic and semantic parsing, but involved a much smaller dataset than current work on ticket databases. Devaney and Ram (2005) dealt with 10,000 logs, all of which pertain to the same machines.…”

“…Mining the Remarks for useful information presented serious obstacles. Because the ECS tickets are created by people working under time pressure, they have the fragmen- 13 ********************** PSC COMPLETED ******************** RV 14 ****************** CHANGE IN PLAN SENT ON '21' ********* ********************************************************** tary language and telegraphic features found to be typical of trouble reports (Linebarger et al 1988;Hirschman et al 1989). They contain a very wide range of categories of information, including repair work ("CUT,CLEARED,P/O RETIED DEFECTIVE SER-VICE LEG & RESTORED FULL POWER"), parking restrictions in the relevant area of the structure ("-NO PARKING 08:00 TO 18:00 MON TO FRI-"), communications with customers ("HAVE A BAD LANGUAGE PROBLEM...WITH CARETAKER...") and other relevant information ("FOUND EXTENSION CORD TIED TO PIGTAIL AT BASE OF LAMP").…”

“…Trouble tickets , maintenance logs (Devaney and Ram 2005), equipment reports (Hirschman et al 1989), safety reports (Oza et al 2009) and similar sets of documents have been handled using the methods of natural language processing, knowledge modeling, text classification and machine learning for a wide range of goals. Relatively early work (Hirschman et al 1989) showed it was possible to handle fragmentary text using full syntactic and semantic parsing, but involved a much smaller dataset than current work on ticket databases.…”