We present a novel commentator system that learns language from sportscasts of simulated soccer games. The system learns to parse and generate commentaries without any engineered knowledge about the English language. Training is done using only ambiguous supervision in the form of textual human commentaries and simulation states of the soccer games. The system simultaneously tries to establish correspondences between the commentaries and the simulation states as well as build a translation model. We also present a novel algorithm, Iterative Generation Strategy Learning (IGSL), for deciding which events to comment on. Human evaluations of the generated commentaries indicate they are of reasonable quality compared to human commentaries.

Topical application of nucleic acids offers many potential therapeutic advantages for suppressing genes in the skin, and potentially for systemic gene delivery. However, the epidermal barrier typically precludes entry of gene-suppressing therapy unless the barrier is disrupted. We now show that spherical nucleic acid nanoparticle conjugates (SNA-NCs), gold cores surrounded by a dense shell of highly oriented, covalently immobilized siRNA, freely penetrate almost 100% of keratinocytes in vitro, mouse skin, and human epidermis within hours after application. Significantly, these structures can be delivered in a commercial moisturizer or phosphate-buffered saline, and do not require barrier disruption or transfection agents, such as liposomes, peptides, or viruses. SNANCs targeting epidermal growth factor receptor (EGFR), an important gene for epidermal homeostasis, are >100-fold more potent and suppress longer than siRNA delivered with commercial lipid agents in cultured keratinocytes. Topical delivery of 1.5 uM EGFR siRNA (50 nM SNA-NCs) for 3 wk to hairless mouse skin almost completely abolishes EGFR expression, suppresses downstream ERK phosphorylation, and reduces epidermal thickness by almost 40%. Similarly, EGFR mRNA in human skin equivalents is reduced by 52% after 60 h of treatment with 25 nM EGFR SNA-NCs. Treated skin shows no clinical or histological evidence of toxicity. No cytokine activation in mouse blood or tissue samples is observed, and after 3 wk of topical skin treatment, the SNA structures are virtually undetectable in internal organs. SNA conjugates may be promising agents for personalized, topically delivered gene therapy of cutaneous tumors, skin inflammation, and dominant negative genetic skin disorders.T he recent development of small molecule inhibitors and antibodies that target components of signaling pathways has revolutionized the treatment of cancers, inflammatory diseases, and genetic disorders, including those that largely manifest in skin (1-3). These protein-based therapeutics, however, are costly, have limited targeting ability, and, when delivered through traditional intravenous or gastrointestinal routes, can lead to systemic toxicity (4, 5). Topical delivery is particularly attractive for the therapy of skin disorders. However, proteins larger than a few hundred daltons cannot easily enter the skin, and high concentrations of proteins must be applied for a cutaneous effect (6). An alternative to protein-based pathway inhibition involves the blocking and/or degradation of precursor mRNA before translation into protein. Gene silencing leads to down-regulation of protein expression and functions with greater specificity than inhibitors of protein function (7). In fact, targeted gene suppression by antisense DNA and siRNA has shown promising preclinical results, and/or is currently in clinical trials for a variety of diseases, including many forms of cancer (e.g., melanoma, neuroblastoma, and pancreatic adenocarcinoma), genetic disorders, and macular degeneration (8). For gene su...

We present a novel framework for learning to interpret and generate language using only perceptual context as supervision. We demonstrate its capabilities by developing a system that learns to sportscast simulated robot soccer games in both English and Korean without any language-specific prior knowledge. Training employs only ambiguous supervision consisting of a stream of descriptive textual comments and a sequence of events extracted from the simulation trace. The system simultaneously establishes correspondences between individual comments and the events that they describe while building a translation model that supports both parsing and generation. We also present a novel algorithm for learning which events are worth describing. Human evaluations of the generated commentaries indicate they are of reasonable quality and in some cases even on par with those produced by humans for our limited domain.

A wealth of pharmacological and behavioral data suggests that spinally projecting serotonergic cells mediate opioid analgesia. A population of medullary neurons, located within raphe magnus (RM) and the neighboring reticular nuclei, contains serotonin and is the source of serotonin in the spinal dorsal horn. To test whether serotonergic neurons mediate opioid analgesia, morphine was administered during recordings from medullary cells that were physiologically characterized as serotonergic (5HTp) by their slow and steady discharge pattern in the lightly anesthetized rat. Selected 5HTp cells (n = 14) were intracellularly labeled, and all contained serotonin immunoreactivity. The discharge of most 5HTp cells was not affected by an analgesic dose of systemic morphine. In a minority of cases, 5HTp cells either increased or decreased their discharge after morphine administration. However, morphine altered the discharge of some 5HTp cells in the absence of producing analgesia and conversely did not alter the discharge of most 5HTp cells in cases in which analgesia occurred. RM cells with irregular discharge patterns and excitatory or inhibitory responses to noxious tail heat were classified as ON and OFF cells, respectively. All ON and OFF cells that were intracellularly labeled (n = 9) lacked serotonin immunoreactivity. All ON cells were inhibited, and most OFF cells were excited by systemic morphine. Because 5HTp cells do not consistently change their discharge during morphine analgesia, they are unlikely to mediate the analgesic effects of morphine. Instead, nonserotonergic cells are likely to mediate morphine analgesia in the anesthetized rat. In light of the sensitivity of morphine analgesia to manipulations of serotonin, serotonin release, although neither necessary nor sufficient for opioid analgesia, is proposed to facilitate the analgesic effects of nonserotonergic RM terminals in the spinal cord.

No abstract