Abstract. In this paper we propose a framework for recognition and retrieval tasks in the context of scene text images. In contrast to many of the recent works, we focus on the case where an image-specific list of words, known as the small lexicon setting, is unavailable. We present a conditional random field model defined on potential character locations and the interactions between them. Observing that the interaction potentials computed in the large lexicon setting are less effective than in the case of a small lexicon, we propose an iterative method, which alternates between finding the most likely solution and refining the interaction potentials. We evaluate our method on public datasets and show that it improves over baseline and state-of-the-art approaches. For example, we obtain nearly 15% improvement in recognition accuracy and precision for our retrieval task over baseline methods on the IIIT-5K word dataset, with a large lexicon containing 0.5 million words.

Abstract-In this paper, we present a solution towards building a retrieval system over handwritten document images that i) is recognition-free, ii) allows text-querying, iii) can retrieve at subword level, iv) can search for out-of-vocabulary words. Unlike previous approaches that operate at either character or word levels, we use character n-gram images (CNG-img) as the retrieval primitive. CNG-img are sequences of character segments, that are represented and matched in the image-space. The word-images are now treated as a bag-of-CNG-img, that can be indexed and matched in the feature space. This allows for recognition-free search (query-by-example), which can retrieve morphologically similar words that have matching sub-words. Further, to enable query-by-keyword, we build an automated scheme to generate labeled exemplars for characters and character n-grams, from unconstrained handwritten documents. We pose this problem as one of weakly-supervised learning, where character/n-gram labeling is obtained automatically from the word labels. The resulting retrieval system can answer queries from an unlimited vocabulary. The approach is demonstrated on the George Washington collection, results show major improvement in retrieval performance as compared to word-recognition and word-spotting methods.

No abstract

Background: Cannabis is purported to provide benefits to cancer patients, however studies supporting efficacy are limited, and use may be associated with increased risk for immunocompromised patients. In Washington (WA) State, medical cannabis was legalized in 1998, and became available for recreational use in July 2014. We set out to assess cannabis use among hematopoietic cell transplant (HCT) recipients during time periods pre-and post-recreational cannabis legalization. Methods: We conducted a retrospective chart review of reported cannabis use among two randomly selected cohorts of 100 HCT recipients at the Fred Hutchinson Cancer Research Center. Both cohorts were selected to include an even split between autologous and allogeneic recipients. The first were transplanted in 2010 when only medicinal cannabis was legal, and the second in 2016 post-recreational legalization. Data were collected using center databases and medical record review. Active users were defined as those noted by the clinical team to have had actively using cannabis during either the immediate pre-(≤1 month) or post-HCT (≤100 days) period, and past users as those with only distant use. Results: HCT recipients in 2016 were older ([median 53 yrs, interquartile range (IQR) 44, 62 vs. [58, IQR 49, 67] P = .03), but otherwise similar to those in 2010. Clinical notes documented that nearly all patients were asked about general drug use (98% vs. 97%, respectively), but documentation targeted to cannabis was more frequent in the latter cohort (33% vs. 96%, P < .001); transplant and social worker teams most frequently captured cannabis-specific histories. More patients were identified as active users in 2016 (5% vs. 18%, P = .008), but when comparing only those with documented cannabis histories, the frequency of past (14/33 vs. 25/96, P = .08) and active users (5/33 vs. 18/96, P = .79) was similar between the cohorts. All active users were identified early, and after counseling (21/23 [91%]) stopped using cannabis pre-HCT. Dronabinol use was more frequent in 2016 (7% vs. 16%, P = .07), especially in active cannabis users (9/23 vs. 14/177, P < .001). Conclusions: An increased number of HCT recipients were documented to be active cannabis users during the period post-recreational legalization in WA. These data suggest an increase in prevalence since recreational legalization, but also that this shift may be at least in part due to heightened provider awareness. Regardless, targeted questions about cannabis use can help to identify patients at risk for adverse events, shift use patterns and allow them to benefit from counseling.