Semantic programming by example with pre-trained models

The ability to learn programs from few examples is a powerful technology with disruptive applications in many domains, as it allows users to automate repetitive tasks in an intuitive way. Existing frameworks on inductive synthesis only perform syntactic manipulations, where they rely on the syntactic structure of the given examples and not their meaning. Any semantic manipulations, such as transforming dates, have to be manually encoded by the designer of the inductive programming framework. Recent advances in large language models have shown these models to be very adept at performing semantic transformations of its input by simply providing a few examples of the task at hand. When it comes to syntactic transformations, however, these models are limited in their expressive power. In this paper, we propose a novel framework for integrating inductive synthesis with few-shot learning language models to combine the strength of these two popular technologies. In particular, the inductive synthesis is tasked with breaking down the problem in smaller subproblems, among which those that cannot be solved syntactically are passed to the language model. We formalize three semantic operators that can be integrated with inductive synthesizers. To minimize invoking expensive semantic operators during learning, we introduce a novel deferred query execution algorithm that considers the operators to be oracles during learning. We evaluate our approach in the domain of string transformations: the combination methodology can automate tasks that cannot be handled using either technologies by themselves. Finally, we demonstrate the generality of our approach via a case study in the domain of string profiling.

The Diverse Landscape of Negative Polarity Items: On the Use of German NPIs as Experimental Diagnostics

The goal of this study is to provide better empirical insight into the licensing conditions of a large set of NPIs in German so that they can be used as reliable diagnostics in future research on negation-related phenomena. Experiment 1 tests the acceptability of 60 NPIs under semantic operators that are expected to license superstrong, strong, weak, and nonveridicality-licensed NPIs, respectively: antimorphic (not), anti-additive (no), downward entailing (hardly), nonveridical (maybe, question). Controls were positive assertions. Cluster analysis revealed seven clusters of NPIs, some of which confirm the licensing categorization from the literature (superstrong and weak NPIs). Other clusters show unclear patterns (overall high or medium ratings) and require further scrutiny in future research. One cluster showed high acceptability ratings only with the antimorphic and the question operator. Experiment 2 tested whether the source of this unexpected distribution was a rhetorical interpretation of the questions. Results suggest that rhetoricity was not the sole source. Overall, the results show gradual rather than categorical differences in acceptability, with higher acceptability corresponding to stronger negativity. The paper provides the detailed results for the individual NPIs as a preliminary normed acceptability index.

Building a Document-Oriented Warehouse Using NoSQL

The traditional data warehousing approaches should adapt to take into consideration novel needs and data structures. In this context, NoSQL technology is progressively gaining a place in the research and industry domains. This paper proposes an approach for building a NoSQL document-oriented warehouse (DocW). This approach has two methods, namely 1) document warehouse builder and 2) NoSQL-Converter. The first method generates the DocW schema as a galaxy model whereas the second one translates the generated galaxy into a document-oriented NoSQL model. This relies on two types of rules: structure and hierarchical rules. Furthermore, in order to help understanding the textual results of analytical queries on the NoSQL-DocW, the authors define two semantic operators S-Drill-Up and S-Drill-Down to aggregate/expand the terms of query. The implementation of our proposals uses MangoDB and Talend. The experiment uses the medical collection Clef-2007 and two metrics called write request latency and read request latency to evaluate respectively the loading time and the response time to queries.

The depth charge illusion, linear order, and “negated worlds” – Competence, performance, or both?

Compositionally incongruous sentences such as "No detail is too unimportant to be left out" are often assigned plausible non-compositional meanings (~ "Don’t leave out details"). The present research shows that moving the incongruous degree phrase to the beginning of the sentence ("Too unimportant to be left out is surely no detail") results in an attenuation of this semantic illusion, implying a role for incremental processing. Moreover, semantic operators cannot be readily applied to the generated propositions, and acceptability varies greatly between sentences and speakers, which is consistent with superficial interpretation. Implications of the present work with regard to the distinction between linguistic competence and performance are discussed, along with future directions for research.

Scalar implicatures in non-monotonic environments

Scalar implicatures (SIs) have been traditionally analyzed as pragmatic inferences that arise after semantic computation. Recent studies, however, have presented various challenges to this classic analysis; for instance, it has been claimed that SIs can be interpreted within the scope of various semantic operators. These observations motivate a grammatical analysis of SIs: SIs are derived during or before semantic computation. Among various kinds of evidence for the grammati- cal approach to SIs, especially convincing one is SIs embedded in non-monotonic (NM) environments, which post-semantic analyses have difficulty deriving. This paper introduces a new example of NM operators that allow the SI-embedding in their scope, and I thereby provide further empirical support for the grammatical analysis. On the other hand, I will also show that not all NM operators behave in the same way with respect to SI-embedding. It will turn out that Strawson non- monotonicity is required to embed SIs in the negative component of NM envi- ronments; i.e. SI-embedding is unavailable if the NMity can be decomposed into monotonic assertion and monotonic presupposition.